KR102115645B1 - Systems and methods for multi-user multi-lingual communications - Google Patents

Abstract

The various embodiments described herein enable multilingual communication. The systems and methods of some embodiments are multi-language, through different communication modes including, for example, Internet-based chat, email, text-based mobile phone communication, posting to online forums, posting to online social media services, and the like. Communication can be enabled. Some embodiments implement a communication system and method for translating text between two or more languages. Users of the system and method can be encouraged to submit corrections for inaccurate or erroneous translations, and receive compensation for these submissions. Systems and methods for evaluating the accuracy of translations are described.

Description

Multi-user multi-language communication system and method {SYSTEMS AND METHODS FOR MULTI-USER MULTI-LINGUAL COMMUNICATIONS}

This application claims the priority and interests of United States Patent Application Serial No. 13 / 908,979, filed on June 3, 2013, the entire content of which is incorporated herein by reference.

The present invention generally relates to language translation, and more particularly, to a language translation including a plurality of users and a plurality of languages.

Before machine-language translation (hereinafter referred to as “machine translation”), translation between two languages was possible only by intervention or interpretation of persons educated in both languages. On the other hand, a typical machine translator generally operates based on statistical / stochastic interpretation of context and grammar, and mainly human intervention / interpretation is not required.

Conventional machine interpretation is prone to errors, especially when the text to be translated has a minimal context. Text with minimal context is mainly found in conversations using simple sentence structures. Additionally, machine translation is commonly found in conversational text, abbreviations, acronyms, diminutives, colloquial words / phrases, proper nouns and common nouns. For (common noun), there are mainly problems.

Various embodiments described herein provide systems and methods related to multilingual communication between multiple users, which may be users of two or more client systems.

Modes of communication enabled by the embodiments include Internet-based chat (eg, Apple® iMessage, Windows® Live Messenger, etc.), email (eg, embedded forum messaging, Yahoo® mail, RFC 5322). Etc.), text-based mobile phone communication (e.g. SMS messages or MMS messages), posting to online forums (e.g. posting to web-based hobby forums) and online social media services ) (For example, Twitter®, Facebook®, etc.). For example, the systems and methods can implement a multi-language, multi-user chat system.

In some embodiments, the provided method includes: identifying a first language and a second language; Receiving an initial message in the first language from a first person communicating in the first language in a first chat client system; And querying a first correspondence message in the second language with the data storage unit, wherein the first correspondence message is based on the initial message in the first language. When the data storage unit includes the first correspondence message, the method may assist in the transmission of the correspondence message from the second chat client system to the second person communicating in the second language. According to the above embodiment, the initial message may include text, emoticons, ASCII-based technology, or other content applicable or customary to a human-readable message transmitted through a network. have. Additionally, the initial message may be part of a larger message (eg, the initial message is one sentence included in a multi-sentence message) communicated between chat client systems.

If the data storage unit does not include the first corresponding message, the method may perform an attempt to convert at least a portion of the initial message into a converted message in the first language. engine). Using the converted message, the method may query the data storage unit for a second corresponding message in the second language based on the converted message.

In certain embodiments, the system or method may attempt to transform the initial message using a series of transform operations before querying the data store for a second corresponding message based on the transformed message. . Alternatively, in some embodiments, the system or method may iteratively perform the transformation and query, where the initial message is transformed using a subset of the transformation operations available, and the data The storage unit is queried for a second corresponding message based on the converted message resulting from the result, and if the second corresponding message is not identified, conversion and other repetition of the query are performed (for example, as the result) The resulting transformed message can be further transformed using a subset of other possible transform operations, and the data storage can be queried for a second corresponding message based on the resulting transformed message) . In some of these embodiments, a subset of transform operations applied to each iteration may be applied to the initial message, or to a recently resulting transformed message.

Consequently, the method can support translating the initial message or the converted message into a corresponding message, in a second language. In some embodiments, if the first corresponding message to the initial message is not in the data storage and the transformation engine does not transform at least a portion of the initial message, the initial message is a corresponding message. Can be translated. Additionally, in various embodiments, the first corresponding message to the initial message is not in the data storage; The conversion engine causes a converted message, including conversion of at least a portion of the initial message; And when the data storage unit does not include a second corresponding message to the converted message, the converted message may be translated into the corresponding message.

According to the embodiment, converting a part of the initial message identifies a chatspeak word or phrase (eg, 'lol', 'gr8') included in the initial message, and the chat term Replace words or phrases with words or phrases other than chat terms, perform a spell check on parts of the initial message, or identify abbreviations of parts of the initial message, and assign the abbreviations to the abbreviations It may include substituting a corresponding (eg, representing) word or phrase (eg, 'CA' to 'California' or 'brb' to 'be right back').

Additionally, converting a portion of the initial message identifies an acronym of a portion of the initial message, and corresponds to the acronym (eg, 'USA') for the acronym (eg, 'USA'). , Indicating a word or phrase, or identifying a colloquial of a part of the initial message and replacing the colloquial language with a word or phrase corresponding to the colloquial language. Further, converting a portion of the initial message identifies a profanity word or phrase of a portion of the initial message, and replaces the profanity word or phrase with a non-profanity word or phrase (eg, the Slang words or phrases, and may include deleting the slang words or phrases from the initial message.

According to some embodiments, converting a portion of the initial message includes marking a portion of the initial message that is not to be translated. For example, if a specific part of the initial message includes a proper noun, a common noun, a diminutive, an abbreviation or an acronym, the method In the part, it can be indicated that it is not translated in a subsequent operation.

In one embodiment, an embodiment is provided that includes various configurations implemented to perform various operations described herein. Similarly, one embodiment of providing a computer program product may include computer instruction code implemented to enable a computer system to perform various operations described herein.

In one aspect, the invention relates to a computer-implemented method. The method comprises selecting a word or phrase from a data store associated with a translation failure that fails to translate a message containing the word or phrase from a first language to a second language; Selecting a user to solicit user feedback for the translation failure; Determining the value of an incentive to provide to the user in exchange for user feedback; Sending a request for feedback to the user's computing device, wherein the request includes an incentive; Receiving user feedback from the computing device, the user feedback comprising each word or phrase in a first or second language; Determining that the user feedback is approved; And based on the approval, depositing into the user's account according to the value of the incentive.

In certain embodiments, the failure is due to an actual failure in translating the message. The failure can be identified or attributed to the user marking the message as potentially incorrect. The step of selecting the user may be based on the user's reliability measure, a quote associated with the user, a previous deposit into the user's account, the user's preference, or the user's language ability. In some embodiments, the word or phrase includes chat terms in a first language. The response may include chat terms in a second language. In some implementations, the query includes fields configured to receive text-based values.

In a particular embodiment, the request includes a set of preselected definitions from which the user can select a definition for the word or phrase. The set of pre-selected definitions includes, for example, at least one definition provided by another user in response to another request, wherein the other request obtains previous user feedback for the word or phrase from the other user. May have been previously created. The other request may include another set of preselected definitions from which the other user has selected the definition. The method may further include evaluating the user feedback to determine the most popular response.

In various embodiments, the method further comprises determining the user's capabilities based on the user feedback. The method may further include updating the translation or translation of the word or phrase from a first language to a second language based on the user feedback. In some embodiments, determining that the user feedback is approved includes: determining that the user feedback is not fraudulent; And / or determining that the user feedback is correct. The step of determining that the user feedback is approved may be based on a comparison of the user feedback with at least one previous user feedback provided by another user in response to another request, the other request being from the other user. It may have been previously generated to get feedback on the word or phrase.

In certain embodiments, the incentive includes (or is an offer for) an in-game currency or an in-game item. The value of the incentive may be determined based on, for example, the importance of the word or phrase, or the complexity of the word or phrase. In some examples, determining the value of the incentive includes: (i) the complexity of the word or phrase, (ii) the importance of the word or phrase, (iii) the response method used by the user, (iv) the Considering the type of word or phrase, and / or (v) the language involved in the translation failure.

In another aspect, the invention relates to a system comprising one or more computers programmed to perform an operation. The operation includes selecting a word or phrase from a data storage unit associated with a translation failure that fails to translate a message containing a word or phrase from a first language to a second language; Selecting a user to solicit user feedback for the translation failure; Determining the value of an incentive to provide to the user in exchange for user feedback; Sending a request for feedback to the user's computing device, the request comprising an incentive; Receiving the user feedback from the computing device, the user feedback comprising each word or phrase in the first or second language; Determining that the user feedback is approved; And based on approval, depositing into the user's account.

In certain embodiments, the failure is due to an actual failure in translating the message. The failure can be identified or attributed to the user marking the message as potentially incorrect. The step of selecting the user may be based on the user's reliability measure, a quote associated with the user, a previous deposit into the user's account, the user's preference, or the user's language ability. In some embodiments, the word or phrase includes chat terms in a first language. The response may include chat terms in a second language. In some implementations, the query includes fields configured to receive text-based values.

In a particular embodiment, the request includes a set of preselected definitions from which the user can select a definition for the word or phrase. The set of pre-selected definitions includes, for example, at least one definition provided by another user in response to another request, wherein the other request obtains previous user feedback for the word or phrase from the other user. May have been previously created. The other request may include another set of preselected definitions from which the other user has selected the definition. The operation may further include evaluating the user feedback to determine the most popular response.

In various embodiments, the operation further comprises determining the capability of the user based on the user feedback. The operation may further include updating the translation or translation of the word or phrase from a first language to a second language based on the user feedback. In some embodiments, determining that the user feedback is approved includes: determining that the user feedback is not fraudulent; And / or determining that the user feedback is correct. The step of determining that the user feedback is approved may be based on a comparison of the user feedback with at least one previous user feedback provided by another user in response to another request, the other request being from the other user. It may have been previously generated to get feedback on the word or phrase.

In certain embodiments, the incentive includes (or is an offer for) an in-game currency or an in-game item. The value of the incentive may be determined based on, for example, the importance of the word or phrase, or the complexity of the word or phrase. In some examples, determining the value of the incentive includes: (i) the complexity of the word or phrase, (ii) the importance of the word or phrase, (iii) the response method used by the user, (iv) the Considering the type of word or phrase, and / or (v) the language involved in the translation failure.

In another aspect, the invention relates to a computer program product stored on one or more storage media for improving language translation through incentivized feedback. The computer program product is executed by a data processing device and causes the data processing device to perform an operation, wherein the operation fails to translate a message containing a word or phrase from a first language to a second language. Selecting the word or phrase associated with the data storage unit; Selecting a user to solicit user feedback for the translation failure; Determining the value of an incentive to provide to the user in exchange for user feedback; Sending a request for feedback to the user's computing device, the request comprising an incentive; Receiving the user feedback from the computing device, the user feedback comprising each word or phrase in the first or second language; Determining that the user feedback is approved; And based on approval, depositing into the user's account.

In certain embodiments, the failure is due to an actual failure in translating the message. The failure can be identified or attributed to the user marking the message as potentially incorrect. The step of selecting the user may be based on the user's reliability measure, a quote associated with the user, a previous deposit into the user's account, the user's preference, or the user's language ability. In some embodiments, the word or phrase includes chat terms in a first language. The response may include chat terms in a second language. In some implementations, the query includes fields configured to receive text-based values.

In a particular embodiment, the request includes a set of preselected definitions from which the user can select a definition for the word or phrase. The set of pre-selected definitions includes, for example, at least one definition provided by another user in response to another request, wherein the other request obtains previous user feedback for the word or phrase from the other user. May have been previously created. The other request may include another set of preselected definitions from which the other user has selected the definition. The operation may further include evaluating the user feedback to determine the most popular response.

In various embodiments, the operation further comprises determining the capability of the user based on the user feedback. The operation may further include updating the translation or translation of the word or phrase from a first language to a second language based on the user feedback. In some embodiments, determining that the user feedback is approved includes: determining that the user feedback is not fraudulent; And / or determining that the user feedback is correct. The step of determining that the user feedback is approved may be based on a comparison of the user feedback with at least one previous user feedback provided by another user in response to another request, the other request being from the other user. It may have been previously generated to get feedback on the word or phrase.

In certain embodiments, the incentive includes (or is an offer for) an in-game currency or an in-game item. The value of the incentive may be determined based on, for example, the importance of the word or phrase, or the complexity of the word or phrase. In some examples, determining the value of the incentive includes: (i) the complexity of the word or phrase, (ii) the importance of the word or phrase, (iii) the response method used by the user, (iv) the Considering the type of word or phrase, and / or (v) the language involved in the translation failure.

In one aspect, the invention relates to a method implemented by a data processing apparatus. The method includes providing a text message chat system to a plurality of users; Receiving an original text message in a first language from a first user; Generating an initial translation in the second language of the original text message; Providing the original text message and the initial translation to a second user; Receiving a translation correction that raises an error in the initial translation from the second user; And (a) identifying the most accurate translation correction from a plurality of translation corrections, wherein the plurality of translation corrections include translation corrections from the second user; And (b) evaluating the accuracy of translation correction from the second user using word-based features, language-based features, and / or word placement features.

In certain embodiments, the method includes providing incentives (eg, virtual goods and / or virtual currencies for use in online games) to encourage the second user to submit translation corrections. Determining the most accurate translation correction comprises: receiving at least one additional translation correction for raising an error in the initial translation from at least one additional user, wherein the at least one additional translation correction and the second The translation correction from the user comprises: receiving an additional translation correction defining the plurality of translation corrections; Receiving feedback from users about the accuracy of the plurality of translation corrections; And based on the feedback, identifying the most accurate translation correction from the plurality of translation corrections.

In some implementations, the method also includes providing a reward (eg, virtual goods and / or virtual currencies for use in online games) to the user who submitted the most accurate translation correction. The method may also include providing a reward (eg, virtual goods and / or virtual currencies for use in online games) to a user who provided feedback used to identify the most accurate translation. The word-based features may include, for example, word counts, character counts, emojis, numbers, and / or punctuation marks. Using the language-based feature may include identifying the part of speech in the original text message and the translation correction from the second user.

In some embodiments, the method also includes identifying the number of verbs in each of the original text message and the translation correction from the second user; And comparing the number of verbs in the original text message with the number of verbs in the translation correction from the second user. The absence of part-of-speech in the original text message and / or translation correction from the second user may indicate a language detection failure. The method may further include rejecting the translation correction from the second user when the translation correction from the second user is the same as the initial translation.

In another aspect, the invention relates to a system comprising a computer readable medium on which instructions are stored, and a data processing apparatus. The data processing apparatus is configured to execute instructions for performing an operation. The operation includes providing a text message chat system to a plurality of users; Receiving an original text message in a first language from a first user; Generating an initial translation in the second language of the original text message; Providing the original text message and the initial translation to a second user; Receiving a translation correction that raises an error in the initial translation from the second user; And (a) identifying the most accurate translation correction from a plurality of translation corrections, wherein the plurality of translation corrections include translation corrections from the second user; And (b) evaluating the accuracy of translation correction from the second user using word-based features, language-based features, and / or word placement features.

In certain embodiments, the operation includes providing incentives (eg, virtual goods and / or virtual currencies for use in online games) to encourage the second user to submit translation corrections. Determining the most accurate translation correction comprises: receiving at least one additional translation correction for raising an error in the initial translation from at least one additional user, wherein the at least one additional translation correction and the second The translation correction from the user comprises: receiving an additional translation correction defining the plurality of translation corrections; Receiving feedback from users about the accuracy of the plurality of translation corrections; And based on the feedback, identifying the most accurate translation correction from the plurality of translation corrections.

In some implementations, the operation also includes providing a reward (eg, virtual goods and / or virtual currencies for use in online games) to the user who submitted the most accurate translation correction. The operation may also include providing a reward (eg, virtual goods and / or virtual currencies for use in online games) to a user who provided feedback used to identify the most accurate translation. The word-based features may include, for example, word counts, character counts, emojis, numbers, and / or punctuation marks. Using the language-based feature may include identifying the part of speech in the original text message and the translation correction from the second user.

In some embodiments, the operation also includes identifying the number of verbs in each of the original text message and the translation correction from the second user; And comparing the number of verbs in the original text message with the number of verbs in the translation correction from the second user. The absence of part-of-speech in the original text message and / or translation correction from the second user may indicate a language detection failure. The operation may further include rejecting the translation correction from the second user when the translation correction from the second user is the same as the initial translation.

In another aspect, the invention relates to a computer program product stored on one or more storage media for controlling a processing mode of a data processing device. The computer program product is executable by a data processing device and causes the data processing device to perform an operation, the operation comprising: providing a text message chat system to a plurality of users; Receiving an original text message in a first language from a first user; Generating an initial translation in the second language of the original text message; Providing the original text message and the initial translation to a second user; Receiving a translation correction that raises an error in the initial translation from the second user; And (a) identifying the most accurate translation correction from a plurality of translation corrections, wherein the plurality of translation corrections include translation corrections from the second user; And (b) evaluating the accuracy of translation correction from the second user using word-based features, language-based features, and / or word placement features.

In certain embodiments, the operation includes providing incentives (eg, virtual goods and / or virtual currencies for use in online games) to encourage the second user to submit translation corrections. Determining the most accurate translation correction comprises: receiving at least one additional translation correction for raising an error in the initial translation from at least one additional user, wherein the at least one additional translation correction and the second The translation correction from the user comprises: receiving an additional translation correction defining the plurality of translation corrections; Receiving feedback from users about the accuracy of the plurality of translation corrections; And based on the feedback, identifying the most accurate translation correction from the plurality of translation corrections.

In some implementations, the operation also includes providing a reward (eg, virtual goods and / or virtual currencies for use in online games) to the user who submitted the most accurate translation correction. The operation may also include providing a reward (eg, virtual goods and / or virtual currencies for use in online games) to a user who provided feedback used to identify the most accurate translation. The word-based features may include, for example, word counts, character counts, emojis, numbers, and / or punctuation marks. Using the language-based feature may include identifying the part of speech in the original text message and the translation correction from the second user.

In some embodiments, the operation also includes identifying the number of verbs in each of the original text message and the translation correction from the second user; And comparing the number of verbs in the original text message with the number of verbs in the translation correction from the second user. The absence of part-of-speech in the original text message and / or translation correction from the second user may indicate a language detection failure. The operation may further include rejecting the translation correction from the second user when the translation correction from the second user is the same as the initial translation.

In one aspect, the invention relates to a method implemented by a data processing apparatus. The method includes identifying a first language and a second language; Receiving a chat language audible message in a first language from a first person in a first chat client system communicating in a first language; Changing the chat term audible message to a chat term text message in a first language; Converting the chat term text message into a clear term text message in a first language; Translating the clear term text message into a corresponding clear term text message in a second language; Converting the corresponding clear term text message into a corresponding chat term text message in a second language; Changing the corresponding chat term text message to a corresponding chat term audible message in a second language; And transmitting the corresponding chat term audible message to a second user of a second chat client system communicating in a second language.

In a specific embodiment, the step of changing the audible message of the chat term to a text message of the chat language in the first language includes providing the audible message of the chat term to the speech recognition system. The step of converting the chat term text message may include identifying a chat term word or phrase in the chat term text message; And replacing the chat term word or phrase with a non-chat term word or phrase. In some examples, the step of converting the corresponding chatspeak text message to a corresponding chatspeak audible message in a second language provides the corresponding chatspeak text message to a text-to-speech system It includes the steps.

In a particular implementation, the speech recognition system and / or the text-to-speech system utilizes data comprising a plurality of accents and dialects for each of the first and second languages. The data may include chat terms and clear term formats for each of the first and second languages. In various embodiments, the method includes receiving feedback from the user regarding the accuracy of the corresponding clear term text message, the corresponding chatspeak text message, and / or the corresponding chatspeak audible message. The method may further include providing an incentive (eg, virtual goods and / or virtual currencies, for use in online games) to the user for providing the feedback. In some examples, the method further comprises processing the feedback to improve the accuracy of the speech recognition system and / or the text-speech system.

In another aspect, the invention relates to a system comprising a computer readable medium on which instructions are stored, and a data processing apparatus. The data processing apparatus is configured to execute instructions for performing an operation. The operation includes: identifying a first language and a second language; Receiving a chat language audible message in a first language from a first person in a first chat client system communicating in a first language; Changing the chat term audible message to a chat term text message in a first language; Converting the chat term text message into a clear term text message in a first language; Translating the clear term text message into a corresponding clear term text message in a second language; Converting the corresponding clear term text message into a corresponding chat term text message in a second language; Changing the corresponding chat term text message to a corresponding chat term audible message in a second language; And transmitting the corresponding chat term audible message to a second user of a second chat client system communicating in a second language.

In a specific embodiment, the step of changing the audible message of the chat term to a text message of the chat language in the first language includes providing the audible message of the chat term to the speech recognition system. The step of converting the chat term text message may include identifying a chat term word or phrase in the chat term text message; And replacing the chat term word or phrase with a non-chat term word or phrase. In some examples, the step of converting the corresponding chatspeak text message to a corresponding chatspeak audible message in a second language provides the corresponding chatspeak text message to a text-to-speech system It includes the steps.

In a particular implementation, the speech recognition system and / or the text-to-speech system utilizes data comprising a plurality of accents and dialects for each of the first and second languages. The data may include chat terms and clear term formats for each of the first and second languages. In various embodiments, the operation includes receiving feedback from the user regarding the accuracy of the corresponding clear term text message, the corresponding chatspeak text message, and / or the corresponding chatspeak audible message. The operation may further include providing an incentive (eg, virtual goods and / or virtual currencies, for use in online games) to the user for providing the feedback. In some examples, the operation further includes processing the feedback to improve the accuracy of the speech recognition system and / or the text-speech system.

In another aspect, the invention relates to a computer program product stored on one or more storage media for controlling a processing mode of a data processing device. The computer program product is executable by a data processing device and causes the data processing device to perform an operation, the operation comprising: identifying a first language and a second language; Receiving a chat language audible message in a first language from a first person in a first chat client system communicating in a first language; Changing the chat term audible message to a chat term text message in a first language; Converting the chat term text message into a clear term text message in a first language; Translating the clear term text message into a corresponding clear term text message in a second language; Converting the corresponding clear term text message into a corresponding chat term text message in a second language; Changing the corresponding chat term text message to a corresponding chat term audible message in a second language; And transmitting the corresponding chat term audible message to a second user of a second chat client system communicating in a second language.

In a specific embodiment, the step of changing the audible message of the chat term to a text message of the chat language in the first language includes providing the audible message of the chat term to the speech recognition system. The step of converting the chat term text message may include identifying a chat term word or phrase in the chat term text message; And replacing the chat term word or phrase with a non-chat term word or phrase. In some examples, the step of converting the corresponding chatspeak text message to a corresponding chatspeak audible message in a second language provides the corresponding chatspeak text message to a text-to-speech system It includes the steps.

In a particular implementation, the speech recognition system and / or the text-to-speech system utilizes data comprising a plurality of accents and dialects for each of the first and second languages. The data may include chat terms and clear term formats for each of the first and second languages. In various embodiments, the operation includes receiving feedback from the user regarding the accuracy of the corresponding clear term text message, the corresponding chatspeak text message, and / or the corresponding chatspeak audible message. The operation may further include providing an incentive (eg, virtual goods and / or virtual currencies, for use in online games) to the user for providing the feedback. In some examples, the operation further includes processing the feedback to improve the accuracy of the speech recognition system and / or the text-speech system.

In one aspect, the invention relates to a method implemented by a data processing apparatus. The method includes old training data (e.g., one or more old text messages whose exact translation is known to another language) and new training data (e.g. one whose exact translation to another language is not known). Or more new text messages); Sending a plurality of each request to the user's client device (or multiple users' multiple client devices) at different times, the requests comprising (i) the user training the old training data and / or the new training; Transmitting, including each request to translate data and (ii) each incentive for translation; After sending a specific request, receiving a translation of the old training data of the specific request from the client device; Comparing the received translation to the correct translation for the old training data; Determining the accuracy of the received translation based on the comparison; And updating a confidence score for the user based on the translation. The reliability score indicates the likelihood that the user will later provide an accurate translation of the text message to the other language.

In certain embodiments, the user is a participant in an online game. Each of the incentives may include, for example, virtual goods and / or virtual currencies for online games. Determining the accuracy of the translation received from the user may include (i) calculating a word error rate (WER) and / or (ii) using bilingual evaluation understudy (BLEU). In some examples, updating the confidence score for the user includes using item response theory to identify deviations from a norm in user translation accuracy. The method may also include canceling the translation privilege of the user when the reliability score falls below a threshold value.

In various implementations, the method includes compensating the respective incentives to the user if the user's translation is determined to be correct. The method may include detecting a conflict between the user and the second user by identifying an existing relationship between the user and the second user. In one example, identifying the existing relationship includes analyzing a social network of at least one of the user and the second user.

In another aspect, the invention relates to a system comprising a computer readable medium on which instructions are stored, and a data processing apparatus. The data processing apparatus is configured to execute instructions for performing an operation, the operation comprising old training data (e.g., one or more old text messages known to the correct translation to another language) and new Selecting training data (eg, including one or more new text messages whose exact translation to another language is unknown); Sending a plurality of each request to the user's client device (or multiple users' multiple client devices) at different times, the requests comprising (i) the user training the old training data and / or the new training; Transmitting, including each request to translate data and (ii) each incentive for translation; After sending a specific request, receiving a translation of the old training data of the specific request from the client device; Comparing the received translation to the correct translation for the old training data; Determining the accuracy of the received translation based on the comparison; And updating a confidence score for the user based on the translation. The reliability score indicates the likelihood that the user will later provide an accurate translation of the text message to the other language.

In certain embodiments, the user is a participant in an online game. Each of the incentives may include, for example, virtual goods and / or virtual currencies for online games. Determining the accuracy of the translation received from the user may include (i) calculating a word error rate (WER) and / or (ii) using bilingual evaluation understudy (BLEU). In some examples, updating the confidence score for the user includes using item response theory to identify deviations from a norm in user translation accuracy. The operation may also include canceling the translation privilege of the user when the reliability score falls below a threshold value.

In various implementations, the operation includes compensating the respective incentives to the user if the user's translation is determined to be correct. The operation may also include detecting a conflict between the user and the second user by identifying an existing relationship between the user and the second user. In one example, identifying the existing relationship includes analyzing a social network of at least one of the user and the second user.

In another aspect, the invention relates to a computer program product stored on one or more storage media for controlling a processing mode of a data processing device. The computer program product is executable by a data processing device and causes the data processing device to perform an action, the action being one or more of old training data (e.g., accurate translation into another language is known. Selecting old training messages (including old text messages) and new training data (eg, including one or more new text messages whose exact translation to another language is unknown); Sending a plurality of each request to the user's client device (or multiple users' multiple client devices) at different times, the requests comprising (i) the user training the old training data and / or the new training; Transmitting, including each request to translate data and (ii) each incentive for translation; After sending a specific request, receiving a translation of the old training data of the specific request from the client device; Comparing the received translation to the correct translation for the old training data; Determining the accuracy of the received translation based on the comparison; And updating a confidence score for the user based on the translation. The reliability score indicates the likelihood that the user will later provide an accurate translation of the text message to the other language.

In certain embodiments, the user is a participant in an online game. Each of the incentives may include, for example, virtual goods and / or virtual currencies for online games. Determining the accuracy of the translation received from the user may include (i) calculating a word error rate (WER) and / or (ii) using bilingual evaluation understudy (BLEU). In some examples, updating the confidence score for the user includes using item response theory to identify deviations from a norm in user translation accuracy. The operation may also include canceling the translation privilege of the user when the reliability score falls below a threshold value.

In various implementations, the operation includes compensating the respective incentives to the user if the user's translation is determined to be correct. The operation may also include detecting a conflict between the user and the second user by identifying an existing relationship between the user and the second user. In one example, identifying the existing relationship includes analyzing a social network of at least one of the user and the second user.

In one aspect, the invention relates to a method implemented by a data processing apparatus. The method comprises: obtaining a text message in a first language, the text message comprising at least one word; Providing the text message to a machine translation system; Obtaining a translation of the text message from the machine translation system; Determining that both the text message and the translation include the at least one word in the first language and that the at least one word is spelled correctly; And next, (a) determining the frequency with which the at least one word appeared in previous text messages; (b) determining a Bayesian probability for neighboring words that appear before and after the at least one word; And (c) performing one or more of the steps of performing k-means clustering to identify a cluster of words containing synonyms. If the frequency exceeds a threshold value, the Bayesian probability exceeds a threshold value, and / or the cluster includes the at least one word, the method includes the at least the dictionary in the data storage unit. And adding a single word.

In certain embodiments, the at least one word may include or be an out of vocabulary word. The at least one word may include or be a new chat term word. The method may include determining whether the dictionary in the data storage unit includes the at least one word. The text message can be received from the client device of the player of the online game. In various examples, the dictionary includes or consists of words in a vocabulary of the first language.

Determining the Bayesian probability includes: (i) reviewing the previous use of the at least one word in previous text messages and (ii) a word appearing before and after the at least one word in the previous text messages. And, if any, identifying. The Bayesian probability can provide an indication of the likelihood that the neighboring words will appear before and after the at least one word in the text message.

In various implementations, identifying the cluster includes reviewing previous text messages and identifying words used in context similar to the at least one word in the text message. The method may also include analyzing the syntax and semantics of the text message to determine parts of speech in the text message.

In another aspect, the invention relates to a system comprising a computer readable medium on which instructions are stored, and a data processing apparatus. The data processing apparatus is configured to execute instructions for performing an operation, the operation comprising: obtaining a text message in a first language, the text message comprising at least one word; Providing the text message to a machine translation system; Obtaining a translation of the text message from the machine translation system; Determining that both the text message and the translation include the at least one word in the first language and that the at least one word is spelled correctly; And next, (a) determining the frequency with which the at least one word appeared in previous text messages; (b) determining a Bayesian probability for neighboring words that appear before and after the at least one word; And (c) performing one or more of the steps of performing k-means clustering to identify a cluster of words containing synonyms. If the frequency exceeds a threshold value, the Bayesian probability exceeds a threshold value, and / or the cluster includes the at least one word, the method includes the at least the dictionary in the data storage unit. And adding a single word.

In certain embodiments, the at least one word may include or be an out of vocabulary word. The at least one word may include or be a new chat term word. The operation may include determining whether the dictionary in the data storage unit includes the at least one word. The text message can be received from the client device of the player of the online game. In various examples, the dictionary includes or consists of words in a vocabulary of the first language.

Determining the Bayesian probability includes: (i) reviewing the previous use of the at least one word in previous text messages and (ii) a word appearing before and after the at least one word in the previous text messages. And, if any, identifying. The Bayesian probability can provide an indication of the likelihood that the neighboring words will appear before and after the at least one word in the text message.

In various implementations, identifying the cluster includes reviewing previous text messages and identifying words used in context similar to the at least one word in the text message. The operation may also include analyzing the syntax and semantics of the text message to determine parts of speech in the text message.

In another aspect, the invention relates to a computer program product stored on one or more storage media for controlling a processing mode of a data processing device. The computer program product is executable by a data processing device and causes the data processing device to perform an operation, wherein the operation comprises obtaining a text message in a first language, wherein the text message is at least one Contains words; Providing the text message to a machine translation system; Obtaining a translation of the text message from the machine translation system; Determining that both the text message and the translation include the at least one word in the first language and that the at least one word is spelled correctly; And next, (a) determining the frequency with which the at least one word appeared in previous text messages; (b) determining a Bayesian probability for neighboring words that appear before and after the at least one word; And (c) performing one or more of the steps of performing k-means clustering to identify a cluster of words containing synonyms. If the frequency exceeds a threshold value, the Bayesian probability exceeds a threshold value, and / or the cluster includes the at least one word, the method includes the at least the dictionary in the data storage unit. And adding a single word.

In certain embodiments, the at least one word may include or be an out of vocabulary word. The at least one word may include or be a new chat term word. The operation may include determining whether the dictionary in the data storage unit includes the at least one word. The text message can be received from the client device of the player of the online game. In various examples, the dictionary includes or consists of words in a vocabulary of the first language.

Determining the Bayesian probability includes: (i) reviewing the previous use of the at least one word in previous text messages and (ii) a word appearing before and after the at least one word in the previous text messages. And, if any, identifying. The Bayesian probability can provide an indication of the likelihood that the neighboring words will appear before and after the at least one word in the text message.

In various implementations, identifying the cluster includes reviewing previous text messages and identifying words used in context similar to the at least one word in the text message. The operation may also include analyzing the syntax and semantics of the text message to determine parts of speech in the text message.

In one aspect, the invention relates to a method implemented by a data processing apparatus. The method comprises: (a) receiving a request to review a portion of the history of text messages from a multi-user chat session, the history comprising a plurality of text messages, each text message being in a respective language A step originated from each chat session participant; (b) performing a plurality of parallel processes, each parallel process selecting (i) each other text message from a portion of the history of the text messages, and (ii) translating the selected text message to a target language. Step comprising; (c) providing the translated text message from the plurality of parallel processes to a user's client device; (d) receiving a request to review another portion of the history of text messages; And (e) repeating steps (b) and (c) for another part of the history of the text message.

In certain embodiments, selecting each of the other text messages includes querying a storage device for a portion of the history of the text messages. Translating the selected text message may include converting at least a portion of the text message from chat terminology to a clear term. In some implementations, the method includes receiving a request from the user to stop reviewing the history of the text messages. The plurality of parallel processes may include one process for each chat session participant. Alternatively or additionally, the plurality of parallel processes may include one process for each language used in each chat session.

In another aspect, the invention relates to a system comprising a computer readable medium on which instructions are stored, and a data processing apparatus. The data processing apparatus is configured to execute instructions for performing an operation, the operation comprising: (a) receiving a request to review a portion of the history of text messages from a multi-user chat session, the history A step comprising a plurality of text messages, each text message being in a respective language and originating from each chat session participant; (b) performing a plurality of parallel processes, each parallel process selecting (i) each other text message from a portion of the history of the text messages, and (ii) translating the selected text message to a target language. Step comprising; (c) providing the translated text message from the plurality of parallel processes to a user's client device; (d) receiving a request to review another portion of the history of text messages; And (e) repeating steps (b) and (c) for another part of the history of the text message.

In certain embodiments, selecting each of the other text messages includes querying a storage device for a portion of the history of the text messages. Translating the selected text message may include converting at least a portion of the text message from chat terminology to a clear term. In some implementations, the operation includes receiving a request from the user to stop reviewing the history of the text messages. The plurality of parallel processes may include one process for each chat session participant. Alternatively or additionally, the plurality of parallel processes may include one process for each language used in each chat session.

In another aspect, the invention relates to a computer program product stored on one or more storage media for controlling the processing mode of the data processing apparatus. The computer program product is executable by a data processing device and causes the data processing device to perform an operation, the operation comprising: (a) requesting to review a portion of the history of text messages from a multi-user chat session. Receiving, wherein the history includes a plurality of text messages, each text message being in a respective language and originating from each chat session participant; (b) performing a plurality of parallel processes, each parallel process selecting (i) each other text message from a portion of the history of the text messages, and (ii) translating the selected text message to a target language. Step comprising; (c) providing the translated text message from the plurality of parallel processes to a user's client device; (d) receiving a request to review another portion of the history of text messages; And (e) repeating steps (b) and (c) for another part of the history of the text message.

In certain embodiments, selecting each of the other text messages includes querying a storage device for a portion of the history of the text messages. Translating the selected text message may include converting at least a portion of the text message from chat terminology to a clear term. In some implementations, the operation includes receiving a request from the user to stop reviewing the history of the text messages. The plurality of parallel processes may include one process for each chat session participant. Alternatively or additionally, the plurality of parallel processes may include one process for each language used in each chat session.

In one aspect, the invention relates to a method implemented by a data processing apparatus. The method includes providing a text message chat system (eg, an online game) to a plurality of users of the online game; Receiving a request to block a second user of the text message chat system from a first user of the text message chat system; And following receipt of the request, blocking text messages from the second user from being displayed to the first user.

In a particular embodiment, following receipt of the request, the method includes blocking future invitations from the second user to the first user to participate in a chat session using the text message chat system. . The method may include receiving a second request from the first user to release the blocking of the second user. In some examples, following receipt of the second request, the method includes allowing a text message from the second user to be displayed to the first user. Following receipt of the second request, the method may include allowing future invitations sent from the second user to the first user to participate in a chat session using the text message chat system. .

In some embodiments, the plurality of users includes or defines an alliance of online games. The method may include translating at least a portion of the text message of the text message chat system from a first language to a second language. The method may also include converting at least a portion of the text message from chatspeak to a clear term. In some embodiments, translation and / or conversion may include or utilize parallel processes. For example, the parallel processes may include or utilize one process for each of the plurality of users of the chat system (or one process for each language used by users).

In another aspect, the invention relates to a system comprising a computer readable medium on which instructions are stored, and a data processing apparatus. The data processing apparatus is configured to execute instructions for performing an operation, the operation comprising: providing a text message chat system (eg, an online game) to a plurality of users of the online game; Receiving a request to block a second user of the text message chat system from a first user of the text message chat system; And following receipt of the request, blocking text messages from the second user from being displayed to the first user.

In a particular embodiment, following receipt of the request, the operation includes blocking future invitations from the second user to the first user to participate in a chat session using the text message chat system. . The operation may include receiving a second request from the first user to release the blocking of the second user. In some examples, following receipt of the second request, the operation includes allowing a text message from the second user to be displayed to the first user. Following receipt of the second request, the operation may include allowing future invitations sent from the second user to the first user to participate in a chat session using the text message chat system. .

In some embodiments, the plurality of users includes or defines an alliance of online games. The operation may include translating at least a portion of the text message of the text message chat system from a first language to a second language. The operation may also include converting at least a portion of the text message from chatspeak to a clear term. In some embodiments, translation and / or conversion may include or utilize parallel processes. For example, the parallel processes may include or utilize one process for each of the plurality of users of the chat system (or one process for each language used by users).

In another aspect, the invention relates to a computer program product stored on one or more storage media for controlling a processing mode of a data processing device. The computer program product is executable by a data processing device and causes the data processing device to perform an operation, wherein the operation provides a text message chat system (eg, an online game) to a plurality of users of the online game. Providing steps; Receiving a request to block a second user of the text message chat system from a first user of the text message chat system; And following receipt of the request, blocking text messages from the second user from being displayed to the first user.

In a particular embodiment, following receipt of the request, the operation includes blocking future invitations from the second user to the first user to participate in a chat session using the text message chat system. . The operation may include receiving a second request from the first user to release the blocking of the second user. In some examples, following receipt of the second request, the operation includes allowing a text message from the second user to be displayed to the first user. Following receipt of the second request, the operation may include allowing future invitations sent from the second user to the first user to participate in a chat session using the text message chat system. .

In some embodiments, the plurality of users includes or defines an alliance of online games. The operation may include translating at least a portion of the text message of the text message chat system from a first language to a second language. The operation may also include converting at least a portion of the text message from chatspeak to a clear term. In some embodiments, translation and / or conversion may include or utilize parallel processes. For example, the parallel processes may include or utilize one process for each of the plurality of users of the chat system (or one process for each language used by users).

Elements of the embodiments described with respect to certain aspects of the invention may be used in various embodiments of other aspects of the invention. For example, it is contemplated that the features of the dependent claims, citing one independent claim, may be used in some devices and / or methods of the other independent claims.

Other features and aspects of the various embodiments will become apparent by the following detailed description in conjunction with the accompanying drawings, which illustrate the features of the embodiments.

According to the present invention, it is possible to increase the accuracy of translation in multi-language communication between a plurality of users.

Various embodiments are described in detail with the following drawings. The drawings are for illustrative purposes only and are not intended to depict certain embodiments. The drawings should not be considered as limiting the breadth, scope or applicability of the embodiments.
1 is a block diagram illustrating an exemplary environment utilizing a multi-language communication system according to various embodiments.
2 is a block diagram illustrating an exemplary communication conversion and translation system in accordance with various embodiments.
3 is a block diagram illustrating an exemplary conversion module in accordance with various embodiments.
4 is a block diagram illustrating an exemplary chat client system according to various embodiments.
5 is a flow chart illustrating an exemplary method of multilingual communication in accordance with various embodiments.
6 is a flow chart illustrating an example method for transforming communication in accordance with various embodiments.
7 is a diagram illustrating an exemplary multi-language chat session between chat client systems in accordance with various embodiments.
8 is a flow chart illustrating the operation of an exemplary multi-language communication method according to various embodiments.
9 is a flow chart illustrating the operation of an exemplary multi-lingual communication method in accordance with various embodiments.
10 is a flowchart illustrating the operation of an exemplary multi-language communication method according to various embodiments.
11 is a flowchart illustrating operation of an exemplary multi-lingual communication method according to various embodiments.
12 is a block diagram illustrating an exemplary digital device according to various embodiments.
13 is a block diagram illustrating an exemplary user feedback system according to various embodiments.
14 is a block diagram illustrating an exemplary user feedback system according to various embodiments.
15 is a flowchart illustrating an exemplary user feedback method according to various embodiments.
16 is a block diagram illustrating data flow of an exemplary user feedback system according to various embodiments.
17 is a screenshot illustrating an example of receiving user feedback for a word according to various embodiments.
18 is a screenshot illustrating an example of skipping user feedback according to various embodiments.
19 is a screenshot illustrating an example of receiving user feedback for a phrase according to various embodiments.
20 is a screenshot illustrating an example of receiving user feedback through a list of select-form responses according to various embodiments.
21 is a screenshot illustrating an example of generating a list of select-form responses according to various embodiments.
22 is a screen shot showing an exemplary incentive notification according to various embodiments.
23 is a screen shot showing an example of when a translation fails between client chat systems according to various embodiments.
24 and 25 are screen shots showing an exemplary list of words or phrases available for user feedback according to various embodiments.
26 is a screenshot illustrating an example of defining a word according to various embodiments.
27 is a screenshot showing an example of a list of select-form responses according to various embodiments.
28 is a screen shot showing an example of a list of statuses for submitted responses according to various embodiments.
29 is a screen shot showing an example of an incentive notification according to various embodiments of the present disclosure.
30 is a flowchart of a method of detecting an unreadable phrase in a language corpus according to various embodiments.
31A is a schematic diagram of a fraud detection module according to various embodiments.
31B is a flow chart of a method for detecting fraud in an incentivized translation according to various embodiments.
32 is a schematic diagram of a system for performing speech-to-text conversion and translation in a group chat environment according to various embodiments.
33A is a schematic diagram of a chat history module for converting and / or translating chat history according to various embodiments.
33B is a flowchart of a method for converting and / or translating chat history according to various embodiments.
34A includes screenshots of a user interface for blocking one or more users of a chat session according to various embodiments.
34B includes screenshots of a user interface for unblocking one or more users of a chat session according to various embodiments.
35 includes a flowchart of a method for blocking one or more users of a chat session according to various embodiments.
36A-36F include screenshots of a user interface that allows a user to correct a language translation of an original message according to various embodiments.
37A and 37B include screenshots of a user interface allowing the user to determine translation corrections submitted by other users in exchange for possible rewards, in accordance with various embodiments.
38 is a schematic diagram of a translation evaluation module for evaluating the accuracy of translation according to various embodiments.
39 is a flowchart of a method for evaluating the accuracy of translation according to various embodiments.

Various embodiments described herein relate to implementations of multilingual communication. Systems and methods according to some embodiments include, for example, Internet-based chat (eg, Apple® iMessage, Windows® Live Messenger, etc.), email (eg, embedded forum messaging, Yahoo® mail, RFC 5322, etc.), text-based mobile phone communication (e.g., SMS messages or MMS messages), posting to online forums (e.g., posting to web-based hobby forums), online social media services (online social media) It is possible to enable multi-language communication through different communication modes including posting to a service (eg, Twitter®, Facebook®, etc.). Certain embodiments may also be used to translate transcripts of communications or conversations that have occurred in the past (eg, deposition transcripts or chat history). Various embodiments include one or more of the following in the text: specialized / sector-specific terms (eg, chatspeak), abbreviations, acronyms, proper noun, usually Two or more languages (e.g. spoken), dealing with / accepting common nouns, diminutives, colloquial words or phrases, and profane words or phrases )) Can implement a communication system and method for translating text between. For example, some of the systems and methods described herein can be utilized to connect chat systems, such as Massive-Multiplayer Online (MMO) games, which can have users chatting in different foreign languages. Can. Through certain embodiments, a chat dialogue between two or more users is translated transparently and can be provided to each user in their respective native language or selected language. have. Additionally, through the use of a multi-tiered / multi-module conversion process, in certain embodiments, for communication between two or more users, compared to what is possible with a conventional translation system alone. Fast translations may be possible (eg, translations between several microseconds).

According to some embodiments, the system or method may perform translation from chatspeak in a first language, such as English, to chat language in a second language, such as French. In other examples, the system or method may include a first language (eg, English) chatspeak from a first language (eg, English) before attempting to translate into a second language (eg, French). , English) to formal speak. In some embodiments, data in which translations manually entered by a human operator, or translations based on previously performed by a translation system (e.g., historical translations performed according to embodiments) are stored. This text translation can be achieved by first querying the storage (eg, translation cache). In embodiments, conversion to one or more parts of the text (e.g., processing of one or more of the following contained in the text: chat terms, acronyms, abbreviations, abbreviations, etc.), suitable for accurate text translation You can try proper nouns, common nouns, colloquial and profanity). For example, in certain embodiments, given text can be transformed to handle (current or past) idiomatic language use between different languages. Embodiments may retry the query to the data storage after conversion for a portion of text. If such a translation lookup attempt fails, in embodiments the text (which may have already been translated) is translated into machine translation services (e.g., third parties, cloud-based translation systems such as Google®translate). You can try to translate using.

In order to further refine the translation of the resulting text, the embodiment converts the translated part of the formal text into a chat language of the new language (eg, formal speak in French). You can try converting into French chatspeak. Thus, certain embodiments enable chat translation between chat terms in a first language (eg, English) and chat terms in a second language (eg, Russian, French, Spanish, Chinese, Hindi, etc.). do.

Some embodiments can help reduce or avoid the need to use machine translation (and thus reduce time, cost, and other indirect costs associated with machine translation), and of text with minimal context or short sentence structure. It can enable accurate translation. The machine translation is implemented by a third-party service or a secure network connection (for example, a Secure-Socket Layer (SSL) connection), and a cost or indirect cost that can be avoided by a specific embodiment may have great meaning. have.

As understood herein, “transformation” means forming a second text segment of the first language by manipulating the first text segment in the first language. The resulting second text segment may also be referred to herein as “transformed text”. “Translation” can be understood to mean converting a text segment of a first language into a text segment of a corresponding second language.

Also, as understood herein, “transformed translation” means, according to the embodiments described herein, that has already been transformed (eg, from the chat language text of the first language, to the official text of the first language). Translation to) means the translation of the text segment (from the first language to the second language). “Untransformed translation” may be understood to mean the translation of the text segment (from the first language to the second language) before the text segment is converted according to the embodiments described herein.

Various embodiments may implement different translation / translation strategies, along with a specific strategy that is well suited to a particular translation application. For example, for a particular chat system application, the conversion strategy implemented may include the following set of conversion related modules: chat term module, acronym module, proper noun module, common noun module, colloquial module, spell check module, abbreviation module, and And applying the profanity module, according to the listed order. In general, the adopted transformation / translation strategy is, which transform operation is performed, and when the transformation operation is performed in the entire translation process (for example, performing a transformation before or after machine translation), Or, in what order, the conversion operations are performed (eg, priority or priority of the conversion operation). The translation / translation strategy also pre-populates any translations in the data storage (eg, translations can be stored in the translation “cache” to speed up the entire process), when in the entire translation process. You can decide whether to utilize the translation cache. In a particular embodiment, the employed translation / translation strategy can be dynamically determined based on the conditions of the environment in which the embodiments are used. For example, if the chat system experiences a load of more users than usual, the conversion / translation strategy can be switched to reduce the processing burden on the chat system (eg, more on machine translation than data storage). will).

1 is a block diagram illustrating an example environment 100 utilizing a multi-language system according to various embodiments. As shown in FIG. 1, the exemplary environment 100 includes clients 102-1 to 102-N, hereinafter collectively referred to as "clients 102" or clients 102 ", and chat server 108. And a translation server 110, each of which can be communicatively coupled to each other via a computer network 106. According to some embodiments, the computer network 106 may be implemented or implemented using one or more local or broadband communication networks, such as the Internet, WiFi networks, WiMax networks, private networks, public networks, and the like. Depending on the embodiment, some or all communication connections with the computer network 106 may be encrypted (e.g., Secure Sockets Layer (for example, Secure Sockets Layer SSL)).

Each client 102, chat server 108 and translation server 110 may be implemented using one or more digital devices, where the digital devices are similar to the digital devices that will be discussed later in connection with FIG. can do. For example, the client 102-1 may receive user input (eg, configured for user interaction), and one or more other clients (eg, client 102- Any type of computing device capable of providing a client user interface to implement communication with any of 2 to 102-N) and performing communication with the chat server 108 through the computer network 106. You can. Such computing devices include mobile phones, tablet computing devices, laptops, desktop computers, personal digital assistants (PDAs), portable gaming units, and wired gaming devices. (wired gaming unit), thin client, set-top box, portable multimedia player (Portable multi-media player) or any other form of network, as known to those skilled in the art, may include user equipment. You can. Furthermore, the one or more chat server 108 and the translation server 110 may include one or more cloud-based services (eg, System-as-a-Service (SaaS), Platform-as-a-Service). (PaaS) or Infrastructure-as-a-Service (IaaS)), or may be implemented using this, and may include one or more servers.

Clients 102 may be configured to communicatively connect with chat server 108 that provides or otherwise enables chat sessions between clients 102. Each of the clients 102-1 to 102-N, the chat client system 104-1 to allow the user of each client 102 to access the chat session through the chat server 108 104-N). Additionally, according to an embodiment, each of the chat client systems 104-1 to 104-N (hereinafter collectively referred to as “chat client systems 104” or “chat client systems 104”) is standalone. ) A chat application, a chat feature built into a non-chat application (eg, video game), or a chat service accessible from a client through a web browser. In some embodiments, those skilled in the art will appreciate that the chat client systems 104 may not be heterogeneous with respect to each other, and that it is possible to establish a chat session between them. The chat client system 104 may input chat (eg, chat messages) from their respective users in a language (and corresponding character set) selected by the user (eg, based on user settings or preferences). And receive chat input to the chat server 108 to be relayed to other users (eg, other users of different chat client systems). The chat client system 104 may also receive chat output (eg, chat session conversations) from the chat server 108 (eg, from other users of other chat client systems), and received chats The output can be displayed in a language (and corresponding character set) selected by the user (eg, based on user settings or preferences).

According to the use of some embodiments, the translation of the chat conversation can be made transparent to the users as the conversation passes between the chat client systems 104. Thus, according to some embodiments, all chat conversations provided in a given chat client system 104, regardless of what language the user has in other chat client systems 104 contributing to the same chat conversation, are given the above chat It may be the native language of the user of the client system 104 (or selected by the user). For example, if the user of the chat client system 104-1 and the user of the chat client system 104-2 form the same chat conversation (ie, are included in the same chat session), the chat client system 104 The user of -1) can choose to input and receive chat conversations in English while the user of the chat client system 104-2 chooses to enter and receive chat conversations in Russian. Although users of the chat client systems 104-1 and 104-2 can view the same chat content, the chat conversations can be displayed in their respective languages.

As shown, the chat server 108 converts according to the chat host system 112 configured to establish and / or run chat sessions between the chat client systems 104 and various systems and methods described herein. And / or a communication transformation and translation (CTT) system 114 configured to perform a translation operation. In some embodiments, chat client systems 104 may establish a chat session with each other through the chat host system 112, where the chat host system 104 chats between the chat client systems 140. It is possible to take advantage of the features of the CTT system 114 to facilitate transparent translation of conversations. In one embodiment, those skilled in the art, the chat host system 112 and the CTT system 114 may be part of separate servers, the entity operating the chat host system 112 (entity) is the CTT system ( It can be appreciated that it may be different from the entity that operates 114). For example, the chat host system 112 may be a third-party chat host system that utilizes the services of the CTT system 114.

Further, as shown, the translation server 110 may include a translation module 116 configured to receive and service a request for machine text translation. According to some embodiments, the CTT system 114 may utilize an operation / service of the translation module 116 that performs machine translation of text. The CTT system 114 may use one or more translation application programming interfaces (APIs) to obtain access to services provided by the translation module 116. Depending on the embodiment, the translation module 116 (and the server 110 on which it is located) may be operated by a third party, providing a paid or free machine translation service, such as Google®. Although the translation module 116 is shown as a component that operates on a separate server from the CTT system 114, those skilled in the art may, in some embodiments, have the translation module 116 on the same server as the CTT system 114. It may be appreciated that it may operate and / or be an integrated component of the CTT system 114.

2 is a block diagram illustrating an exemplary communication conversion and translation system 114 according to various embodiments. As shown, the CTT system 114 includes a communication conversion and translation (CTT) control module 202, a communication conversion and translation (CTT) communication module 204, a language module 206, a conversion module 208, A translation data storage unit 210 and a translation application application programming interface (API) module 212 may be included. The CTT control module 202 is configured to control and / or direct the performance of various operations within the CTT system 114 such that the CTT system 114 performs translation or translation, according to the systems and methods described herein. Can be. In some embodiments, the CTT control module 202, the CTT communication module 204, the language module 206, the conversion module 208, the translation data storage unit 210 and the translation API module 212 ), The operation of other components of the CTT system 114 can be controlled.

The CTT communication module 204 is configured to perform communication between external systems and components with respect to the CTT system 114, such as the CTT system 114 and the chat server 108 and / or the translation server 110. Can be configured. Accordingly, through the CTT communication module 204, the CTT system 114 can receive a chat conversation (including at least one or more chat messages) to be converted or translated by the CTT system 114, The translated chat conversation resulting from the CTT system 114 may be output.

The language module 206 may be configured to identify one or more languages used in connection with the chat conversation received by the CTT system 114. In some embodiments, the language module 206 may identify the language through analysis of the content of the received chat conversation, and / or each chat client system participating in the received chat conversation ( For example, from the chat client system 104, language preference / setting information may be obtained.

The conversion module 208 may be configured to perform a conversion operation on a chat conversation (including one or more chat messages) received from the CTT system 114, according to the systems and methods described herein. . According to some embodiments, the conversion operation performed by the conversion module 208 may include, without limitation, chat terms, acronyms, abbreviations, proper nouns, common nouns, colloquial and profanity. Additional details of the conversion module 208 are discussed in FIG. 3.

When the CTT system 114 performs transformed translation and untransformed translation according to some systems and methods described herein, the translation data storage 210 may have previously been translated. It can be configured to store the text and then provide it to the CTT system 114. As described herein, the translation data storage 210 operates as a cache of translations previously performed by the CTT system 114 and / or by a human operator (eg, a translation training system). You can save translations that are manually entered and saved. In some embodiments, the translation data storage 210 may be filled with translations that can speed up the operation of the CTT system 114 in relation to a particular chat context. For example, the CTT system 114 may be utilized in connection with a chat system related to an MMO game, and the translation data storage unit 210 may, in particular, be associated with an MMO game (converted and unconverted). Can be filled with translations (eg, by the operator of the CTT system 114). In a particular embodiment, the multi-tiered / multi-module approach to converting text, used by the conversion module 208, is actually very complex, but in particular chat in MMO games Very suitable for text processing.

According to an embodiment, the data storage unit 210 is an untranslated translation (eg, <English Formal> 'you'-> <Official French (French Formal)-> 'vous'), You can save the translated translation (eg <English Chatspeak> 'u'-> <official French>-> 'vous'), or both. In some embodiments, the translation data storage unit 210 may store a translation such that a corresponding chat message can be identified using a hash value / tag. For example, in order to store the Spanish translation of the original message in English, the Spanish translation may be stored based on the hash value of the English message, whereby the Spanish translation may display the hash value of the English message. By using it, the location can be located and acquired later. Those skilled in the art may recognize that, in some embodiments, the translation data storage unit 210 may include separate data storage units for translation between two specific languages. Therefore, when a chat message is converted / translated between English and French, the corresponding data English-French data store is utilized for operations related to the translation data storage 210. Can be.

The translation API module 212 may be configured to provide a connection to a machine translation service provided outside the CTT system 114 to the CTT system 114 (for example, the translation server 110 ) (By the translation module 116). As described herein, when a translation is not found in the translation data storage 210, the translation API module 212 may be utilized by the CTT system 114.

3 is a block diagram illustrating an example conversion module 280 according to various embodiments. As shown, the conversion module 208 is a chat term module 302, an acronym module 304, a proper noun module 306, a common noun module 307, a colloquial module 310, a spell check module 312 , Abbreviation module 314 and / or profanity module 316. According to some embodiments, during operation, the conversion module 208 may process all or part of the chat message (eg, divide the message into tokens or logical portions, and then these tokens) / Process part). In some embodiments, various modules of conversion module 208 may be called in parallel.

The chat term module 302 may identify one or more words or phrases related to a chat jargon (ie, chat term) in a chat message, and furthermore, the identified word or It may be configured to suggest a replacement word or phrase for a phrase (eg, a corresponding official term / ie non-chat term). In some embodiments, the chat term module 302 may display the identified chat term words or phrases to be omitted or otherwise ignored in a sequential machine translation process (eg, by the translation module 116). have. Additionally, in some embodiments, the identified chatspeak words or phrases can then be displayed for review and disposition by human operators (eg, administrators of the CTT system 114). . To identify chat term words or phrases, and / or their corresponding (official) words or phrases, some embodiments may include chat term words or phrases, and / or between chat term words or phrases and their corresponding words or phrases It is possible to utilize a dataset (stored in a data storage, for example), including mapping of. The dataset may be built through a training or learning system, and may be owned by an owner (eg, “in-house” manually collected by the administrator of the CTT system 114). , Or commercially available, or derived from publicly available Internet knowledge bases. For example, the chat terminology module 302 can apply statistical machine translation to its function. In some embodiments, the applied statistical machine translation can be trained using parallel text and / or phrase-level pairs extracted from transformations containing contextual information, and / or Or else, you can add grammar to sentences that don't fit the grammar. The result by the chat term module 302 is the chat term word or phrase marked to be ignored by the chat term module 302, the proposed replacement (replacement), or the chat term module 302 to the message. It may include non-chatspeak words or phrases that are inserted (eg, in place of the identified chat term words or phrases). According to an embodiment, the message resulting from the chat term module 302 may be provided to another conversion module (in the conversion module 208) for further processing, or the proposed replacement may be The message converted by the chat term module 302 may be provided to the CTT control module 202 to determine whether the message is stored in the data storage unit 210.

The acronym module 304 may be configured to identify one or more acronyms in a chat message, and further may include replacement words or phrases corresponding to the acronyms (eg, represented by acronyms). It can be configured to suggest. In some embodiments, the acronym module 304 may indicate the identified acronyms to omit or ignore during subsequent machine translations (eg, by the translation module 116). Additionally, in some embodiments, the identified acronyms can then be displayed for review and disposition by a human operator (eg, an administrator of the CTT system 114). A dataset, including mapping between acronyms, and / or acronyms and their corresponding words and phrases, in some embodiments, to identify acronyms and / or corresponding words or phrases , For example, may be utilized. The dataset may be built through a training or learning system, and may be owned by an owner (eg, “in-house” manually collected by the administrator of the CTT system 114). , Commercially available, or derived from publicly available Internet knowledge bases. The result by the acronym module 304 may be inserted into the message by the acronym module 304, a suggested acronym, suggested replacement, or by the acronym module 302 ( For example, it may include words or phrases (instead of identified acronyms). According to an embodiment, the message resulting from the acronym module 304 may be provided to another transformation module (in the transformation module 208) for further processing, or the acronym module 304 ) To determine whether the message converted by the data storage unit 210, the proposed replacement may be provided to the CTT control module 202.

The proper noun module 306 may be configured to identify one or more proper nouns in a chat message, and further propose a replacement word or phrase corresponding to the proper noun (eg, represented by a proper noun) It can be configured to. In some embodiments, the proper noun module 306 may indicate the identified proper noun to be omitted or ignored during subsequent machine translation (eg, by the translation module 116). Additionally, in some embodiments, the identified proper noun can then be displayed for review and disposition by a human operator (eg, an administrator of the CTT system 114). In order to identify proper nouns and / or corresponding words or phrases, in some embodiments, proper nouns (eg, well-known nouns or common names of individuals, such as Disneyland®), and / or nouns and their It is possible to utilize a dataset (stored in a data store, for example), including mapping between corresponding words and phrases. The dataset may be built through a training or learning system, and may be owned by an owner (eg, “in-house” manually collected by the administrator of the CTT system 114). , Or commercially available, or derived from publicly available Internet knowledge bases. The result by the proper noun module 306 will be inserted into the message by the proper noun module, proposed replacement, or the proper noun module 302, which is marked to be ignored by the proper noun module 306 (eg , On behalf of the identified noun). According to an embodiment, the message resulting from the proper noun module 306 may be provided to another transform module (in the transform module 208) for further processing, or the proper noun module 306 ) To determine whether the message converted by the data storage unit 210, the proposed replacement may be provided to the CTT control module 202.

The common noun module 308 may be configured to identify one or more common nouns in a chat message, and further propose a replacement word or phrase corresponding to the common noun (eg, represented by a common noun) It can be configured to. In some embodiments, the common noun module 308 may indicate the identified common noun to be omitted or ignored during subsequent machine translation (eg, by the translation module 116). Additionally, in some embodiments, the identified common noun can then be displayed for review and disposition by a human operator (eg, the manager of the CTT system 114). In order to identify common nouns and / or their corresponding words or phrases, in some embodiments, a dataset, comprising common nouns and / or mapping between common nouns and their corresponding words and phrases dataset, for example). The dataset may be built through a training or learning system, and may be owned by the owner (eg, “in-house” manually collected by the administrator of the CTT system 114). , Or commercially available, or derived from publicly available Internet knowledge bases. The results by the common noun module 308 may be inserted into the message by the common noun module, proposed replacement, or by the common noun module 308, for example, indicated by the common noun module 308 to be ignored. Words or phrases (instead of identified common nouns). According to an embodiment, the message resulting from the common noun module 308 may be provided to another conversion module (in the conversion module 208) for further processing, or the common noun module 308 ) To determine whether the message converted by the data storage unit 210, the proposed replacement may be provided to the CTT control module 202.

The colloquial module 310 may be configured to identify one or more colloquial words or phrases in a chat message, and further replacement of the identified word or phrase (eg, corresponding official / ie colloquial language Not) can be configured to suggest words or phrases. In some embodiments, colloquial module 310 may display the identified colloquial word or phrase to omit or ignore during subsequent machine translation (eg, by translation module 116). Additionally, in some embodiments, the identified colloquial words or phrases can then be displayed for review and disposition by human operators (eg, administrators of the CTT system 114). . To identify colloquial words or phrases and / or corresponding (official) words or phrases, in some embodiments, colloquial words or phrases, and / or mapping between colloquial words or phrases and their corresponding words and phrases A dataset, including (mapping), may be utilized. The dataset may be built through a training or learning system, and may be owned by an owner (eg, “in-house” manually collected by the administrator of the CTT system 114). , Or commercially available, or derived from publicly available Internet knowledge bases. The result by the colloquial module 310 will be inserted into the message by the colloquial module 310, or a colloquial word or phrase marked as to be ignored by the colloquial module 310, or by the colloquial module 310 (eg, identification Non- colloquial words or phrases (in lieu of spoken words or phrases). According to an embodiment, the message derived as a result by the colloquial module 310 may be provided to another conversion module (included in the conversion module 208) for further processing, or the colloquial module 310 In order to determine whether the message converted by is in the data storage unit 210, the proposed replacement may be provided to the CTT control module 202.

The spell check module 312 may be configured to identify one or more misspelled words or phrases in a chat message, and further replace (eg, corrected) the identified words or phrases ) Can be configured to suggest words or phrases. For example, the spell check module 312 may be configured to automatically correct the word or phrase with the suggested replacement word or phrase. According to some embodiments, the spell checking module 312 may display the misspelled word or phrase that is identified to be omitted or ignored during subsequent machine translation (eg, by the translation module 116). have. Additionally, in some embodiments, the identified misspelled words or phrases are then marked for review and disposition by a human operator (eg, an administrator of the CTT system 114). Can be. To identify misspelled words or phrases and / or corresponding (corrected) words or phrases, in some embodiments, misspelled words or phrases, and / or misspelled words or phrases and their corresponding words And a mapping between verses (dataset, for example, stored in a data storage). The dataset may be built through a training or learning system, and may be owned by an owner (eg, “in-house” manually collected by the administrator of the CTT system 114). , Or commercially available, or derived from publicly available Internet knowledge bases. The results by the spell check module 312 will be inserted into the message by the misspelled word or phrase, suggested replacement, or by the spell check module 312, which is marked to be ignored by the spell check module 312 ( For example, it may include a corrected word or phrase (in place of a misspelled word or phrase). According to an embodiment, the message derived as a result by the spell checking module 312 may be provided to another conversion module (included in the conversion module 208) for further processing, or the spell checking module ( In order to determine whether the message converted by 312) is in the data storage unit 210, the proposed replacement may be provided to the CTT control module 202.

The abbreviation module 314 may be configured to identify one or more abbreviations in a chat message, and further to suggest replacement words or phrases corresponding to the identified abbreviations (eg, represented by abbreviations). Can be configured. According to some embodiments, abbreviation module 314 may display the identified abbreviations to be omitted or ignored during subsequent machine translations (eg, by translation module 116). Additionally, in some embodiments, the identified abbreviations may then be displayed for review and disposition by a human operator (eg, the manager of the CTT system 114). To identify an abbreviation and / or a corresponding word or phrase, in some embodiments, a dataset, including an abbreviation and / or mapping between the abbreviation and their corresponding words and phrases For example, it can utilize (stored in the data storage). The dataset may be built through a training or learning system, and may be owned by the owner (eg, “in-house” manually collected by the administrator of the CTT system 114). , Or commercially available, or derived from publicly available Internet knowledge bases. The result by the abbreviation module 314 may be inserted into the message by the abbreviation module 314, an abbreviation marked to be ignored by the abbreviation module 314, or the abbreviation module 314 (eg, the identified abbreviation). Instead) words or phrases. According to an embodiment, the message resulting from the abbreviation module 314 may be provided to another transformation module (included in the transformation module 208) for further processing, or the abbreviation module 314 In order to determine whether the message converted by is in the data storage unit 210, the proposed replacement may be provided to the CTT control module 202.

The profanity module 316 may be configured to identify one or more profanity words or phrases (hereinafter referred to as “profanity words”) in a chat message, and further, a replacement word or phrase corresponding to the identified profanity ( For example, it may be configured to suggest appropriate substitutions (eg, soft euphemism). According to some embodiments, the profanity module 316 may indicate the profanity identified to be omitted or ignored during subsequent machine translation (eg, by the translation module 116). Additionally, in some embodiments, the identified profanity may then be displayed for review and disposition by a human operator (eg, an administrator of the CTT system 114). To identify profanity and / or corresponding words or phrases, in some embodiments, a dataset, e.g., including profanity and / or mapping between the profanity and their corresponding words and phrases For example, it can utilize (stored in the data storage). The dataset may be built through a training or learning system, and may be owned by an owner (eg, “in-house” manually collected by the administrator of the CTT system 114). , Or commercially available, or derived from publicly available Internet knowledge bases. The result by the profanity module 316 may be inserted into the message by the profanity module, slang proposed by the profanity module 316 to be ignored, or by the profanity module 316 (eg, the identified profanity Instead) words or phrases. According to an embodiment, the message resulting from the slang module 316 may be provided to another conversion module (included in the conversion module 208) for further processing, or the slang module 316 In order to determine whether the message converted by is in the data storage unit 210, the proposed replacement may be provided to the CTT control module 202.

According to some embodiments, one or more of the various modules of the conversion module 208 may insert one or more of the chat message by inserting a predetermined character before and / or after the portion to be displayed. The part can be marked. For example, when the chat term module 302 displays the word “LOL” included in a part of the chat message, the chat term module 302 ignores the displayed part by the translation module 116 To indicate that it should be, a preset character (“_”) can be inserted before and / or after the word (eg, “_LOL_”).

According to some embodiments, the conversion module 208 may perform two or more conversion operations in parallel on the initial message, and in response, each two or more conversions will return a separate response. From there, the conversion module 208 can select one converted message for further processing (eg, used in operation 514). According to the above embodiment, each response may include a displayed text area, a suggested replacement, or a word or phrase inserted in the initial message. Thereafter, the selected converted message may be subject to a priority of selection, which may determine which converted message will be selected for further processing, according to what source. In some embodiments, the priority selection, which conversion operation is easy to generate a converted message suitable for a subsequent search in the translation data storage unit 210, or is easy to generate a converted message suitable for a subsequent machine translation Can follow. Additionally, according to some embodiments, the priority of the selection may depend on which transform operation produces the most formal transformed message. The priority of selection may be determined according to the conversion / translation strategy selected by the embodiment.

Table 1 provides an example of how the conversion module processes a portion of the chat message, according to various embodiments. As illustrated, the conversion module 208 may process a chat message based on a token or a proximal token, and once conversion is performed, processing of a specific token may be stopped. .

Chat Message Processing Example Token for chat messages
(Token (s) from a Chat Message) Conversion processing
(Transformation Processing) Token = 'USA' Chatting term module ('USA') → Acronym module ('USA') → Marking not to translate Token = 'brb' Chat Glossary Module ('brb') → Acronym Module ('brb') → Proper Noun Module ('brb') → Common Noun Module ('brb') → Spoken Language Module ('brb') → Spell Checking Module ('brb) ') → Abbreviation module (' brb ') → Convert to' be right back ' Token = '9' Chat Term Module ('9') → Converted to 'parents watching over shoulder' Token = '99' Chat Term Module ('99') → Converted to 'parents stopped watching over shoulder' Proximal tokens = 'go gabe' Chat Glossary Module ('go gabe') → Acronym Module ('go gabe') → Proper Noun Module ('going') → Common Noun Module ('go gabe') → Indicates that there is a possibility of being a common noun String = 'Your going to attack him?'
Token # 1 = 'Your'
Token # 2 = 'going'
Token # 3 = 'to'
Token # 4 = 'attack'
Token # 5 = 'him' Spell check module ('Your') → Corrected to 'you're' based on the proxy word token 'going' (ie, use a proximal context for spell checking)

Chat Glossary Module ('going') → Acronym Module ('going') → Proper Noun Module ('going') → Common Noun Module ('going') → Colloquial Module ('going') → Spell Checking Module ('going) ') → Abbreviation module (' going ') → Slang module (' going ') → Do not convert.

Chat Glossary Module ('to') → Acronym Module ('to') → Proper Noun Module ('to') → Common Noun Module ('to') → Colloquial Module ('to') → Spell Checking Module ('to ') → Abbreviation module (' to ') → Profanity module (' to ') → Do not convert.

Chat Glossary Module ('attack') → Acronym Module ('attack') → Proper Noun Module ('attack') → Common Noun Module ('attack') → Colloquial Module ('attack') → Spell Checking Module ('attack) ') → Abbreviation module (' attack ') → Slang module (' attack ') → Do not convert.

Chat Glossary module ('him') → Acronym module ('him') → Proper noun module ('him') → Common noun module ('him') → Spoken language module ('him') → Spell check module ('him ') → Abbreviation module (' him ') → Profanity module (' him ') → Do not convert. String = 'Sup bro, sup yall?'
Token # 1 = 'Sup'
Token # 2 = 'bro'
Token # 3 = 'sup'
Token # 4 = 'yall' Replace the chat term module ('Sup') with "How is it going."

Chat Term module ('bro') → Acronym module ('bro') → Proper noun module ('bro') → Common noun module ('bro') → Spoken language module ('bro') → Spell check module ('bro ') → Abbreviation module (' bro ') → Replaced with "brother"

Replace the chat term module ('sup') → "how is it going."

Chat Term Module ('yall') → Replaced with "you all."

4 is a block diagram illustrating an exemplary chat client system 104 according to various embodiments. As illustrated, the chat client system 104 may include a chat client control unit 402, a chat client communication module 404, and a chat client graphical user interface (GUI) module 406. The chat client control module 402 is within the chat client system 104 as the chat client system 104 performs actions related to chat (eg, communication chat conversation with the chat server 108). It can be configured to control and / or direct the performance of various operations. In some embodiments, the chat client control module 402 includes other components of the chat client system 104, such as, for example, the chat client communication module 404 and the chat client GUI module 406. You can control the operation.

The chat client communication module 404 is configured to implement communication between a chat client system 104 and external systems and components for the chat client system 104, such as the chat server 108. Can be configured. Thus, through the chat client module 404, the chat client system 104 is provided from the chat server 108 to the chat client system 104 (eg, via the chat client GUI module 406). A chat conversation may be received, and a chat conversation received from a user of the chat client system 104 may be transmitted to a chat server (eg, via the chat client GUI module 406).

The chat client GUI module 406 may be configured to provide users of the chat client system 104 with graphical input / output access to chat sessions with other chat client systems. Thus, in some embodiments, the chat client GUI module 406 can provide a client GUI to a user of the client 102, adapted to receive user interaction through the client 102. have. According to some embodiments, the chat client GUI module 406 provides the user with chat conversations (eg, from the chat server 108) in their selected language (eg, according to the user's preferred / set language). Can be configured to provide). Additionally, the chat client GUI module 406 may be configured to receive chat input from the user in their chosen language (eg, according to the user's preferred / set language). Here, as described, in the chat client system 104, the language used to display and receive the chat conversation may be different from the language used to display and receive the chat conversation in another chat client system. . More on the chat client GUI module 406 is described with respect to FIG. 7.

5 is a flow diagram illustrating an exemplary method 500 for multilingual communication, in accordance with various embodiments. As described below, in some embodiments, the method represented by the method 500 includes a chat client system 104-1, a chat client system 104-2, and a CTT system 114 (eg For example, an operation may be performed in connection with the chat server 108 and the translation module 116 (eg, the translation server 110).

The method 500 can begin at operation 502, and the language module 204 (of the CTT system 114) is the first used by the user of the first chat client system (eg, 104-1). The language and the second language used by the user of the second chat client system (eg, 104-2) can be identified. According to some embodiments, the language module 204 may identify the first language and the second language according to the language preference / setting obtained from each chat client system 104.

In operation 504, the CTT communication module 204 (of the CTT system 114) may receive an initial message in a first language. In some embodiments, the CTT communication module 204 may receive the initial message from the chat host system 112, which has received the initial message from a chat client system (eg, 104-1).

In operation 506, the CTT control module 202 (of the CTT system 114) queries the translation data storage 210 for a corresponding message in a second language, corresponding to the initial message. You can. In operation 508, the CTT control module 202 may determine whether a corresponding message is found in the translation data storage 210. If there is one, in operation 510 the CTT communication module 204 may support sending the corresponding message to a second chat client system (eg, chat client system 104-2). In some embodiments, the correspondence message may be sent to the chat host system 112, which may relay the correspondence message to a second chat client system (eg, 104-2). The method 500 can then end.

If a corresponding message does not exist in the translation data storage 210, in operation 512, the conversion module 208 converts at least a part of the initial message into a converted message in a first language. You can try. As described herein, the message derived by the conversion module 208 may remain converted or unchanged (eg, the conversion operation of the conversion module 208 is not applied to the initial message). When not). In some embodiments, the conversion module 208 may perform two or more conversion operations in parallel on the initial message, and in response, each two or more conversion operations may generate a separate response. Can reply, and the conversion module 208 can then select one converted message for further processing (eg, used in operation 514). According to an embodiment, each response may include a portion of the displayed text, a suggested replacement, or a word or phrase to be inserted into the initial message. Thereafter, the selected converted message may be in accordance with the priority of the selection, where the priority of the selection may determine which converted message is selected for further processing and according to which precedent. In some embodiments, the priority selection, which conversion operation is easy to generate a converted message suitable for a subsequent search in the translation data storage unit 210, or is easy to generate a converted message suitable for a subsequent machine translation Can follow. Additionally, according to some embodiments, the priority of the selection may depend on which transform operation produces the most formal transformed message. The priority of selection may be determined according to the conversion / translation strategy selected by the embodiment.

In operation 514, assuming that the conversion module 208 has converted the message, the CTT control module 202 (of the CTT system 114) is in a second language, corresponding to the converted message. The corresponding message may be queried by the translation data storage unit 210. In operation 516, the CTT control module 202 may determine whether the corresponding message is found in the translation data storage 210. If one exists, in operation 518, the CTT communication module 204 may support sending a corresponding message to a second chat client system (eg, chat client system 104-2). . In some embodiments, the correspondence message may be sent to a chat host system 112 that relays the correspondence message to the second chat client system (eg, 104-2). Method 500 may then end.

In some embodiments, if the corresponding message is still not present in the translation data storage 210, in operation 520, the CTT control module 202 is a conversion module that can be performed on a chat message that has not yet been performed. It may be determined whether there is an additional conversion operation of 208.

If there is an additional conversion operation, method 500 returns to operation 512 to perform additional conversion operation (s). According to an embodiment, the additional conversion operation (s) may include, by the conversion module 208, application of a different conversion operation to those already applied on the initial message, and the same conversion to other parts of the English chat message. Motions, or the same combination thereof. For example, if during the first execution of operation 512, the conversion module 208 applies a chat term-related operation to the initial message (to generate the first converted message), the operation 512 During the second execution, the conversion module 208 may apply the abbreviation-related operation to the second converted message. Then, with the next execution of operation 512, the method 500 can continue with operations 514 and 516. In operation 514 and 516, the CTT control module 202 queries the translation data storage 210 again for a corresponding message in a second language corresponding to the most recently derived converted message, The CTT control module 202 determines whether a corresponding message is found in the translation data storage 210. By performing transform and query operations in this iterative manner, some embodiments may be able to find a corresponding message before performing all possible transform operations. Those skilled in the art, in certain embodiments, only the query operation (eg, operation 514) is performed after all possible transformation operations provided by the transformation module 208 are performed on the chat message, while the transformation is performed. And it can be appreciated that the query operation can be performed serially.

If a corresponding message does not exist in the translation data storage 210 and there is no additional conversion operation (of the conversion module 208), in operation 522 (via the translation API module 212), the translation module 116 may support translation of an initial message or a converted message into a corresponding message in a second language. Thereafter, in operation 524, the CTT communication module 204 may support sending a corresponding message to a second chat client system (eg, chat client system 104-2). According to some embodiments, the correspondence message may be transmitted to the chat host system 112, which may relay the correspondence message to the second chat client system (eg, 104-2). The method 500 can then end.

In some embodiments, the conversion module 208 may be utilized to convert a correspondence message in a second language before the correspondence message is sent to the chat host system 112. As described herein, the corresponding message can be submitted for further conversion processing, to further refine the translation for the user of the second chat client system 104-2. For example, if the initial message includes chat terms in the first language (eg, English), further conversion processing may add chat terms in the second language to the extent possible.

Although the method of steps above has been described and described in a particular order, those skilled in the art will appreciate that the order of steps performed may vary between embodiments. Additionally, those skilled in the art will appreciate that in connection with method 500, the components described above are merely examples of components that may be used in the method, and in some embodiments, other components may also be utilized as some embodiments. You can.

6 is a flow diagram illustrating an example method 600 for transforming communications, in accordance with various embodiments. As described below, in some embodiments, method 600 performs operations in connection with transformation module 209 (eg, of CCT system 114).

The method begins at operation 602, where an initial message is received at the conversion module 208 for conversion processing. Based on some embodiments, the conversion module 208 may receive an initial message for conversion after the translation data storage 210 fails to identify the message, and the initial message is a third-party or owner translation process ( It may also be possible to receive an initial message before being machine translated by a proprietary translation process (eg, translation module 116 that can be provided as a cloud-based service). As described herein, the conversion module 208 can be used in various embodiments to facilitate or improve text translation. In this case, the text should include minimal context, simple sentence structure, and terminology limited to professional / sector (eg, chat term for Internet-based chatting), abbreviations, acronyms, colloquial language, proper nouns, common nouns, profanity, or combinations thereof. It can contain. Text translations that can benefit from the operation of the translation module 208 include conversation (eg, transcript), offline or online Internet-based chat (eg, instant messaging) and mobile phone messaging (eg For example, it may include translation of text originating from SMS or MMS), but is not limited thereto.

In operation 604, the chat term module 302 may identify one or more words or phrases associated with a chat jargon (ie, chat term) in the initial message, and the identified words or phrases It is possible to suggest more words or phrases for replacement (eg, corresponding formal / ie non-chat terms). In some embodiments, chat terminology module 302 may indicate that the identified word or phrase is to be omitted or ignored during the next machine translation (eg, by translation module 116). Additionally, in some embodiments, the identified chatspeak words or phrases may be displayed for review and disposition by a human operator (eg, manager of CTT system 114). . In order to identify chat term words or phrases and / or their corresponding (official) words or phrases, some embodiments may include chat term words or phrases and / or between chat term words or phrases and their corresponding words or phrases It is possible to utilize a dataset (stored in a data storage, for example), including mapping of. The dataset may be built through a training or learning system, and may be owned by the owner (eg, “in-house” manually collected by the manager of the CTT system 114), and commercially available. It may be obtained from, or may be derived from publicly available Internet knowledge. The message (hereinafter referred to as 'first intermediate message') resulting from the operation 604 is displayed by the chat term module 302 to be ignored or a chat term word or phrase, suggested replacement, or chat term module 302 ) Into the initial message (eg, instead of the identified chat term word or phrase). For example, the first intermediate message may be the same as the initial message (eg, when no replacement is performed by the chat term module 302). According to an embodiment, the first intermediate message generated from the chat term module 302 may be provided to another conversion module (in the conversion module 208) for further processing, or the proposed replacement may be a chat term. The message converted by the module 302 may be provided to the CTT control module 202 to determine whether it is in the data storage 302. After operation 604, the first intermediate message can be provided to the next operation (eg, operation 606) of the conversion module 208 for processing.

In operation 606, the acronym module 304 can identify one or more acronyms in the chat message, and can further suggest replacement words or phrases (represented by) corresponding to the acronyms. In some embodiments, the acronym module 304 can indicate that the identified acronym is omitted or ignored during the next machine translation (eg, by the translation module 116). Additionally, in some embodiments, the identified acronyms can be displayed for review and disposition by a human operator (eg, an administrator of the CTT system 114). A dataset comprising, in some embodiments, acronyms, and / or mappings between acronyms and their corresponding words and phrases, to identify acronyms and / or their corresponding words or phrases ( dataset, for example). The dataset may be built through a training or learning system, and may be owned by the owner (eg, 'internally' collected manually by the manager of the CTT system 114), or may be commercially obtained. Or, it may come from publicly available Internet knowledge. The message (hereinafter referred to as 'second intermediate message') resulting from operation 606 is displayed by the acronym module 304 and is inserted into the message by the acronym to be ignored, the proposed replacement, or the acronym module 304 May include words or phrases (eg, in place of the identified acronyms). In some examples, the second intermediate message may be the same as the first intermediate message (eg, when no replacement is performed by the acronym module 304). According to an embodiment, the second intermediate message generated from the acronym module 304 may be provided to another conversion module (in the conversion module 208) for further processing, or the proposed replacement may include the acronym module ( 304) may be provided to the CTT control module 202 to determine whether the message converted by the data storage 210. After operation 606, a second intermediate message may be provided to the next operation (eg, operation 608) of the conversion module 208 for processing.

In operation 608, the proper noun module 306 can identify one or more proper nouns in the chat message, and further adds a replacement word or phrase corresponding to the proper noun (eg, represented). I can suggest. In some embodiments, the proper noun module 306 can indicate that the identified proper noun is omitted or ignored during the next machine translation (eg, by the translation module 116). Additionally, in some embodiments, the identified proper noun may be marked for post-mortem review and disposition by a human operator (eg, manager of the CTT system 114). To identify proper nouns and / or their corresponding words or phrases, in some embodiments, proper nouns (eg, common names of well-known nouns or individuals such as Disneyland®) and / or nouns and their It is possible to utilize a dataset (stored in a data store, for example), including mapping between corresponding words and phrases. The dataset can be built through a training or learning system, can be owned by the owner (eg, 'internally' collected manually by the administrator of the CTT system 114), or can be commercially obtained. Or it may come from publicly available Internet knowledge. The message resulting from operation 608 (hereinafter referred to as the 'third intermediate message') is displayed by the proper noun module 306 and is to be ignored by the proper noun, proposed replacement, or proper noun module 306. It may contain words or phrases embedded in the message (eg, in place of its identified proper noun). In some examples, the third intermediate message may be the same as the second intermediate message (eg, when no replacement is performed by the proper noun module 306). According to an embodiment, the third intermediate message generated from the proper noun module 306 may be provided to another transform module (in the transform module 208) for further processing, or the proposed replacement may be a proper noun. The message converted by the module 306 may be provided to the CTT control module 202 to determine whether it is in the data storage 210. After operation 608, a third intermediate message may be provided to the next operation (eg, operation 610) of the conversion module 208 for processing.

In operation 610, the common noun module 308 may identify one or more common nouns in the chat message, and further add a replacement word or phrase corresponding to the common noun (eg, represented). I can suggest. In some embodiments, common noun module 308 may indicate that the identified common noun is omitted or ignored during the next machine translation (eg, by translation module 116). Additionally, in some embodiments, the identified common noun can be marked for follow-up review and disposition by a human operator (eg, the manager of the CTT system 114). A dataset comprising a common noun, and / or a mapping between the common noun and their corresponding words and phrases, in some embodiments, to identify the common noun and / or its corresponding word or phrase (dataset, for example, stored in data storage) can be utilized. The dataset may be built through a training or learning system, may have an owner (eg, 'internally' collected manually by the administrator of the CTT system 114), or may be commercially obtained. Or, it may come from publicly available Internet knowledge. The message resulting from operation 610 (hereinafter referred to as the 'fourth intermediate message') is displayed by the common noun module 308 and is to be ignored by the common noun, the proposed replacement, or the common noun module 308. It may contain words or phrases embedded in a message (eg, in place of its identified common noun). In some examples, the fourth intermediate message may be the same as the third intermediate message (eg, when no replacement is performed by the common noun module 308). According to an embodiment, the fourth intermediate message generated from the common noun module 308 may be provided to another conversion module (in the conversion module 208) for further processing, or the proposed replacement may be a common noun. The message converted by the module 308 may be provided to the CTT control module 202 to determine whether it is in the data storage 210. After operation 610, a fourth intermediate message may be provided to the next operation (eg, operation 612) of the conversion module 208 for processing.

In operation 612, the colloquial module 310 may identify one or more colloquial words or phrases in the chat message, and also replace the identified words or phrases (eg, corresponding official / ie, Non-verbal) more words or phrases can be suggested. In some embodiments, colloquial module 310 may indicate that the identified colloquial word or phrase is omitted or ignored during the next machine translation (eg, by translation module 116). Additionally, in some embodiments, the identified colloquial words or phrases can be marked for follow-up review and disposition by a human operator (eg, manager of the CTT system 114). To identify colloquial words or phrases and / or their corresponding words or phrases, in some embodiments, colloquial words or phrases and / or mapping between colloquial words or phrases and their corresponding words and phrases It may utilize a dataset (dataset, for example, stored in the data storage), including. The dataset may be built through a training or learning system, may have an owner (eg, 'internally' collected manually by the administrator of the CTT system 114), or may be commercially obtained. Or, it may come from publicly available Internet knowledge. The message (hereinafter referred to as 'Fifth Intermediate Message') as a result of the operation 612 is displayed by the colloquial module 310 to be ignored by the colloquial word or phrase, proposed replacement, or colloquial module 310. May include non- colloquial words or phrases embedded in the message (eg, in place of the identified colloquial words or phrases). In some examples, the fifth intermediate message may be the same as the fourth intermediate message (eg, when no replacement is performed by colloquial module 310). According to that embodiment, the fifth intermediate message generated from the colloquial module 310 may be provided to another conversion module (in the conversion module 208) for further processing, or the proposed replacement may be a colloquial module The message converted by 310 may be provided to the CTT control module 202 to determine whether it is in the data storage 210. After operation 612, a fifth intermediate message may be provided to the next operation (eg, operation 614) of the conversion module 208 for processing.

In operation 614, the spell check module 312 can identify one or more misspelled words or phrases in the chat message, and also replace (eg, corrected) the identified words or phrases You can suggest more words or phrases. For example, the spell check module 312 can automatically correct a word or phrase with the suggested replacement word or phrase. In some embodiments, the spell checking module 312 can indicate that the identified misspelled word or phrase is omitted or ignored during the next machine translation (eg, by the translation module 116). . Additionally, in some embodiments, the identified misspelled words or phrases may be marked for follow-up review and disposition by a human operator (eg, manager of CTT system 114). To identify misspelled words or phrases and / or their corresponding (corrected) words or phrases, in some embodiments, misspelled words or phrases and / or misspelled words or phrases and their corresponding A dataset (stored in a data store, for example), including mapping between words and phrases, can be utilized. The dataset may be built through a training or learning system, may have an owner (eg, 'internally' collected manually by the administrator of the CTT system 114), or may be commercially obtained. Or, it may come from publicly available Internet knowledge. The message (hereinafter referred to as the 'sixth intermediate message') resulting from the operation 614 is displayed by the spelling check module 312 and the misspelled word or phrase to be ignored, the proposed replacement, or the spell checking module ( 312) may include a corrected word or phrase inserted into the message (eg, in place of the misspelled word or phrase). In some examples, the sixth intermediate message may be the same as the fifth intermediate message (eg, when no replacement is performed by the spell checking module 312). According to that embodiment, the sixth intermediate message generated from the spell checking module 312 can be provided to another conversion module (in the conversion module 208) for further processing, or the proposed replacement is spelled The message converted by the inspection module 312 may be provided to the CTT control module 202 to determine whether or not it is in the data storage 210. After operation 614, the sixth intermediate message can be provided to the next operation (eg, operation 616) of the conversion module 208 for processing.

In operation 616, the abbreviation module 314 may identify one or more abbreviations in the chat message, and may further suggest alternate words or phrases (eg, representative) corresponding to the abbreviations. have. In some embodiments, abbreviation module 314 may indicate that the identified abbreviation is omitted or ignored during the next machine translation (eg, by translation module 116). Additionally, in some embodiments, the identified abbreviations may be marked for post-mortem review and disposition by a human operator (eg, manager of CTT system 114). To identify an abbreviation and / or its corresponding word or phrase, in some embodiments, a dataset, including an abbreviation and / or mapping between the abbreviation and their corresponding words and phrases, For example, it may utilize (stored in the data storage). The dataset may be built through a training or learning system, may have an owner (eg, 'internally' collected manually by the administrator of the CTT system 114), or may be commercially obtained. Or, it may come from publicly available Internet knowledge. The message resulting from operation 616 (hereinafter referred to as the 'seventh interim message') is displayed by the abbreviation module 314 and is inserted into the message by the abbreviation, proposed replacement, or abbreviation module 314 to be ignored. May include words or phrases (eg, in place of its identified abbreviations). In some examples, the seventh intermediate message may be the same as the sixth intermediate message (eg, when no replacement is performed by abbreviation module 314). According to that embodiment, the seventh intermediate message generated from the abbreviation module 314 may be provided to another conversion module (in the conversion module 208) for further processing, or the proposed replacement may be an abbreviation module. The message converted by 314 may be provided to the CTT control module 202 to determine whether it is in the data storage 210. After operation 616, the seventh intermediate message can be provided to the next operation (eg, operation 618) of the conversion module 208 for processing.

In operation 618, the profanity module 316 may identify one or more profanity words or phrases (hereinafter, referred to as 'profanity') in a chat message, and may also substitute words or phrases corresponding to the profanity ( For example, a suitable alternative word may be further proposed (eg, a soft euphemism). In some embodiments, profanity module 316 may indicate that the identified profanity is omitted or ignored during the next machine translation (eg, by translation module 116). Additionally, in some embodiments, the identified profanity can be marked for follow-up review and disposition by a human operator (eg, the manager of the CTT system 114). To identify a slang and / or its corresponding word or phrase, in some embodiments, a dataset, including mapping between the slang and / or the slang and their corresponding words and phrases, For example, it may utilize (stored in the data storage). The dataset may be built through a training or learning system, may have an owner (eg, 'internally' collected manually by the administrator of the CTT system 114), or may be commercially obtained. Or, it may come from publicly available Internet knowledge. The message resulting from operation 618 (hereinafter referred to as the 'eighth intermediate message') is displayed by the profanity module 316 and inserted into the message by the profanity, suggested replacement, or profanity module 316 to be ignored. May include words or phrases (eg, in place of the identified profanity). In some examples, the eighth intermediate message may be the same as the seventh intermediate message (eg, when no replacement is performed by the slang module 316). According to that embodiment, the eighth intermediate message generated from the slang module 316 can be provided to another conversion module (in the conversion module 208) for further processing, or the proposed replacement is a slang module The message converted by 316 may be provided to the CTT control module 202 to determine whether it is in the data storage 210. After operation 618, the eighth intermediate message may be provided to the next operation of conversion module 208 for processing. The method 600 then ends.

According to some embodiments, the result message ultimately generated from translation module 208 (eg, the eighth intermediate message generated as a result of operation 618), later, serves as a translation to the result message. It can be used to query the corresponding message to the translation data storage 210. In some examples, a message generated from the transformation module 208 (eg, a message used to query the translation data storage 210 later) is when no transformation is applied to the initial message (eg, It is apparent to those skilled in the art that the initial message may be the same as the initial message received (eg, operation 602) without any transformation applied and passed operations 604-618.

It will be apparent to those skilled in the art that various embodiments may perform more or less operations than described, may perform other operations than those described, and may perform operations in a different order. In general, the type of conversion operation performed and the order in which it is performed can depend on the conversion strategy employed in the embodiment. As mentioned herein, various embodiments may implement different transformation / translation strategies in achieving each translation, which strategy is suitable for a particular translation application or translation context. The conversion / translation strategy employed may determine which conversion operation is performed, when the conversion operation is performed, or the order in which the conversion operation is performed. The translation / translation strategy can also determine when translations are transferred to the translation data store and when the translation data is used in the entire conversion / translation process.

In some embodiments, the intermediate message that is the result of the actions in method 600 may exert an impact and / or cascade effect on the message that is the result of the following actions in method 600. Additionally, in some embodiments, when the chat message is processed by method 600, each operation of flowchart 600 may be performed on the chat message before the method ends. Alternatively, in some embodiments, the method of flowchart 600 may terminate prematurely as soon as only a portion of the illustrated operation is performed (eg, after at least one operation has caused the conversion of a chat message). . According to some embodiments, performing each operation of the flowchart 500 may lead to querying the translation data storage 210 for a corresponding message in a desired language based on the converted message, which is a recent result. If a corresponding message is identified, the method of flowchart 500 may end prematurely.

In various embodiments, method 600 may perform operations 604-612 in parallel. For example, the CTT control module 202 can provide the initial message in parallel in two or more actions 604-612, and receive separate responses from each of the two or more actions. . Each response may include a portion of the displayed text, a suggested replacement, or a word or phrase inserted in the initial message. Thereafter, the CTT control module 202 may select one of the received responses for the next processing (eg, translation data storage 210 query or translation by the translation module 116), a possible selection priority ( For example, it is selected for further processing and according to which precedent can determine which translated message is selected).

For example, during the method 600, the CTT control module 202 may process the initial message, an operation 604 for processing chat term identification, an operation 610 for common noun processing, an operation for abbreviation processing 616 ). In response, operation 604 may return the initial message with the chat term converted, operation 610 may return the initial message with no change, and operation 616 returns the initial message with the abbreviation converted. can do. Next, based on the selection priority, the CTT control module 202 can select the converted message returned from operation 616 for further processing.

In some embodiments, time limits may be applied to perform various operations of method 600. The time limit causes the conversion of method 600 to stop performing if the Response / Result is not received before the time limit expires. If so, various embodiments can ensure that any conversion operation does not unnecessarily interfere with the entire conversion / translation process.

Although the operations of the above-described method are shown and described in a specific order, it is apparent to those skilled in the art that the order in which the operations are performed may vary between embodiments. Additionally, it is apparent to those skilled in the art that the components described in connection with the method of flowchart 600 are only examples of components that may be used in the method, and other components may also be used in some embodiments.

FIG. 7 is a diagram 700 illustrating an exemplary multi-language chat session between chat client systems 104 (eg, 104-1 and 104-2), in accordance with various embodiments. As shown, the chat client system 104-1 includes a chat client GUI module 406-1, and the chat client system 104-2 includes a chat client GUI module 406-2. . As described herein, each chat client GUI module 406-1, 406-2, graphically inputs a chat session shared between them to users of the chat client system 104-1, 104-2, respectively. / Can be configured to provide output access. In some embodiments, the chat client GUI modules 406-1 and 406-2 may provide each user with a client GUI tailored to receive user interactions in relation to the transmitted and received chat conversations.

As the chat conversation 712 (indicated by the double arrow in FIG. 7) passes between the chat client systems 104-1, 104-2, the chat client GUI modules 406-1, 406-2 are each chat client system. A chat conversation 712 can be provided in a language (implicitly or explicitly) selected by the user of (104-1, 104-2). As shown, the chat client GUI module 406-1 provides a chat conversation 712 in a first language (eg, English) to the output area 708 and a first language in the second area 710. And a chat dialog box 702 configured to receive chat input. The chat client GUI module 406-2 provides a chat conversation 712 in a second language (eg, French) to the output area 720 and receives chat input in a second language in the second area 722. And a configured chat dialog box 714. In some embodiments, when a chat conversation 712 is provided in the dialog boxes 702 and 714, an indication of a user name (eg, user online identifier) associated with the user entering a chat message in the chat conversation 712 is provided. It can contain.

In the embodiment of FIG. 7, the language selected in the chat client system 104-1 is English and the language selected in the chat client system 104-2 is French. Therefore, the chat message 704 (“LOL”) and the chat message 706 (“Who u laughin at?”) Are displayed in English in the dialog box 702 of the chat client GUI module 406-1. On the other hand, their respective chat messages 716 (“MDR”) and chat messages 718 (“Qui te fair rire?”) Are displayed in the dialog box 714 of the chat client GUI module 406-2. It is displayed in French. Translation of chat messages 704, 706, 716 and 718 can be accomplished through various systems and methods described herein. More details regarding the translation of messages similar to chat messages 704, 706, 716 and 718 are described with reference to FIGS. 8-10.

8 is a flow chart illustrating the operation of an exemplary multi-lingual communication method 800, in accordance with various embodiments. As described below, in some embodiments, method 800 includes chat client system 104-1, chat client system 104-2, and CTT system 114 (eg, chat server 108). Related actions can be performed. In particular, FIG. 8 describes translating an English chat message containing the text 'LOL' into a French chat message, according to some embodiments. This situation may occur when the language used by the user of the first chat client system 104-1 is English, and the language used by the user of the second chat client system 104-2 is French. According to some embodiments, the CTT system 114 may automatically detect language selection / preferration of the chat client systems 104-1 and 104-2.

As shown, in operation 802, the first chat client system 104-1 provides an English message to be sent to the second chat client system 104-2 (eg, a chat host system ( 112). The English messages can be routed to the CTT control module 202 of the CTT system 114 for translation processing.

In operation 804, the CTT control module 202 queries the translation data storage 210 for a chat message corresponding to the English chat message (“LOL”), which is pre-translated in French. do. In response, in operation 806, the translation data storage 210 returns a French message (“MDR”) corresponding to the English chat message (“LOL”) to the CTT control module 202. Next, in operation 808, the CTT control module 202 helps to send the corresponding French chat message (“MDR”) to the second chat client system 104-2 (eg, the CTT system). (114) sends a corresponding French chat message to the chat host system 112 for transmission).

9 is a flow chart illustrating operation of an exemplary multi-lingual communication method 900, in accordance with various embodiments. In some embodiments described below, the method 900 described through the flowchart is a chat client system 104-1, a chat client system 104-2, a CTT system 114 (eg, a chat server (108) and translation module 116 (eg, translation server 110). In particular, FIG. 9 describes translating an English chat message containing the text “LOL” into a French synonym chat message, according to some embodiments. Unlike the illustrated embodiment of FIG. 8, FIG. 9 shows use for translation module 208 (eg, CTT system 114) and translation module 116.

As shown, in operation 902, the first chat client system 104-1 is directed to a second chat client system 104-2 having a French-speaking user (eg, the chat host system 112). It is possible to provide an English chat message for transmission. The English chat message is routed to the CTT control module 202 of the CTT system 114 for translation processing.

In operation 904, the CTT control module 202 may query the French synonym chat message corresponding to the English chat message ("LOL") to the translation data storage 210. In operation 906, the translation data storage unit 210, as a response, a query failure indicating that the translation data storage unit 210 does not hold the French chat message corresponding to the English chat message (“LOL”) (query failure) to the CTT control module 202. In this case, in operation 908, the CTT control module 202 may deliver the English chat message to the conversion module 208 in order to proceed with conversion processing according to some embodiments. As described herein, the conversion module 208 may include a number of conversion related modules 932 configured to convert chat messages into more suitable messages for further translation processing.

In operation 910, the chat term module 302 of the conversion module 208 may convert the English chat message ("LOL") into a converted English chat message ("laugh out loud"), and further processing To this end, the converted English chat message may be returned to the CTT control module 202. Those skilled in the art will understand that in some embodiments, before the converted English chat message is returned to the CTT control module 202, the English chat message may be processed by an additional module of the conversion module 208.

In operation 912, the CTT control module 202 may query the French synonym chat message corresponding to the converted English chat message ("laugh out loud") to the translation data storage 210. Operation 914 ), The translation data storage 210 is a query indicating that the translation data storage 210 does not have a French chat message corresponding to the converted English chat message ("laugh out loud") as a response. The failure can be returned to the CTT control module 202. In this case, in operation 916, the CTT control module 202 translates the translated English chat message for machine translation processing according to some embodiments. Module 116.

In operation 918, the translation module 116 may reply to a machine translated French chat message (“mort de rire”) corresponding to the converted English chat message. The machine-translated French chat message ("mort de rire") is an example of a translated translation of the English chat-term chat message ("LOL").

In operation 920, the CTT control module 202 converts the machine translated French chat message (“mort de rire”) to the machine translated French message (“mort de rire”), according to some embodiments. ). As mentioned herein, machine translated text can be provided for further conversion processing so that the French translation is further refined. For example, if the original English chat message includes English chat terms, the additional variant processing may add French chat terms to the extent possible. Thus, in operation 922, the chatspeak module 302 of the conversion module 208 converts the machine translated French chat message ("mort de rire") into a converted French chat message ("MDR"). It is possible to return the converted French chat message to the CTT control module 202 for further processing.

Eventually, in operation 924, the CTT control module 202 helps to send the corresponding French chat message (“MDR”) to the second chat client system 104-2 (eg, the CTT system ( 114) sends the corresponding French chat message to the chat host system 112 for transmission). Additionally, in operation 926, the CTT control module 202 translates the translation mapping of the translated translation between the original English chat message ("LOL") and the translated French chat message ("MDR") translation data storage 210 ). Once the mapping is stored in translation data storage 210, the mapping can be used to store translation entries to improve future translation speed, for example, as shown in FIG. . As mentioned herein, the translation data storage 210 may store untranslated translations and mappings of the translated translations.

In some embodiments, CTT control module 202 may also store equivalent (translated and untranslated) translation mappings determined during operation of method 900. In some embodiments, the translation mapping includes chat messages that are not originally stored in the translation data storage 210 (eg, chat messages shown in operation 904 and chat messages shown in operation 912), Mapping between corresponding messages determined during an operation after the search of the translation data storage unit 210 (for example, a query to the translation data storage unit 210 that does not receive a result and a conversion module 208 and / or translation after the query) It may be a mapping between corresponding chat messages determined through module 116).

For example, as shown in FIG. 9, the CTT control module 202 queries the translation data storage 210 for the original English chat message (“LOL”) in operation 904, and in operation 912 , Query the translated English chat message ("laugh out loud") to the translation data storage 210, and the CTT control module 202 (for each of the operations 906 and 914) for both queries, No results are received from the translation data storage 210. However, in operation 916, the CTT control module 202 eventually forwards the translated English message (“laugh out loud”) to the machine translation module 116 for machine translation and in response to operation 918 Receive a machine translated French chat message ("mort de rire"). Accordingly, in operation 928, the CTT control module 202 stores translation data of the translation mapping of the translated translation between the original English chat message ("LOL") and the machine translated French chat message ("mort de rire"). It can be stored in the unit 210. Similarly, in operation 930, the CTT control module 202 translates translation mappings of translation translations between translated English chat messages ("laugh out loud") and machine translated French chat messages ("mort de rire"). It may be stored in the storage unit 210. Through this process, the next method 900 queries the original English chat message ("LOL") or the converted English chat message ("laugh out loud") with the translation data storage unit 210, and the translation data storage unit 210 will provide the corresponding translated translation.

10 is a flow chart illustrating operation of an exemplary multi-lingual communication method in accordance with various embodiments. As described below, in some embodiments, the method 1000 includes a chat client system 104-1, a chat client system 104-2, and a CTT system 114 (eg, chat server 108). ) And the translation module 116 (eg, the translation server 110). In particular, FIG. 10 describes translating an English chat message composed of the text “Who u laughin at?” Into a French chat message, according to some embodiments.

As shown in operation 1002, the first chat client system 104-1 is in English for transmission to the second chat client system 104-2 (eg, via the chat host system 112). You can deliver a chat message. The English chat message is routed to the CTT control module 202 of the CTT system 114 for translation processing.

In operation 1004, the CTT control module 202 may query the French synonym chat message corresponding to the English chat message ("Who u laughin at?") To the translation data storage 210. In operation 1006, the translation data storage unit 210, in response, indicates that the data storage device 210 does not hold a French chat message corresponding to the English chat message (“Who u laughin at?”). The query failure may be returned to the CTT control module 202. In this case, in operation 1008, the CTT control module 202 may deliver the English chat message to the conversion module 208 in order to proceed with conversion processing according to some embodiments. As described herein, the conversion module 208 may include a number of conversion related modules 1036 configured to convert chat messages into messages that are more suitable for further translation processing.

In operation 1010, the chat term module 302 of the conversion module 208 converts the English chat message ("Who u laughin at?") Into a converted English chat message ("Who you laughin at?"). In addition, for further processing, the converted English chat message may be delivered to an additional module of the conversion module 208 such as the spell check module 312.

As discussed, various modules of conversion module 208, including chat term module 302, identify one or more words or phrases in a chat message, and recommend replacement words or phrases for the identified words or phrases It can be configured to. Thus, those skilled in the art will understand that, in some embodiments, the conversion performed / suggested by one module of the conversion module 208 may include a word-to-verse or phrase-to-verse conversion of a chat message. For example, in operation 1010, the chat term module 302 alternatively replaces the phrase “who u” in the conversion process (after the replacement / suggestion to replace the word “laughin” with “laughing”) “who are you” You can convert the English chat message ("Who u laughin at?") Into a converted English chat message ("Who are you laughing at?"). Through this process, various modules of the transformation module 208, such as the chat term module 302, will provide grammatical improvements in each transformation, and to the transformation module 208 to realize grammar improvement. The need for a separate module for this would be eliminated.

In some embodiments, before the English chat message is delivered to an additional module of the conversion module 208, the chat term module 302 may deliver the converted English chat message from operation 1010 to the CTT control module 202. have. Eventually, the CTT control module 202 may query the French synonym chat message corresponding to the converted English chat message (“Who you laughin at?”) To the translation data storage 210 (in operation 1012). In operation 1014, the translation data storage 210 stores, in response, the French chat message corresponding to the converted English chat message (“Who you laughin at?”). The query failure indicating not being performed may be returned to the CTT control module 202.

In operation 1016, the spell check module 312 may, in operation 1018, perform spell checking processing on the converted English chat message (“Who you laughin at?”). While performing the spell check processing, the spell check module 312 may correct the converted English chat message into a corrected English chat message ("Who you laughing at?"), And the corrected English chat message. It is possible to reply to the CTT control module 202. Those skilled in the art will understand that in some embodiments, the corrected English chat message can be processed by an additional module of the conversion module 208 before the converted English chat message is returned to the CTT control module 202.

In operation 1020, the CTT control module 202 may query the translation data storage 210 for a French synonym chat message corresponding to the corrected English chat message (“Who you laughing at?”). At 1022, the translation data storage 210 does not hold the French chat message corresponding to the corrected English chat message ("Who you laughing at?") As a response. A query failure indicating C may be returned to the CTT control module 202. In this case, in operation 1024, the CTT control module 202 may revise the corrected English chat for machine translation processing according to some embodiments. The message can be delivered to the translation module 116.

As mentioned herein, the machine translated text can be delivered for further conversion processing to further refine the text translation. For example, if the original English chat message includes English chat terms, the additional conversion processing may add French chat terms to the extent possible. In operation 1030, the conversion module 208 is configured for further processing (e.g., if the module of the conversion module 208 has not applied any changes to the machine translated French chat message), the unchanged machine The translated French chat message (“Qui te fait rire?”) May be returned to the CTT control module 202.

In operation 1032, the CTT control module 202 assists in sending a machine translated French chat message (“Qui te fait rire?”) To the second chat client system 104-2 (eg, CTT). The system 114 delivers the corresponding French chat message to be transmitted to the chat host system 112). In addition, in operation 1034, the CTT control module 202 translates the translation mapping between the original English chat message ("Who u laughin at?") And the translated French chat message ("Qui te fait rire?"). It may be stored in the storage unit 210. As described herein, in additional operations (not shown in the figure), the CTT control module 202 may also query previously failed translation data storage 210 and this query (eg, the operation of FIG. 9). (Similar to 928 and operation 930), the synonym translation mapping based on the determined corresponding message may be stored in the translation data storage 210.

According to some embodiments, the conversion operation performed by the conversion module 208 may include performing a predetermined conversion operation in parallel, and may include performing a predetermined conversion operation in series. The order of conversion operations performed in parallel and serially may vary between various embodiments. As described herein, some embodiments in which the conversion operation is performed in parallel may adopt the priority of selection to determine which conversion message is selected for further conversion and according to some precedent.

11 is a flow chart illustrating operation of an exemplary multi-lingual communication method in accordance with various embodiments. In some embodiments described below, the method 1100 includes a chat client system 104-1, a chat client system 104-2, a CTT system 114 (eg, of the chat server 108), and An operation related to the translation module 116 (eg, of the translation server 110) may be performed. In particular, FIG. 11 describes translating an English chat message containing the text “Who u laughin at?” Into a French chat message through a parallel conversion operation, according to some embodiments.

As shown in operation 1102, the first chat client system 104-1 is an English language that is intended to be transmitted (eg, via the chat host system 112) to the second chat client system 104-2. You can deliver a chat message. The English chat message can be routed to the CTT control module 202 of the CTT system 114 for translation processing.

In operation 1104, the CTT control module 202 may query the French synonym chat message corresponding to the English chat message ("Who u laughin at?") To the translation data storage 210. In operation 1106, the translation data storage unit 210, in response, indicates that the translation data storage unit 210 does not have a French chat message corresponding to the English chat message (“Who u laughin at?”). The query failure may be returned to the CTT control module 202.

In this case, the CTT control module 202 may deliver the English chat message to the conversion module 208 to perform conversion processing according to some embodiments. As described herein, the conversion module 208 may include a plurality of conversion related modules 1130 configured to convert chat messages into messages that are more suitable for further translation processing. 11, during operation 1108, the CTT control module 202 may deliver the English chat message in parallel to two or more conversion related modules 1130 included in the conversion module 208. have. In addition, during operation 1108, the CTT control module 202 receives results from the transformation related modules 1130 in parallel, and forwards queries to the translation data storage 210 in parallel based on the transformation results. You can.

Accordingly, in operation 1110a, the CTT control module 202 may transmit the English chat message (“Who u laughin at?”) To the chat term module 302 for conversion processing. In parallel, in operation 1110b, the CTT control module 202 may deliver the English chat message ("Who u laughin at?") To the spell check module 312 for conversion processing. Subsequently, the CTT control module 202 may receive the first converted English chat message ("Who you laughin at?") From the chat term module 302 in operation 1112a, and simultaneously operate 1112b In CTT control module 202 may receive a second translation English chat message ("Who u laughing at?") From the spell check module 312. According to each conversion processing time, the chat term module 320, the spell check module 312, and other conversion related modules 1130 may respond to each other, serially or in parallel, with the CTT control module 202. have.

Subsequently, in operation 1114a, the CTT control module 202 queries the translation data storage 210 for a French synonym chat message corresponding to the first translated English chat message ("Who you laughin at?"). Can be. In operation 1114b, the CTT control module 202 may query the translation data storage 210 for a French consent chat message corresponding to the second converted English chat message ("Who u laughing at?"). . For some embodiments, during operations 1114a and 1114b, the CTT control module 202 may query the translation data storage 210 serially or in parallel. In some embodiments, the query timing may depend on when each of the transformation related modules 1130 of the transformation module 208 returned a response. As shown in FIG. 11, the translation data storage unit 210 may return a query failure (eg, <FAIL>) for the query in operations 1116a and 1116b.

As a result, the CTT control module 202 can select one converted message from two or more messages resulting from the parallel operation 1108 for further processing. When any one of the conversion-related modules 1130 generates a converted message, the CTT control module 202 can select the specific converted message for further processing. As mentioned herein, the CTT control module 202 may select the converted message based on the selection priority, and the selection priority may be determined according to the conversion / translation strategy selected by the embodiments. In some embodiments, the priority of the selection is whether the converted message has the most formal content, whether the converted message has the best conversions, or whether the converted message is suitable for machine translation. It can be based on whether the result is from a conversion related module known to have a high probability of generating a message.

If any one converted message is selected, in operation 1118, the CTT control module 202 may deliver the converted English chat message to the translation module 116 for machine translation processing according to some embodiments. For example, as shown in FIG. 11, the CTT control module 202 generates a first translated English chat message ("Who you") generated by the chat term module 302 for delivery to the translation module 116. laughin at? ").

In operation 1120, the translation module 116 (even though it contains the misspelled word “laughin”) corresponds to the first translated English chat message and a machine translated French chat message (“Qui te fait rire”) ? "). In operation 1122, the CTT control module 202 converts the machine translated French chat message (“Qui te fait rire?”) For further translation processing of the machine translated French chat message in accordance with some embodiments. 208).

As described herein, the machine-translated text can be provided as an additional conversion process to further refine the text translation. For example, if the original English chat message includes English chat terms, further modification processing may add French chat terms to the extent possible. In operation 1124, the translation module 208, for further processing (eg, if the module of the translation module 208 did not apply any changes to the machine translated French chat message), the unchanged machine The translated French chat message (“Qui te fait rire?”) May be returned to the CTT control module 202.

In operation 1126, the CTT control module 202 helps to send a machine translated French chat message (“Qui te fait rire?”) To the second chat client system 104-2 (eg, CTT). The system 114 delivers the corresponding French chat message to be sent to the chat host system 112). Additionally, in operation 1128, the CTT control module 202 translates the translation mapping between the original English chat message ("Who u laughin at?") And the translated French chat message ("Qui te fait rire?"). It can be stored in the storage 210. As described herein, in additional operations (not shown in the figure), the CTT control module 202 may query previously failed with the translation data storage 210 and this query (e.g., the operation of FIG. 9 ( Similar to 928 and 930), synonym translation mapping based on the corresponding determined message afterwards may also be stored in the translation data storage 210.

In some embodiments, conversion operations may be performed in a hybrid serial / parallel fashion, or some conversion operations may be performed in parallel and other conversion operations may be performed in series. For example, as illustrated in FIG. 11, the English chat message (“Who u laughin at?”) Is performed in parallel in the operation 1110a and operation 1110b, the chat term module 302 and the spell check module ( 312). Subsequently, if any one of the resulting converted messages is selected (e.g., based on the selection priority), other conversion related modules 1130 of the conversion module 208 (e.g., acronym module ( 304), the proper noun module 306, and the common noun module 308 may process the selected conversion message in parallel.

12 is a block diagram of an exemplary digital device 1200. The digital device 1200 is communicatively connected via a bus 1214, a processor 1202, a memory system 1204, a storage system 1206, a communication network interface 1208, an I / O interface 1212 and a display interface 1212. The processor 1202 is configured to execute executable instructions (eg, programs). In some embodiments, the processor 1202 includes any processor or circuit capable of processing the executable instructions.

The memory system 1204 is any memory configured to store data. An example of the memory system 1204 is a storage device such as RAM or ROM. The memory system 1204 may include a ram cache. In various embodiments, data is stored in memory system 1204. Data stored in memory system 1204 may be deleted or ultimately transferred to storage system 1206.

The storage system 1206 is any storage configured to retrieve and store data. Examples of the storage system 1206 are flash drives, hard drives, optical drives and / or magnetic tapes. In some embodiments, the digital device 1200 includes a memory system 1204 formed of RAM and a storage system 1206 formed of flash data. Both the memory system 1204 and the storage system 1206 may include computer readable media storing instructions or programs executable by a computer processor including the processor 1202.

The communication network interface 1208 (com. Network interface) may be coupled to a network (eg, the computer network 106) via a link 1216. The communication network interface 1208 may support communication through, for example, Ethernet connection, serial connection, parallel connection, or ATA connection. In addition, the communication network interface 1208 may support wireless communication (eg, 802.11 a / b / g / n, WiMAX). It will be apparent to those skilled in the art that the communication network interface 1208 can support many wired or wireless standards.

The optional input / output (I / O) interface 1210 is any device that receives input from a user and outputs data. The optional display interface 1212 is any device configured to output graphics and data to the display. As an example, the display interface 1212 is a graphics adapter.

Those skilled in the art will understand that the hardware components of the digital device 1200 are not limited to those shown in FIG. 12. The digital device 1200 may include more or fewer hardware components than shown. In addition, hardware components can be maintained and share functionality within the scope of the various embodiments described herein. As an example, encoding and / or decoding may be performed by the processor 1202 and / or a coprocessor located in the CPU (eg, Nvidia®).

The functions and components described above may be composed of instructions stored in a storage medium such as a computer readable medium. The instructions can be retrieved and executed by the processor. Some examples of instructions are software, program code and firmware. Some examples of storage media are memory devices, tapes, disks, integrated circuits, and servers. The instructions operate when executed by a processor, in accordance with some embodiments, to direct the operation to the processor. Those skilled in the art are familiar with instructions, processor and storage medium.

Various embodiments are described herein as examples. It will be apparent to those skilled in the art that various modifications can be made and other embodiments can be used without departing from the scope of the present invention. It should be understood that various modifications according to the exemplary embodiments belong to the present invention.

13 is a block diagram illustrating an exemplary user feedback system according to various embodiments. For purposes of explanation, the system for user feedback shown in FIG. 13 shows a portion of an exemplary communication conversion and translation (CCT) system 1300 according to various embodiments. For example, an exemplary user feedback system is a chat translation system employed in various multi-language chat systems, including a game chat system (eg, an in-game chat system) available with MMO games. It can be part of. With the use of an exemplary user feedback system, various players in an MMO game may receive user feedback for a displayed word or phrase, in exchange for an in-game currency / credit transaction / item that is an incentive for feedback / approved feedback. For example, through an in-game chat system, you can choose to provide. It will be appreciated by those skilled in the art that, in accordance with some embodiments, some or all of the exemplary user feedback system, and components related thereto, may be separated and / or independent from the exemplary communication translation and translation system 1300.

As used herein, “feedback” should be understood to indicate “user feedback” or “user feedback response” that responds as much as possible to queries requesting feedback for translation or translation. It should also be understood that the user feedback may include user feedback for translation, or user feedback for translation. User feedback may include a definition for a particular word or phrase, which definition is: (i) the particular word or phrase in the first language (same) in that particular word or phrase in the first language. Allow conversion to the corresponding word or phrase, and (ii) allow the given word or phrase to be converted from that particular word or phrase in the first language to the corresponding word or phrase in the (other) second language Do so, thereby performing a translated translation of a word or phrase; And / or (iii) allow that particular word or phrase to be translated from a first language to a second language, thereby performing an untranslated translation of the word or phrase. As described herein, a user feedback response may be received in response to a query requesting user feedback in connection with a translation failure.

As used herein, “failure to translate” or “failure to translate” is a substantial failure in translating the text (eg, translated text is the same as the original text), or potentially a translation failure. It can be understood to include text translations marked as including (eg, displayed by a user who reads translated text and believes that the translated text contains a suspicious or incorrect translation).

As shown in FIG. 13, the CTT system 1300 includes a communication conversion and translation (CTT) control module 1302, a communication conversion and translation (CTT) communication module 1304, a language module 1306, and a conversion module ( 1308), a translation data storage unit 1310 and a translation API module 1312. In some embodiments, CTT control module 1302, CTT communication module 1304, language module 1306, translation module 1308, translation data storage 1310 and translation API module 1312 are described herein. As may be, it may be similar to the components of the CTT system 114.

As shown in FIG. 13, the CCT system 1300 includes a translation failure management module 1314, a user feedback (UF) query generation module 1316, a user feedback (UF) query crowd selection module 1318, and a query / response. It may include an evaluation module 1320, a query API module 1322, a response evaluation module 1324, a translation data storage update module 1326, an audience capability evaluation module 1328, and an incentive reward module 1330.

The translation failure management module 1314 may be configured to enable management of translation failures (eg, failure to translate text from a first language to a second language). In some embodiments, the translation failure management module 1314 attempts to translate the chat message sent by the system 1300 from a first user who speaks a first language to a second user who speaks a second language during a chat session. , It may be suitable to handle one or more translation failures that may be experienced by the CCT system 1300. As described herein, text to be translated by the CCT system 1300 may be a chat message that may include chat terms, abbreviations, colloquial expressions, idioms, and the like. During the conversion or translation of a chat message, some or all of the chat message may result in a translation failure, due to the inability of the CCT system 1300 to transform and / or translate the portion that causes the chat message to fail to translate. I can understand.

Features provided by the translation failure management module 1314 may include, without limitation: automatically detecting when a failure to translate the text occurs; Automatically detecting certain words or phrases in the text when they are not translated; Allowing the user to display some or all of the “translated” text containing substantial or potential translation errors after the translation process; Managing which translation failures are selected for user feedback from various audience members (eg, chat members such as players using an in-game chat system); Managing which words or phrases associated with translation failures are selected for user feedback (eg, based on how often / when words are used / in the text to be translated); And managing when translation failure is no longer a selection candidate for user feedback (eg, words or phrases associated with translation failure are removed from the selection pool).

In various embodiments, the translation failure management module 1314 may manage the overall availability of words or phrases for user feedback of audience members. As used herein, “user feedback opportunity” may include words or phrases available for user feedback of spectator members. The translation failure management module 1314 may generate one or more user feedback opportunities based on various conditions such as availability of computing resources for user feedback, incentives compensated in connection with user feedback, and quality of feedback responses received in the past. You can defer or postpone availability. For example, continuing to reward incentives (eg, in-game currency or in-game items) for an approved user feedback response will have a negative impact on the economy associated with incentives (eg, in-game economy). You can decide if you can give. Accordingly, some embodiments may (temporarily) postpone the availability of user feedback opportunities to reward incentives when detecting that current incentive strategies are detrimental to the relevant economy (eg, in-game economy). Or adjust the rewarded incentives for user feedback opportunities.

As described herein, an audience member for whom user feedback is requested is a person who volunteers to provide such feedback, is a person like a participant in the user feedback program, and / or when their submitted feedback has been approved (eg For example, it may be a person who volunteers to provide such feedback when corrected, accepted, or in exchange for an incentive (useful for defining a translation or translation). In some embodiments, the user feedback system described herein may be part of a gaming system, such as an online MMO game, in which the spectator members may find useful incentives (eg, in-game currency or in-game) in the gaming system. For game items), game players have chosen to participate in user feedback opportunities. In this way, the user feedback system can be provided as a game feature that the game-playing spectator member regards as a "game" "played" for rewards, thereby expanding competition inherent in the game environment. More on spectator member selection is described with respect to UF query spectator selection module 1318.

As described herein, a word or phrase may be indicated by a communication system user who believes that the word or phrase of interest may interfere with translation or cause an incorrect translation of the communicated text. For example, in a multi-language multi-user chat system associated with an online game, game players display the entire chat message they have received, potentially having translation problems or failures, or certain parts of the translation chat message (eg , Words or phrases). Words or phrases associated with translation failures include, for example, specialized / domain related terminology, abbreviations, acronyms, proper nouns, common nouns, idioms, colloquial words or phrases, and profanity words or phrases You can. Additionally, a word or phrase can be displayed by a system or method that automatically detects that a word or phrase cannot be translated and can interfere with the translation of a larger phrase or sentence.

The translation failure management module 1314 may be responsible for selecting words or phrases that are the subject of user feedback, where the selected words or phrases may cause actual or potential failures to translate text from a first language to a second language. Can be displayed when associated. The translation failure management module 1314 can select a word or phrase based on a number of factors, some of which are based on the current importance of the word or phrase in translation (eg, the overall usage of the word or phrase) Based importance), the complexity of the word or phrase (e.g., the difficulty of the word or phrase, or the period during which the word or phrase was an issue), and / or the ability of the user to choose / volunteer to provide user feedback (e.g. For example, the user has a language ability for the second language), and / or a preference of the user who is / are selected to provide user feedback. It will be appreciated by those skilled in the art that other factors for selecting words or phrases for user feedback can be utilized by various embodiments.

In various embodiments, users selected to provide user feedback may be people who volunteer to provide such feedback. For example, a user may set an associated user preference, select an in-chat system advertisement requesting user feedback, and / or list one or more user feedback opportunities available for selection. By browsing through sections, you can choose to provide user feedback.

As described herein, the translation failure management module 1314 may manage when a particular word or phrase associated with actual or potential translation failure is no longer a selection candidate for user feedback from spectator members. Various conditions can lead the translation failure management module 1314 to make such a decision. The decision may be made, for example, when a specific number of feedback responses are received in connection with the associated translation failure or in connection with a particular word or phrase; When a certain number of consistent feedback responses are received in connection with the associated translation failure or in connection with a particular word or phrase; And when a specific feedback response has been approved as valid for the user feedback that has been searched (eg, from a set of unique feedback responses, a specific response is manually approved by the administrator as an accurate response).

The UF query generation module 1316 may be configured to generate a query for obtaining user feedback for a specific word or phrase selected for user feedback from a selected audience member. As described herein, selected spectator members, user feedback, in return for incentives when submitted user feedback has been approved as being accurate, accepted feedback response, or useful for purifying controversial translations It may be a user who volunteers to provide. More about spectator member selection is described in relation to UF query spectator selection module 1318.

In some embodiments, the generated query includes one or more methods for receiving query responses from selected audience members. For example, the generated query, as a response method, includes a predefined list of responses, and the audience member from the list of responses to the generated query (also referred to herein as “select-form response”). You can choose as their response. In another example, the generated query includes, as a response method, one or more fields configured to receive as their responses to the generated query, and text values are entered into the fields by an audience member (also here. Referred to as "pre-form response"). Other response methods may include a graphical user interface (GUI), text values, or some combination thereof.

The one or more responses included in the generated query may be subject to a number of factors including, for example: audience member preferences; The importance and / or complexity of the word or phrase for which user feedback is requested; The number of user feedback responses received so far for the word or phrase for which user feedback is requested; The number of consistent feedback responses received so far for the word or phrase for which user feedback is requested; And whether there is enough pre-form response to generate a select-form response. In some embodiments, the spectator member responding to the generated query can select between two or more response methods to the generated query.

Additionally, user feedback opportunities may be provided depending on the language of the audience member, and may be subject to a number of factors including, for example: whether the audience member is multilingual by monitoring the audience member's chat history; Whether the audience member's language ability meets or exceeds a certain level of language confidence; And the language set on the user device by the audience member to use to participate in the user feedback opportunity (eg, the default language set on the audience member's device). For example, the default device language of a spectator member's device is German, and if the spectator member chooses to participate in a user feedback opportunity, the query generated and sent to the spectator member may be based on German (eg , Query to define an English word or phrase as a German word or phrase, or vice versa). In some embodiments, the generated query provides an option to select the language of the response to be submitted (eg, when the generated query does not specify or require a specific language) and / or two or more responses. Each response can offer different language options to audience members. In some such embodiments, the audience member may be provided with language response options according to various language factors, such as whether the audience member is multilingual and the language set on the user device of the audience member, as described herein.

The UF query audience selection module 1318 may be configured to select an audience member to request user feedback. In some embodiments, UF query spectator selection module 1318 is a spectator member in a pool of spectator members who volunteer to provide user feedback in exchange for incentives (which may be received when user feedback is approved / accepted). You can choose As described herein, a spectator member can be a volunteer who participates in a translation, and at that time, the spectator member is a pool of spectator members that the UF query audience selection module 1318 selects for user feedback requests. Can be included in In some embodiments, when the UF query spectator selection module 1318 selects a spectator member, a query generated by the UF query generation module 1316 may be provided to the spectator member to obtain user feedback. The query generated by the UF query generation module 1316 may be provided to an audience member selected by the UF query audience selection module 1318 as part of a list of user feedbacks that can be participated by the selected audience member. Once the generated query is provided to the selected audience member, the selected audience member initiates a user feedback session, and the generated query is provided to the audience member in the user feedback session (e.g., the generated audience member) Have one or more response methods associated with the query), and the selected audience member can provide one or more responses (eg, via one or more response methods included in the generated query) .

According to an embodiment, when a word or phrase for user feedback is selected before an audience member is selected, the audience member is based on the selected word or phrase (eg, the selected word or phrase is the ability of the audience member) Or whether it matches the preference). Alternatively, if an audience member is selected before a word or phrase for user feedback is selected, the selected word or phrase may be based on the selected audience member. The ability of a particular spectator member may be determined based on one or more feedback responses previously provided by the particular spectator member (eg, with respect to a query previously created and provided to that particular spectator member), and And / or based on one or more language skills of the audience member. Audience members can be evaluated, for example, to see if they have the ability to provide user feedback on one or more words or phrases associated with a translation failure when they achieve a certain number of approved feedback responses. . In another example, a spectator member is in a particular language based on at least a feedback response to a specific language submitted previously, the spectator member's preferences, and / or information related to the spectator member in their particular language. You can be evaluated for your ability. In another example, a spectator member rated with a previously submitted feedback response being incorrect or deceiving (eg, an absurd or fictional response) may be determined to have a low ability.

In some embodiments, if a particular audience member provides a response related to a word or phrase (and the response is approved as possible), that particular audience member may not be selected again to provide user feedback for the same word or phrase . In certain embodiments, an exception to this limitation may be given when the particular spectator provides a response in a language different from that of a response previously submitted for the same word or phrase.

When an incentive is provided as a reward for user feedback (e.g. as soon as the feedback response is approved), various embodiments are based on time-based quotas (e.g. hourly, daily, monthly, for submitting feedback responses) Audience members may be selected based on quotas, such as limits per year) or incentive-income quotas (eg, limits on in-game currency or items given for incentive-based user feedback). The quarter is then associated with an individual spectator member, spectator member group, or some combination thereof. In some embodiments, the UF query spectator selection module 1318 is based on various conditions such as availability of computing resources for user feedback, rewarded incentives in relation to user feedback, quality of feedback responses received in the past, or the like. The availability of user feedback opportunities may be suspended or postponed for further audience members.

The query / response (QR) evaluation module 1320 is configured such that the query or query response can determine the value or importance of the query or query response based on the word or phrase requesting user feedback. Examples of factors that are considered by various embodiments when determining the value of a query or question answer include word or phrase complexity (eg, the higher the complexity, the higher the value), the word or phrase in the translation / translation process. The importance of (for example, the higher the value, the higher the value), the response method employed by the query or query response (eg, the value of the pre-form response method is higher than the select-form response method), Types of words or phrases (eg, chat terms, abbreviations or colloquial), or queries involving one or more involved languages (eg, English words or phrases and receiving a French response), the English words or phrases And higher value than a query to receive an English response). In some embodiments, incentives rewarded for specific user feedback may be based on a value associated with a query or query response. In some embodiments, QR evaluation module 1320 values the value of a query or query response based on the efficacy of previous queries or query responses in soliciting beneficial, useful, or accurate user feedback. It can be configured to determine. Accordingly, in some embodiments, the value can be dynamically adjusted based on recent efficacy achieved by previous queries or query responses. In certain embodiments, the QR evaluation module 1320 may be configured by an administrator (eg, of the CCT system 1300), or other authorized user to manually respond to queries and / or query responses associated with one or more user feedback. It is possible to allocate or adjust the value of

The query API module 1322 may be configured to facilitate sending user feedback queries to selected audience members and / or receiving query responses from the selected audience members. In some embodiments, query API module 1322 may provide a list of user feedback opportunities available to the selected audience member, and provide queries related to one or more user feedback opportunities selected by the selected audience member. And receive a response to one or more user feedback opportunities selected by the selected audience member, and / or the current status of the query response ever submitted for approval (eg, approved, rejected or pending) / Awaiting approval status) to the selected audience member. In some embodiments, query API module 1322 may obtain one or more preferences associated with the particular audience member from the chat client system used by that particular audience member interacting with CCT system 1300. As described herein, preferences associated with a spectator member can determine a selection of a word or phrase for which user feedback is requested from the spectator member, and / or the spectator member receives a query for user feedback of a specific word or phrase You can decide whether or not it is selected.

The response evaluation module 1324 may be configured to evaluate the submitted query response in association with the generated query to obtain user feedback. In some embodiments, the response evaluation module 1324 may evaluate query responses in one or more steps.

During the validation phase, the response evaluation module 1324 disposes one or more unique responses, such as approval or rejection of the particular generated query, submitted for that particular generated query. can do. In some embodiments, a unique response submitted in relation to a query generated to obtain user feedback is either approved as valid for the word or phrase associated with the generated query, or a word associated with the generated query, or Until it is rejected as invalid for the passage, it may be considered pending approval. As used herein, a particular “unique response” may not be an exact match in syntax, but may include a set of similar (eg, different punctuation or spacing) responses, received via two or more response methods. Can be. Native responses may also be referred to herein as “unique query responses” and “unique user feedback responses”. According to an embodiment, one or more unique query responses may be approved for a particular generated query. For example, one or more unique query responses may be approved when defining a specific word or phrase related to the user response being sought. Responses can be passively evaluated, for example, by an administrator reviewing and deploying responses (eg, only unique responses possible). Additionally, or alternatively, the response can be evaluated by an automated process that approves and / or rejects the submitted response based on the response's count and / or threshold associated with the response count. In some embodiments, an automatic validation process can filter the best submitted responses for a particular word or phrase, and provide the best submitted responses to managers, etc. for review and placement. The verification phase may be performed by the response evaluation module 1324 periodically (eg, based on a schedule) or condition (eg, when the number of submitted responses awaiting approval satisfies or exceeds a threshold). You can.

In some embodiments, when a plurality of unique user feedback responses (eg, definitions) are provided for a specific word or phrase, the response evaluation module 1324 is the highest rated unique user feedback among the plurality of unique user feedback responses Responses (eg, the most popular responses) can be approved automatically (and / or via manual admin input), but only when the highest rated unique user feedback response is statistically important. For example, when a first unique user feedback response has been received 26 times for a particular word, and a second unique user feedback response has been received 24 times for that same particular word, the first unique user feedback response, although it is Even a unique user feedback response rated best for that particular word may not be accepted in response to that particular word. This is because 26 is not statistically significant for 24. The first unique user feedback may not be selected, for example, until the first unique user feedback response remains the highest rated unique user feedback response and the response reaches 40.

Considering that the verses are not the same and can be substantially similar and convey the same intent, in some embodiments, a Word Error Rate (WEB) can be used to group unique user feedback responses that contain the verse. . For both passages, the word error rate can measure the replacement, deletion, and insertion of words to convey similarity between passages.

In various embodiments, if multiple unique user feedback responses (eg, definitions) are provided for a particular phrase, the response evaluation module 1324 automatically (and / or through manual administrator input) is the best rated unique User feedback responses can be approved.

During the check phase, the response evaluation module 1324 may determine whether the response submitted by the spectator member is placed in an approval, rejection or review pending (eg, pending approval). In some embodiments, the data storage unit may maintain a state as to whether a unique response submitted by an audience member with respect to a particular word or phrase has been approved or rejected as a valid definition for that particular word or phrase. Accordingly, the checking step may determine the placement of the response submitted for the specific word or phrase by consulting with a data storage that maintains the placement of the unique response previously submitted for the specific word or phrase; The submitted response shares the deployment status of the unique response corresponding to the submitted response. According to an embodiment, the cheat step for a submitted response may be performed immediately or soon after the response has been submitted. If it is determined that the submitted response is still pending pending during the check step, the check step may be re-performed at a time following a validation step that affects the status of the same or similar submitted response. The status of the submitted response can be updated according to the current batch of the submitted response determined during the check phase. As described herein, the current status of one or more responses submitted by a particular audience member can be provided as a list reflecting the current status for those responses. More about the response status will be described later in conjunction with FIG. 28.

The response evaluation module 1324 can include determining whether the response is approved. The response may be approved based on at least one previous response provided by the other person in response to another query, the other query being previously generated to obtain feedback on the word or phrase from the other person. If the response is determined to accurately define a word or phrase, the response can be approved.

Translation data storage update module 1326 updates translations or translations (eg, stored in translation data storage 210) based on evaluation of responses submitted to user feedback queries by selected audience members. It can be configured to. For example, when a response submitted in the first language is determined during the response evaluation to an approved definition for a specific word in the (same) first language, that specific word in the first language and the first language Transforms that map responses can be added or updated. In another example, a response submitted in a second language maps the response in a second language to that specific word in a first language when a response is determined by an approved definition of a specific word in the first language during response evaluation. The conversion to be done can be added or updated. The translation or update of the translation may be performed by the translation data storage update module 1326 during or after the check step in which the disposition of the submitted response is changed to approval.

The audience capacity assessment module 1328 may be configured to determine a crowd member's ability to indicate a confidence level associated with the audience member ability to provide an accurate and / or useful user feedback response. As described herein, the ability of a particular spectator member may be determined based on one or more feedback responses previously provided by a particular spectator member, and / or one or more language abilities of the spectator member. For example, an audience member can be evaluated to see if that audience member has the ability to provide user feedback for one or more words or phrases associated with a translation failure when a certain number of approved feedback responses have been achieved. In another example, an spectator member is assessed to have a proficiency in a particular language based at least on feedback responses previously submitted in a particular language, spectator member preferences, and / or information related to the spectator member indicating the fluency of that particular language. Can be. In another example, a spectator member with a previously submitted feedback response that has been evaluated as inaccurate or deceptive (eg, gibberish response) may be determined to have a low ability.

The incentive reward module 1330 may be configured to reward an incentive member to an audience based on an evaluation of a response submitted by the audience member in relation to a query for user feedback. As described herein, if the submitted response is approved, the spectator member can be rewarded with incentives. The amount or type of rewarded intent can be determined based on a number of factors including, but not limited to: The factor is the value of the query or query response assigned by the QR evaluation module 1320, the response method used by the spectator member in response to the query, already (e.g., related to a specific word or phrase or incentive based user feedback) The amount of incentives rewarded to that spectator member or to all spectator members via, the language of the query or the language of the response provided, and the type of word or phrase in which the response was submitted (eg, chat terms, abbreviations or specific domain words or Verse). Rewarded incentives include real world currencies, or virtual currencies such as in-game currencies or in-game items that may or may not have value outside the associated virtual economy (e.g., currency values in the real world economy). can do. In some embodiments, incentives may include real world goods or services or virtual goods or services that may have an associated currency value. It will be appreciated by those skilled in the art that other forms of incentives may be compensated in other embodiments.

In some embodiments, the incentive reward module 1330 notifies the spectator member when one or more responses submitted by the spectator members are approved, and / or when the submitted response is rewarded by an approved spectator member. You may be responsible. In various embodiments, the incentive reward module 1330 notifies spectator members of the incentive rewards through an update to the list of statuses for notification messages (eg, in-chat messages such as pop-up messages) and / or submitted responses. You can.

In various embodiments, a system for user feedback may include more or fewer components than those illustrated in FIG. 13, and each component illustrated in FIG. 13 is described for each component It will be apparent to those skilled in the art that more or fewer operations can be performed.

14 is a block diagram illustrating an exemplary user feedback client system according to various embodiments. For illustrative purposes, the user feedback client system in FIG. 14 is shown as part of a chat client system 1400 according to various embodiments. For example, an exemplary user pigback client system may be part of a game chat client system (eg, an in-game chat client system) available with MMO games, and various players of the MMO game, if possible, To reward in-game currency / credits / items as an incentive for feedback, one can choose to provide user feedback for the identified word or phrase. It may be appreciated by those skilled in the art that in some embodiments, the exemplary user feedback system, some or all of the components associated therewith, can be separated from the exemplary communication conversion and translation system 1300.

As illustrated in FIG. 14, the chat client system 1400 may include a chat client control unit 1402, a chat client communication module 1404, and a chat client GUI module 1406. In some embodiments, chat client control unit 1402, chat client communication module 1404, and chat client GUI module 1406 may be similar to components of chat client system 104, as described herein. .

As shown in FIG. 14, the chat client system 1400 includes a conversion / user feedback (UF) query preference module 1408 and a conversion / user feedback (UF) query graphical user interface (GUI) module 1410. can do. In some embodiments, UF query preference module 1408 and / or UF query GUI module 1410 facilitate user feedback interaction with respect to CTT system 1300. In the context of chat client system 1400, a chat user of chat client system 1400 may be an audience member associated with a user feedback system (eg, CTT system 1300) of various embodiments.

The UF query preference module 1408 allows chat users to preview, redefine, and / or adjust preferences regarding user feedback features provided in connection with a user feedback system (eg, CTT system 1300) of some embodiments. It can be configured to manage and enable what to do. Examples of preferences managed by the UF query preferences module 1408 include, for example, language preferences associated with user feedback (eg, the language of the word or phrase in which the user feedback is requested and / or the language of the user feedback found). Preferred response methods for user feedback queries (e.g., select-form responses rather than pre-form responses), or preferred word or phrase types (e.g., abbreviations, chat terms, physics related or idioms), and the like.

As used herein, a select-form response is a response that is predetermined and selectable from at least two or more select-form responses. According to the embodiment, the list of select-form responses may allow audience members to select two or more responses when applicable. The pre-form response is a response that includes text-based values (eg, character values, or string values) entered in the field by an audience member.

The UF query GIU module 1410 receives a response from a chat user via an indication of a query generated for user feedback and provided to the chat user, an indication of one or more response methods associated with the query, and / or the displayed response method It makes it possible to do it graphically. The UF query GUI module 1410 may also enable management of preferences through the UF query preference module 1408. More on the graphical user interface is described below in relation to FIGS. 17-23 and 24-31.

In various embodiments, the chat system for user feedback may include more or fewer components than those illustrated in FIG. 14, and each component illustrated in FIG. 14 is for each component It is apparent to those skilled in the art that it is possible to perform more or less actions than the described actions.

15 is a flowchart illustrating an example method for user feedback according to various embodiments. At step 1502, translation failure management module 1314 may identify potential failures in the translation or translation of text, for example, from a first language to a second language. At step 1504, translation failure management module 1314 also selects a word or phrase for user feedback from the identified potential failure. In step 1506, the UF query spectator selection module 1318 may select spectator members to request user feedback. In step 1508, the UF query generation module 1316 generates a query to obtain user feedback from the selected audience member. In step 1510, the response evaluation module 1324 may receive a response to the generated query. The query API module 1322 may be responsible for providing the generated query to the selected audience member and receiving a response to the generated query. In step 1512, the response evaluation module 1324 can evaluate the received response. In step 1514, the spectator ability evaluation module 1328 evaluates the ability of the selected spectator member based on the evaluation provided in step 1510 and / or the received response performed in step 1512. You can. In step 1516, the incentive reward module 1330 may reward the selected audience member based on the response evaluation. As described herein, upon evaluating the response and determining that the response is approved, the incentive reward module 1330, to the audience member, depending on the value of the query and / or query response determined by the QR evaluation module 1320 Rewards can be rewarded. In step 1518, the translation data storage update module 1326 may update the translation or translation based on the response evaluation. As described herein, upon evaluating the response and determining that the response is approved, the translation data store update module 1326 updates the word or phrase of the query and the translation or translation corresponding to the submitted query response. You can.

16 is a block diagram illustrating an exemplary data flow 1600 for user feedback according to various embodiments. As shown, the data flow 1600 includes a chat client system 1400, a translation failure management module ( 1314), a UF query generation module 1316, a UF query crowd selection module 1318, a response evaluation module 1324, a translation data storage update module 1326, and an incentive reward module 1330. The data flow 1600 includes an undefined word / phrase data storage unit 1602, a recorded response data storage unit 1604, a feedback spectator data storage unit 1606, a response approval data storage unit 1608, The chat data storage unit 1610 and the dictionary data storage unit 1612 are related.

The undefined word / phrase data storage unit 1602 may include words or phrases related to translation failures and user feedback is being sought. The undefined word / phrase data storage 1602 indicates a sample sentence (eg, word or phrase context) in which the word or phrase is used, along with the word or phrase, and how important the word or phrase is (eg, word importance). Reliability measures, source language for words or phrases, target language for user feedback found. In some embodiments, the word importance of a word or phrase in the undefined word / phrase data storage 1602 is when all words are equal initially, but the words or phrases conflict, have problems and / or cannot be translated Can increase.

The recorded response data storage unit 1604 may include a user feedback response received from an audience member and recorded for a word or phrase included in the undefined word / phrase data storage unit 1602. In some embodiments, the user feedback response includes a response received for a query generated to obtain user feedback regarding a word or phrase. The undefined word / phrase data storage unit 1602 records, along with the recorded user feedback response, an identifier for an audience member submitting the user feedback response, a timestamp when the user feedback response is received and / or recorded It includes an indication of whether the user feedback response is approved, a timestamp when the recorded user feedback response is approved, and the like.

The feedback spectator data storage unit 1606 may include a set of identifiers for spectator members selected to participate in user feedback for words or phrases included in the undefined word / phrase data storage unit 1602. The feedback spectator data storage unit 1606 b, along with each identifier for the spectator member, scores a confidence level reflecting the consistency, competence, and / or confidence of the spectator member providing user feedback. It may include.

The response approval data storage unit 1608 may include each unique user feedback response received in association with a word or phrase included in the undefined word / phrase data storage unit 1602. Response acknowledgment data storage 1608, along with each unique user feedback response, indicates whether the unique user feedback response is an approved response (eg, an accurate response), rejected response (eg, incorrect response), pending review It can include either a medium response (eg, a response that needs to be reviewed), or a response with some other status. In some embodiments, the response approval data storage 1608 may be employed to determine when a user feedback response received from an audience member and recorded in the recorded response data storage 1604 is approved.

According to some embodiments, the translation failure management module 1314 reviews chat logs that may be provided by the chat data storage 1610, and one or more words or phrases associated with actual or potential translation failures. It can be configured to identify. In various embodiments, translation failure management module 1314 may include standard dictionaries (eg, Oxford dictionaries) and / or words that the embodiments described herein can parse, recognize, and / or process. Or it may be configured to exclude such words or phrases, as defined in the dictionary data storage 1612, which may include a dictionary of phrases (eg, chat term words or phrases). Words or phrases identified by the translation failure management module 1314 may be added to the undefined word / phrase data storage 1602, whereby these added words or phrases receive user feedback from selected audience members. Can be selected.

The translation failure management module 1314 may be configured to select one or more words or phrases from the undefined word / phrase data storage unit 1602 for user feedback. In some embodiments, the translation failure management module 1314 sets a word or phrase designated as having the highest importance in the undefined word / phrase data storage unit 1602 (eg, undefined word / phrase data storage unit) (Selected from the top 10 important words or phrases in 1602). In some embodiments, the translation failure management module 1314 may select two or more words or phrases so that the audience member is provided with a set of two or more user feedbacks selectable for response (eg For example, allowing audience members to select these user feedbacks they feel is the most trusted response). The selection process by the translation failure management module 1314 from the undefined word / phrase data storage unit 1602 includes the importance of the word or phrase, the word or phrase in the undefined word / phrase data storage unit 1602, The preference of the selected audience member by the UF query audience selection module 1318, whether the audience member selected by the UF query audience selection module 1318 has already responded to the word or phrase to be selected (eg, a word to be selected or Based on checking the recorded response data storage 1604 for a verse) and the like.

The UF query spectator selection module 1318 may be configured to select one or more spectator members from the feedback spectator data storage 1606 where user feedback can be found. As described herein, user feedback may be searched for the word or phrase selected by the translation failure management module 1314, possibly from the chat data storage 1610. The choice of spectator members from the feedback spectator data storage 1606 may depend on the level of competence associated with the spectator members.

The UF query generation module 1316 generates one or more queries for words or phrases selected by the translation failure management module 1314 from the undefined word / phrase data storage unit 1602 for user feedback. Can be configured. As shown, the translation failure management module 1314 may provide the selected word or phrase in which one or more queries will be generated, to the UF query generation module 1316. As described herein, the UF query generation module 1316 is a word selected for user feedback by the preference and translation failure management module 1314 of the spectator members selected, for example, by the UF query audience selection module 1318. Or, when creating a query that includes phrases, there are many different factors to consider. As a result, the UF query generation module 1316 may generate one or more queries generated by the UF query generation module 1316 which may generate different queries for each word selected and provided by the translation failure management module 1314. , Can be provided to the chat client system 1400.

As a result, one or more queries generated by the UF query generation module 1316 may be provided to the chat client system 1400, and the chat client system 1400 in turn queries the provided queries to the chat client system 1400. Provides for user's choice in. According to an embodiment, the UF query generation module 1316 may provide the generated query to the chat client system, or alternatively, other components may be responsible for providing the generated query to the chat client system. . Once the generated query is provided for selection in the chat client system 1400, a user of the chat client system 1400 can choose to respond to one or more of the provided queries, and the chat client system 1400 These responses provided by) may be added (eg, recorded) to the recorded response data storage 1604.

When a response is added to the recorded response data storage 1604, some embodiments check whether the added response can be evaluated by the response evaluation module 1324. As described herein, the response evaluation module 1324 can evaluate the response by checking the response, and can arrange the status of the response.

As shown in FIG. 16, the response evaluation module 1324 includes a response check module 1614, which may be configured to perform a status check of the user feedback response during evaluation of the user feedback response. The response check module 1614 retrieves a user feedback response from the recorded response data storage unit 1604, and checks the status of the unique response corresponding to the retrieved user feedback in the response approval data storage unit 1608, thereby chatting client The status of the user feedback response from system 1400 may be checked. While doing so, the response check module 1614 can determine whether a particular user feedback response is approved or rejected. The approval status of the retrieved user feedback response in the recorded response data storage 1604 may be updated according to the latest status check performed by the response check module 1614. When the response check module 1614 determines whether the retrieved user feedback response has been approved, the approval status of the retrieved user feedback response in the recorded response data storage 1604 reflects the approval and the approval It can be updated to include a timestamp for when the status was updated. Eventually, for the retrieved user feedback response, the approval reflected in the recorded response data storage 1604 concludes that the incentive reward module 1330 compensates the incentive to the audience member who submitted the approved user feedback response. Can be.

If the status of the retrieved translation response is still pending review, the response check module 1614 may check again the status of the retrieved user feedback response later (eg, according to a predetermined schedule). If the status of the retrieved translation response is rejected, the approval status of the retrieved user feedback response in the recorded response data storage 1604 can be updated to reflect the rejection.

If a unique response corresponding to the retrieved user feedback response is not found, the retrieved user feedback response is sent to the response approval data storage unit 1608, to the user of the chat client system 1400, for example. Can be added as a unique response to a word or phrase. Additionally, if the retrieved user feedback response is added to the response acknowledgment data storage 1608 as a unique response, the unique response may have an initial status pending approval, in which the status of the native response is manually or automatically deployed. (Eg, via evaluation response module 1324).

As shown in FIG. 16, the response evaluation module 1324 also includes a response verifier 1616, which can be configured to place the status of the unique user response into approval, rejection, or pending approval. As described herein, a unique user response submitted in connection with a query generated to obtain user feedback is either approved, or generated, that the unique user feedback response is valid for the word or phrase associated with the generated query. It can be considered pending approval until it is rejected as invalid for the word or phrase associated with the query.

In some embodiments, response evaluation module 1324 may be configured to determine when a particular word or phrase no longer requires additional user feedback. The response evaluation module 1324 is based on example factors such as how many unique user feedback responses are approved for a specific word or phrase, and the number of unique and approved user feedback responses equals or exceeds a certain threshold. You can make a decision. When it is determined that a specific word or phrase requires additional user feedback for a longer time, the response evaluation module 1324 may be configured to remove the specific word or phrase from the undefined word / phrase data storage 1602, This removes the word or phrase from future selection for user feedback (eg, by translation failure management module 1314).

As described herein, the incentive reward module 1330 may be a word or phrase in which a user feedback response provided by an audience member (eg, by an audience member via the chat client system 1400), was provided with a user feedback response. If approved as a valid response to, you can reward the spectator members with incentives. The incentive reward module 1330 may include, in the recorded response data storage unit 1604, recently approved (eg, the approval status has been recently updated to reflect the approval) and / or the incentive reward module 1330. The recorded response data storage 1604 may identify one or more user feedback responses that have been approved since the last attempt to identify one or more user feedback responses with an approved status. The incentive reward module 1330 may determine a time at which the specific user feedback response was last approved based on the approval timestamp included for the user feedback response in the recorded response data storage 1604. In some embodiments, if the incentive is compensated for the translated feedback response in the recorded response data storage 1604, the translated feedback response may be removed from the recorded response data storage 1604. Alternatively, if the incentive is compensated for the translated feedback response in the recorded response data storage 1604, the translated feedback response may be, for example, when the incentive was compensated, the amount of the incentive compensated, the compensated incentive It can be updated in the recorded response data storage 1604 to indicate the type of, when the audience member was notified of the reward, and / or how the audience member was notified of the incentive reward, and the like.

17 shows example screenshots for receiving user feedback for words in accordance with various embodiments. In particular, FIG. 17 is a screenshot 1702, 1704 and 1706 showing an example of a GUI that can be provided to an audience member (eg, via the chat client system 1400) to facilitate the user feedback process. Screenshot 1702 shows an example of a banner 1708 encouraging one or more audience members to participate in user feedback on a word or phrase associated with a translation failure. Audience members can direct user feedback by selecting banner 1708, which can lead to the start of a user feedback session and / or lead a spectator member to a list of available user feedback opportunities from which the spectator member can choose to participate. You can choose to participate. As described herein, user feedback opportunities can allow audience members to provide definitions for words or phrases associated with actual or potential translation failures. According to some embodiments, the spectator member may select one of the available user feedback opportunities associated with the word or phrase, and when prompted to do so, provide a definition of the associated word or phrase.

Screenshot 1704 shows a list 1710 of available user feedback opportunities for various words (eg, “Skrilla”, “Booty”, “Cray”, “Hecka” and “Freshness”). Screenshot 1706 provides an example of a query 1712 presented to an audience member to get user feedback on the word “Skrilla”. As shown, query 1712 provides an example context 1714 in which the word “Skrilla” is used, and also provides a field 1716 configured to receive a pre-form response to query 1712. . When the spectator member selects user feedback for the word “Skrilla” from the list 1710 of the screenshot 1704, the spectator member is directed to the screenshot 1706.

18 shows example screenshots for skipping user feedback according to various embodiments. In particular, FIG. 18 is a screenshot 1802, 1804 showing an example of a GUI that may be presented to an audience member (eg, via the chat client system 1400) to enable a user feedback process. Screenshot 1802 shows a list 1806 of user feedback opportunities that are selectable by an audience member. As shown, list 1806 provides spectator members an option to skip one or more of the user feedback opportunities listed.

Screenshot 1804 shows an example of a query 1808 presented to an audience member to get user feedback for various words. As shown, query 1808 provides an audience member with the option to skip the process of providing a response to query 1808. Certain embodiments enable avoiding inaccurate and / or fictional responses to various user feedback queries by providing an audience member with the option to skip specific user feedback opportunities and / or various user feedback queries.

19 shows example screenshots for receiving user feedback for a phrase in accordance with various embodiments. In particular, FIG. 19 is a screenshot 1902, 1904 showing an example of a GUI that can be presented to an audience member (eg, via the chat client system 1400) to facilitate the user feedback process. Screenshot 1902 shows a list of user feedback opportunities 1906 selectable by an audience member. As shown, list 1906 represents available user feedback opportunities for various words and phrases (eg, “Skrilla” and “Pardon my french”).

Screenshot 1904 shows an example of a query 1908 presented to an audience member to get user feedback for the phrase “Pardon my french”. As shown, query 1908 provides an example context 1910 in which the phrase “Pardon my french” is used, and also provides a field 1912 configured to receive a pre-form response to query 1910 do. When user feedback for the phrase “Pardon my french” from the list 1906 of the screenshot 1902 is selected by the audience member, the audience member can be directed to the screenshot 1904.

20 shows an example screenshot for receiving user feedback through a list of select-form responses according to various embodiments. In particular, FIG. 20 is a screenshot (2002, 2004, 2006) showing an example of a GUI that can be presented to an audience member (eg, via the chat client system 1400) to facilitate the user feedback process. Screenshot 2002 shows an example of a banner 2008 that requests one or more spectator members to participate in user feedback for a word or phrase associated with a translation failure. Spectator members can choose to participate in user feedback by selecting banner 2008, which banner 2008 can guide to the start of a user feedback session and / or the audience member is available to choose to participate A list of user feedback opportunities can direct the audience member.

The screenshot 2004 provides an example of a query 2010 presented to an audience member to get user feedback on the word “Skrilla”. The query 2010 includes a select-form response 2012 listing possible responses that can be selected by an audience member. Screenshot 2006 presents rewarded incentives (eg, 5 gold coins) by notifications (2014 and 2016) when the correct response “money” is selected for the word “Skrilla”.

21 shows example screenshots for generating a list of selections in accordance with various embodiments. In FIG. 21, screenshot 2102 provides an example of a query 2106 presented to an audience member to obtain user feedback for the word “Skrilla”. As shown, query 2106 provides an exemplary context 2108 in which the word “Skrilla” is used, and also provides a field 2110 configured to receive a pre-form response to query 2106.

According to some embodiments, a select-form response method used to obtain user feedback for a particular word or phrase includes a list of predefined responses selected from pre-form responses collected for that particular word or phrase. Can be. Accordingly, various spectator members provide a pre-form response to the word “Skrilla” through the field 2110 (eg, “a lot of momey”, “Cash”, “Money” and “Really Rich”), The collected responses can be useful for generating a list of select-form responses 2112, as shown in screenshot 2104.

22 illustrates exemplary incentive notifications in accordance with various embodiments. In FIG. 22, the screenshot 2200 informs the spectator member of the approval of the spectator member's response to the word “Skrilla”, “momey”, and rewards the incentive (eg, XXXX Gold) rewarded for the approved response. Here is an example of a notification to an audience member. The screenshot 2202 shows an example of a notification informing the audience member of the rejection of “money”, which is the response of the audience member to the word “Skrilla”. Screenshot 2204 shows an example of a push notification informing an audience member of the approval of an audience member's response.

23 shows a screenshot showing an example of when translation fails between chat client systems, according to various embodiments. In FIG. 23, screenshot 2300 represents an exemplary interface of the first chat client system, and screenshot 2302 represents an exemplary interface of the second chat client system. The double arrow 2304 indicates chat communication between the first and second chat client systems. As shown, when the chat user “Aramis” enters chat communication in English into the interface of the first chat client system, the entered chat communication is translated into French and the interface of the second chat client system of the chat user “tapir” Is displayed on. Similarly, when the chat user “tapir” enters chat communication in French into the interface of the second chat client system, the entered chat communication is translated into English and displayed on the interface of the first chat client system of the chat user “Aramis”. .

As shown in FIG. 23, the chat communication 2306 (ie, “Tru dat bro?”) Input to the interface of the first chat client system by the chat user “Aramis” is the second of the chat user “tapir”. Translation fails when sent to the interface of the chat client system. The chat communication 2308 presented to the chat user “tapir” (ie “Tru dat bro?”) Presents the original chat communication entered by the chat user “Aramis” and the chat communication to the chat user “tapir” is the chat user By indicating that it is the original chat message entered by "Aramis", it reflects the translation failure.

The translation failure described in FIG. 23 may benefit from user feedback in accordance with some embodiments. According to some embodiments, the translation failure described in FIG. 23 may be identified by the translation failure management module 1314, one or more from the original chat communication 2306 (ie, “Tru dat bro?”) The above words may be added to the undefined word / phrase data storage unit 1602 for future selection for user feedback from participating audience members. For example, each word “Tru”, “dat”, and “bro” may be added to the undefined word / phrase data storage unit 1602 for future user feedback on each.

24 and 25 illustrate screenshots illustrating an exemplary list of words or phrases available for user feedback, according to various embodiments. In FIG. 24, screenshot 2400 presents a list 2402 of user feedback opportunities available to the audience member, for words and phrases (including “Tru”) selectable by the audience member. In FIG. 25, screenshot 2500 presents another list 2502 of user feedback opportunities, available for audience member selection, for verses selectable by the audience member. 24 and 25, screenshots 2400 and 2500 can be part of an in-game chat system that allows game players to provide user feedback for a particular word or phrase, and as soon as the user feedback is approved, In-game credits (eg in-game gold) can be given.

26 is a screen shot illustrating an example of defining a word according to various embodiments. In FIG. 26, screenshot 2600 includes an example context 2604 in which the word “Tru” is used, and also provides a field 2606 configured to receive a pre-form response to query 2602. The query 2602 is presented.

27 shows a screenshot illustrating an exemplary list of select-form responses in accordance with various embodiments. In FIG. 27, screenshot 2700 presents a list 2702 of responses the audience member can select to define the word “nemor”.

28 shows a screenshot illustrating an exemplary list of statuses for responses submitted in accordance with various embodiments. As shown in FIG. 28, the list 2802 of the submitted response status is pending status for the first response 2804, and the approved status for the second and third responses 2806 and 2808 ( approved) states. In some embodiments, the list 2802 can provide additional information about the status of the response, including, for example, why the particular response was approved, or why it was pending pending review.

29 shows screenshots illustrating exemplary incentive notifications in accordance with various embodiments. In particular, FIG. 29 provides a screenshot 2900 presenting an exemplary notification 2902 to a member who correctly defined the phrase “U still thr” as “You still there?” During the user feedback process according to one embodiment. . The notification, as an incentive for the user feedback provided, informs him that he will be rewarded with 10 valuable or useful gold in an in-game currency. As shown, the notification also provides a summary of user feedback (ie, the word or phrase being a question, and the user feedback response provided).

In certain embodiments, a learning system for data selection is provided in which feedback from the user is automated by a machine learning system that checks and balances for player consistency. The system adds parallel sentences received from the player to a parallel corpora that can sometimes be used to retrain a statistical machine translation (SMT) system.

The chat conversion system may be a system that converts chat terms into plain speak, or may include the system. For example, the chat conversion may convert “U r da king” (chat term message) to “You are the king” (obvious term message). In certain embodiments, “plain speak” refers to an ordinary language spoken and / or written by an ordinary person outside the electronic chat environment in which the chat term is dominant. Clear terms tend to be more grammatical than chat terms.

The learning system may also use or include a language translation system that translates one language into another. For example, a language translation can translate “How are you doing kind sir” (English message) to “¿Como te va amable senor” (Spanish message).

In some embodiments, “parallel corpus” can be understood to mean two texts arranged in parallel in each language such that n lines of one text correspond to n lines of the second translated text. Parallel corpus can also be referred to in this context as “training corpora”.

In various embodiments, “machine learning” will be understood as being referred to as supervised learning, semi-supervised learning or autonomous learning, which can learn from patterns of input data and develop mechanisms for discovery, transformation or behavior prediction. Can be.

In general, the construction of a chat conversion system and a language translation system requires an appropriate amount of syntactic laws or a large amount of parallel corpus for statistical learning. The systems and methods described herein generally use a reliable statistical chat transformation and language translation system using parallel corpus. However, in certain situations, such initial training data may be limited in content and scope. For example, new chat words can be created and added to the chat room every day. In order to maintain an accurate and reliable conversion and translation system, these new chat words must be increased in the chat conversion training corpus.

In various embodiments, systems and methods are provided for identifying words that are out of vocabulary (OOV) (eg, words that are not presented in a particular lexicon). Referring to FIG. 30, in some embodiments, a method 3000 for detecting and processing OOV words is provided. In step 3002, OOV words are initially detected by sending words through a translator system such as CTT system 114 or CCT system 1300 and / or one or more modules thereof. When the output from the translator system is the same as the input for a particular word, the translator system indicates a lack of translatability and suggests that the word can be OOV. To further evaluate that a word is a potential OOV word, the system and method can determine if the word is a new word rather than a misspelled word (step 3004), and in both cases may appear as an OOV. Accordingly, a word that can be corrected with a spell checker may be considered as a misspelled word rather than an OOV word.

Additionally, OOV words that appear frequently in chat tend to be more OOV words (eg, new chat term words). For example, when a word was used by a user in a previous text message, such previous use suggests that the word may be an OOV word. In some embodiments, a total of machine learning and language processing methods are used in parallel to detect whether a word is an OOV word (step 3006).

Additionally or alternatively, the Bayesian probability can be calculated to provide a statistical probability when the OOV is a new word rather than a misspelled word (step 3008). True chat terminology tends to be used before and after certain commonly used words. In comparison, spelling errors will have less consistent distribution with adjacent words. Calculating Bayesian probabilities before and after can help distinguish between useful OOV words that can be added to a dictionary and spelling errors that should not be added to a lexicon. For example, when considering the phrase “Wassup, how ’s it going.”, “Wassup” is considered an OOV word when it is not in the standard lexicon. However, “Wassup” is almost always followed by the word “How ’s it going”, or is usually used at the beginning of a sentence. This pattern or consistent behavior is captured by Bayesian probability. The system is trained on text that contains misspelled words but not chat words.

Alternatively, or additionally, machine learning methods such as k-means clustering can be used to distinguish new words, misspelled words, or different kinds of OOV words such as junk. k-means clustering tends to elicit potential similarities between classes of words. Words belonging to similar subjects tend to be grouped together, indicating a potential synonymous relationship between them. Consider again the example of “Wassup, how ’s it going”. When a group of sentences is clustered using the k-means algorithm, a cluster of greeting words such as “Hi,” “What's up,” “Hello,” “Hi !,” etc., within the cluster, between them Reveals "Wassup" included in Conversely, spelling errors are placed at the periphery of the cluster, or not at all in any defined cluster. This potential relationship helps to distinguish useful OOV words from errors. The syntax and semantics of a sentence can be analyzed to determine what kind of OOV word (eg, verb, noun, or adjective) the sentence contains.

When the system and method detect a new chat word or other OOV word, the new chat word can be presented to a human translator to define a chat translated version or a language translated version of the new chat word. Translated or translated versions of new chat words can be added to the translation dictionary and used by the systems and methods described herein.

As described herein, when incentives are provided for manual translation of chats between languages, there is a possibility that users will manipulate the system maliciously to benefit from incentives (eg, in-game currency). The systems and methods described herein are generally tolerant of human-translator's ability to detect malicious submissions.

When a user of the system acts as a translator, the user translates one or more words or sentences into a specified target language. However, fraud can be fraudulent when a user provides a false, incomplete, or immoral translation for the purpose of betting on the system or for obtaining incentives without fulfilling the purpose of the system.

Referring to Figure 31A, in certain embodiments, the systems and methods described herein utilize fraud detection module 3100. The fraud detection module 3100 can detect fraud in incentivized translations by providing users with new and old training data (eg, parallel corpus). Old training data corresponds to a translation for which the correct answer is known. On the other hand, the new training data corresponds to a translation for which an exact correct answer is unknown. The percentage (%) of new data versus old data can change over time. For example, older data may be provided more initially and then gradually reduced in percentage.

In some embodiments, fraud detection may be performed by checking the accuracy of old data translations received from the user. Reliability scores are assigned to each user based on this accuracy. A large or sudden change in translation accuracy or constant low accuracy indicates a user's fraud or low translation ability. Even after establishing confidence in the translator's ability, it is desirable that old data is randomly introduced at least 10-20% of the time for periodic fraud checks.

Using this basic structure, the fraud detection module 3100 may include and / or use a supervised fraud detection module 3102 and / or an unsupervised fraud detection module 3104. Through the map fraud detection module 3102, the reporting tool may, for example, present output from each user to the user interface. Then, the user interface has the following fields: a graph showing the proposed input sentence, the obtained translation, the currently known true translation, the user's current confidence score, and the change in the translator's confidence score over time. When the translation is reviewed, the human supervisor can accept or reject the translation and adjust the user's confidence score accordingly. The supervisor can remove the user if the accumulated report shows fraudulent behavior (ie, cancel the user's translation privilege). The removal of the user or cancellation of the translation privilege of the user may be performed using the translation privilege module 3106.

Alternatively, or additionally, the fraud detection module 3100 can use the autonomous fraud detection module 3104. Through the autonomous fraud detection module 3104, the accuracy of translation can be calculated using various metrics such as word error rate (WEB) and BLEU (machine translation accuracy metric that compares machine translation to a reference reference of good quality). . Reliability in user translation ability can be checked for changes or fluctuations (eg, rising or falling). Similar sentences presented to one translator can be presented to another independent translator using the system. Internet translator reliability can also be calculated. For example, through random sampling, social networking analysis (e.g., verifying that two translators are not socially connected or having no existing relationship), and repetitive between users in chat sessions and / or online games. By detecting interactions, conspiracy among translators can be prevented. Two users who regularly interact together online (eg online games or chat sessions) are more likely to join the competition. In some embodiments, item response theory (eg, the theory used in psycholinguistics and black theory) is used to increase the measure of translator confidence using translator capabilities. Fraud detection can be performed using item response theory to autonomous fraud detection in a translation enhancement system with an incentive mechanism. The item response theory dictates how translator accuracy can be measured over time with respect to peers and with respect to themselves to measure consistency. Deviations from the norm can be identified in this way. Intra-translator capability can also be calculated by presenting the same sentence back to the translator after a set period of time. Various thresholds in reliability and translator confidence can be set, and if the reliability of the translator falls below that threshold, the translator is removed using the translation privilege module 3106 and blocked from the system. (Eg, the user's translation privileges may be revoked). In some embodiments, translations from high reliability systems are added to a translation pair dictionary (lexicon).

31B is a flow diagram of a method 3110 for detecting fraud in an incentivized translation, in accordance with certain embodiments of the present invention. The method includes selecting a mixture of old training data and new training data (step 3112). The old training data includes one or more old text messages whose exact translation to another language is known. The new training data includes one or more new text messages whose exact translation to another language is not known. Each of the plurality of requests is sent to the user's client device at different times (step 3114). The request includes (i) requesting the user to translate old training data and / or new training data and (ii) incentives for translation. After sending the specific request, a translation for the old training data of the specific request is received from the user device (step 3116). The received translation is compared to the correct translation for old training data (step 3118). The accuracy of the received translation is determined based on the comparison (step 3120). Next, the confidence score is updated for the user based on the translation (step 3122). The reliability score indicates the likelihood that the user will provide an accurate translation of the text message in another language at a later time.

In various embodiments, the systems and methods described herein utilize speech translation or speech recognition to translate audible speech in one language to another language for users of a group speech chat system. The system and method can be implemented for chat terms. The speech-to-text change system converts user chat terms into text, and then the text is translated into a clear language (ie, non-chat term) and implemented for chat terms that are translated into foreign languages. Can be. The final conversion is then performed to produce a foreign chat voice that is output to the end user through a foreign language text-to-speech system. The system and method preferably utilize speech recognition technology and statistical machine translation technology as very fast decoders.

FIG. 32 is a block diagram of a group chat system 3200 that allows groups 3202 of people speaking different languages to verbally interact using chat terms. As described herein, system 3200 may recognize a language spoken by a person participating in group chat system 3200. When the first user 3204 wishes to send an audible chatspeak message to the second user 3206, the first user 3204 may be sent to a user input device (eg, a microphone in a chat client system). Enter audible chat term message 3208 in one language. The speech recognition module 3210 changes the audible chat term message to the first term chat term text message 3212. A conversion module 3214 is used to convert the chatspeak message 3212 to a clear text (eg, non-chatspeak) text message 3216 in the first language. Next, a translation module 3218 is used to translate the clear term text message 3216 into the corresponding clear term text message 3220 in a second language (eg, French). The conversion module 3222 is then used to convert the corresponding clear term text message 3220 to the corresponding chatspeak text message 3224 in a second language. Those skilled in the art will recognize that the conversion module 3222 may be the same as or form part of the conversion module 3214. The text-to-speech module 3226 is then used to change the corresponding chatspeak text message 3224 to the corresponding chatspeak audible message 3228 in the second language. Finally, the corresponding chat term audible message 3228 is delivered to the second user 3206 using an output device (eg, the speaker of the second chat client system).

In various embodiments, the speech recognition module 3210 changes the audible chat term text message 3208 into a chat term text message 3212 to use a hidden Markov model, DTW (Dynamic Time Warping) based speech recognition. , And / or neural networks. Similarly, the text-to-speech module 3226 can use speech synthesis to change the corresponding chatspeak message to the corresponding chatspeak audible message. Speech synthesis is concatenative synthesis (e.g., unit selection synthesis, diphone synthesis, and / or domain-specific synthesis) that can be understood by those skilled in the art. ), Formant synthesis, articulatory synthesis, HMM-based synthesis, and / or sinewave synthesis.

An important aspect of creating something like a speech processing system involves collecting speech samples from multiple accents and dialects for the language being processed. The essence of voice data includes chat terms and clear terminology formats for each language, in order to maintain the relevance of the system and the domain it addresses. The incentivized feedback mechanism described herein can be used to alter this speech sample, which in turn can be used to train the speech recognition module 3210 and / or text-speech module 3226. Domain adaptation techniques may be necessary in cases of chatspeak speech samples where data tends to be sparse. For example, voice data collected in a game domain (eg, online game) can be replaced with clear voice term data that is abundantly available. Domain adaptation preferably governs rules that govern minor voice fluctuations from chat-obvious terms in a particular language (e.g., in certain languages, changing chat terms to clear terms or clear terms to chat terms). Rules). Clear term sentences that do not have speech samples in equivalent chat terms can be changed to chat terms using the domain level rules. The user feedback loop can be used to tune the acoustic model parameters (eg, speech recognition module 3210 and / or text-to-speech module 3226) to a level that makes the acoustic model domain special and more accurate. For example, if the speech recognition module 3210 continues to have difficulty with a particular accent, audible samples of additional various words may be provided to the system with the accent (eg, by users). This will help the speech recognition module 3210 learn how to better recognize words pronounced with the accent.

As mentioned, embodiments of the systems and methods described herein are used to translate text or chat messages from a group chat environment into other languages. Archiving these translated chats allows a large number of texts in different languages to be kept in the repository.

Referring to FIG. 33A, in certain embodiments, to facilitate review of the chat history by the user and to reduce storage requirements, the chat history module 3300 chats in real time when the chat history is browsed by the user. Used to translate histories. The chat history module 3300 includes a chat storage module 3302 (eg, a register or other storage device) for storing chat history from various users. The chat history module 3300 also includes a chat history conversion module 3304 that converts text messages before and / or after the text messages are translated into other languages. For example, the chat history conversion module 3304 performs real-time conversion of chat history text messages from chat terms to formal or explicit terms. In some embodiments, chat history transformation module 3304 is the same or similar to transformation module 208. Chat history module 3300 can also be used to perform real-time translation of chat history text messages (e.g., in official or clear terms) to other languages (e.g., French to English), chat history translation module 3306 ). Chat history translation module 3306 can be and include other modules or components described herein, such as language module 206 and / or translation data storage 210.

When the user completes reviewing the chat history, the converted and / or translated text generated by the chat history module 3300 may be deleted or removed from the memory. This reduces storage requirements for systems and methods. If the user wants to review the chat history later, the chat history module 3300 can be used again to convert and translate text in the chat history, if required.

In certain embodiments, the chat history module 3300 translates the chat history for the user in real time. The chat history module 3300 receives a request from a user to review the history of a text message in a chat session. The chat history module 3300 receives a history of text messages including text messages in a plurality of languages from the chat history storage module 3302. The chat history conversion module 3304 and chat history translation module 3306 are then used to convert and / or translate the initial portion of the chat history to the language used by the user when required. After viewing the translated first portion of the chat history, the user may want to view another portion of the chat history. Then, the chat history module 3300 may receive a request from the user to view another part of the history of the text message. The chat history conversion module 3304 and chat history translation module 3306 are then used to convert and / or translate other parts of the chat history into the language used by the user when required. The chat history module 3300 preferably converts and / or translates in real time when the user scrolls the chat history.

In a particular example, the scrolling of the chat history represents a problem of scale and data storage. The provision of infinite strokes in the chat history represents the problem of fast real-time access to data spanning multiple databases and multiple users. This can happen by spawning multiple processes in parallel that fetch historical messages from all users in the chat room. Translation of these messages and associated chat transformations are performed in parallel when messages are retrieved from the data store. The result output, realized by the end user, is the result output of a seamless transition from one screen of the chat to the next, where data lookup from the database has already been done. This can start for an infinite number of screens when the systems and methods described herein do not have any limitations on data storage, and parallel computation is recycled between derived processes earlier.

33B is a flow diagram of a method 3310 for translating chat history in real time, in accordance with certain embodiments of the present invention. The method 3310 includes receiving a request from one person to review the history of the text message from the chat session (step 3312). The history preferably includes text messages in multiple languages and from multiple users. At least two parallel processes are performed (step 3314). Each parallel process includes: (i) receiving or selecting a text message generated by each user of a chat session (eg, the text message forms at least a portion of the history of text messages), and (ii) text. And translating the message into the target language. Text messages translated from multiple parallel processes are provided to the person's client device (step 3316). A request is received from the person to review other parts of the history of the text message (step 3318). Steps 3314 and 3316 are repeated for other parts of the history of the text message.

In some examples, users of the systems and methods described herein may want to avoid interaction with certain other users in a group chat or gaming environment. In previous chat systems, the ban and silence of chat users is typically handled by the chat server's administrator or coordinator. However, embodiments of the systems and methods described herein allow the user to directly control who can send user chat messages and / or chat contact invitations. For example, user A may be allowed to block user B, such that user A may no longer see user B's communication in any chat room, and / or user A no longer has a personal chat contact from user B (I.e. exclusive chat) You may not receive an invitation.

In various embodiments, an alliance is a group of players that can be grouped together as a unit to enhance gameplay in the game. Each alliance preferably has a single chat room where members of the alliance can communicate with each other and send text messages. This sometimes indicates the need to block certain users from allied chat rooms.

34A includes a screenshot of a user interface 3400 that allows a first user of a gaming system to block communication from a second user of the gaming system, according to certain embodiments. As shown, the first user selects the management settings icon 3402 that opens the settings window 3404 (eg, by tapping the touch screen). Then, the first user selects the "block from alliance" button 3406 in the setting window 3404. A message window 3408 appears to inform the first user that future communication from the second user will be blocked. The second user can be added to the list of other users blocked by the first user. The first user can have the option of editing this list to add or remove users from the list. For example, referring to FIG. 34B, the next time the user selects the management settings icon 3402, the settings window 3404 includes an unblock from alliance button 3410. When the first user selects the unblock button 3410 from the federation, future communication from the second user may not be blocked, and a message window 3412 may inform the first user that such communication is not blocked appear.

In some embodiments, the complexity of the system is increased or decreased by the scale at which blocking and unblocking is performed. Parallel computation can provide the flexibility to perform blocking and unblocking in real time without the penalty of time delays seen in conventional systems. For example, parallel processing can be used to translate and / or translate text messages in a text message chat system. Separate parallel processing can be deployed to each user of the chat session and / or to each language used in the chat session. Such parallel processing can simplify the task of blocking and unblocking the user. For example, separate parallel processing can be added or removed from the chat system, respectively, when users are blocked or unblocked.

35 is a flow chart of a method 3500 for blocking a user in a chat session, and the method 3500 includes providing a text message chat system to a plurality of users of an online game (step 3502). A request for blocking the second user of the text message chat system is received from the first user of the text message chat system (step 3504). Following receipt of the request, a text message from the second user is blocked from being displayed to the first user (step 3506). In some examples, text messages in a chat session are translated and / or translated using the systems and methods described herein. Parallel processing can be used to perform translation and / or conversion of text messages. For example, separate parallel processing may be arranged to handle translation and / or translation of text messages of each particular user in the chat session, and / or may be placed in each language involved in the chat session.

Automated translation services are not always accurate, and sometimes human intervention can benefit from correcting certain errors. In some implementations, the translation system and method described herein can allow a user to identify translation errors and suggest corrections to correct those errors. For example, bilingual users or foreign language users (eg, French players in online games) can view the chat window and see incorrect translations (eg, in French or from French). The user can submit a correction proposal for the erroneous translation, and the user can receive a reward for the correction submission (eg, in-game currency or virtual item).

In some implementations, the original text message and corresponding translation are displayed on a single screen, which provides an opportunity for experienced users in the language to immediately provide feedback on the translation. For example, the user recognizes a translation error and selects an option to submit a corrected translation. The user can then enter and submit the corrected translation and receive compensation if and when the corrected translation is approved. Upon submitting the corrected translation, the user is blocked from submitting additional corrected translations of the original message. This prevents users from getting multiple rewards from a single erroneous translation.

In some examples, the system and method cannot translate the original message because the original message is not entered correctly by the user. For example, FIG. 36A shows an original Spanish message 3602 quoting “Eres el peor!”. The automated English translation 3604 of this message is shown in Figure 36B and quotes "You are the best!". 36C and 36D, the user can recognize that the original message was not entered in the correct Spanish and was translated incorrectly. To cope with this error, the user selects the "correct translation button" 3606, which causes the user to enter a correction window 3608 to enter a correction to the original message. 36E and 36F, in this case, the user enters “Ustedes son los mejores!” In the correction window 3608 and selects a submit button 3610. A confirmation window 3612 appears informing the user that the submission will be processed. When the confirmation window 3612 is closed, the user returns to the original chat page.

37A and 37B, a user interface 3700 is provided that allows a user to review translation corrections received from other users. Users who review translation corrections can be rewarded for their efforts, and can select a specific type of reward they have received (eg, virtual goods or currency for online gaming). In general, after a translation correction has been submitted by a user, another user can determine whether the correction was better than the original translation and other translation corrections that were submitted by other users. When it is determined that the user's translation is the best translation, the user can receive the reward, and the user's translation can be added to the translation dictionary (translation data storage 210). Users who participate in judging various translations can also receive rewards. However, these rewards can only be given to users who have selected the translation correction best selected by all judges.

In general, by allowing users to submit proposed translation corrections and judge other users' submissions, the system and method utilize user feedback that users are free to provide. The data collected in this process, once approved, can be used to correct the translation cache entry, thereby improving all translation capabilities of the systems and methods described herein. This can ensure that when the original message is resubmitted for translation, the correct translation will be shown in the future.

In a typical implementation, there are two types of users who can submit and / or determine translation corrections: users who speak one language, and users who speak two languages. Users who speak two languages can generally understand the original language sentences and provide more accurate translations in other languages. Conversely, a user who speaks one language is unable to understand the original language verse, but can still review translations (provided in the user's device language) and submit corrections in exchange for rewards. Translations obtained from both types of users tend to differ in content. Bilingual users generally provide more accurate translations. It is preferred that the system and method be able to determine or detect whether the user is a user who speaks a single language or a bilingual user, and the user's feedback can be evaluated accordingly. For example, users can confirm to the system and method whether they are monolingual or bilingual users.

In certain situations, most of the users are monolingual and speak the same language (English). When a large number of people speaking a language are supplied, the number of users submitting a translation offer for that language is generally reduced, and the demand for correction of translations into or from that language is generally reduced. In order to stimulate the supply of translation corrections to other languages, the user can be compensated according to the demand for translation corrections. For example, when multiple users speak a language and there is not a shortage of translation corrections given in that language, those users may receive less compensation (eg, 75% of nominal total) for the submission of translation corrections. have. At the same time, a small number of users speaking in different languages can receive greater rewards (eg, 125% of the nominal total) due to the greater demand for translation corrections in those different languages.

The number of translations that a user can correct over a certain period of time (eg, one day) may or may not be limited. For example, there may be no limit to the number of translation corrections when no compensation is given for submitting corrections. On the other hand, when the user is compensated for such submission, the user may be allowed to submit a limited number of translation corrections during that period. This limitation prevents users who speak bilingual, or those who tend to submit a large number of translation corrections, from being overly rewarded, thereby unfairly benefiting the underlying game (eg multiplayer online games). It can be prevented from getting.

In certain instances, feedback on incorrect translations that are difficult to determine the accuracy of translation submission and difficult to automatically generate rewards can be received from only a small number of users. For example, chats occur in a continuous stream, and many users may focus more on chatting with other users and / or focusing on playing in the underlying game, and less on submitting translation corrections. Users can also choose a chat based on what they see in their window, and not many users can choose the same chat. Accordingly, when many translation corrections are received, the proposed correction can be made to be judged by other users in exchange for compensation, in an effort to obtain a consensus on the correct translation.

Rewards for submitting translation corrections can be given to users according to a raffle system. In such a system, rewards are not given for all submissions, but can be given randomly to users who submit more corrections that are likely to get rewards. This approach reduces the likelihood that certain players will gain an unfair advantage over other users due to their ability and / or desire in the translation of the message, compared to their ability or effort in the underlying game.

In addition to allowing users to correct incorrect translations, users can also submit feedback regarding incorrectly detected languages, unfiltered profanity, and named entity detection. For example, when viewing the original message and the translated message, the user can recognize that the automated translation system has incorrectly detected the original language. The user can then inform the system of this language detection mistake in exchange for a possible reward. Likewise, the user can inform the system about profanity appearing in the message, thereby allowing the system to filter or delete such profanity from future messages. Users can also inform the system about the entity name appearing in the message, such as company, brand, trademark, and so on. This allows the system and method to recognize when the entity name appears in the message and allows those entities to be properly named and / or identified.

In general, translation corrections submitted by the user require careful evaluation to ensure that the user is only compensated for the correct correction. This will improve the overall accuracy of the system and prevent users from cheating by submitting fraudulent corrections. In some implementations, the accuracy of translation correction uses word-based features, language-based features, and other features (eg, word alignment match features) to prevent users from unfairly using the system. It is evaluated automatically. A language model based on a part of speech (POS) can be used to check sentences for grammar correction. Additionally, some users may submit translation corrections that are grammatically correct but independent of the original message. In such a case, the word placement matching analysis feature can be useful and can act as a periodic process to approve and / or reject user submissions. The machine learning approach can be used to authenticate rare user feedback in the translation systems and methods described herein.

Table 2 shows examples of proposed translation corrections submitted by users according to certain embodiments of the present invention. In this example, the original message in the source language is “aaa bbb ccc” and the correct translation in the target language is “xxx yyy zzz”. The column “Shown Translation” contains an example of the initial translation proposed by the automated system described herein.

Examples of user corrections and desirable results The original message Translation shown User correction Explanation state aaa bbb ccc

.;$@#!

.; $ @ #! xxx yyy zzz User corrects incorrect translation Approved xxx yyy xxx yyy zzz Improves the quality of existing machine translations that are partially accurate by the user Approved aaa xxx uuu xxx 'yyy' zzz ' Users do not know the target language, but simply use an online translation service Approved xxx yyy zzz xxx zzz yyy The user unreasonably uses the system for compensation by rearranging the words a little in the translation shown. Rejected xxx yyy zzz xxx yyy zzz Copy and paste the same translation for cheating Rejected xxx yyy zzz

$ @% ^ & / dse reyyfwf If copying is rejected, type a random message / junk for cheating Rejected xxx yyy zzz sss ddd fff The user submits a grammatically correct message in the target language, but the translation does not matter (copy and paste the translation of the previous message) Rejected xxx yyy zzz XXX yyY zzz User attempts to submit a true translation, but his correction is worse than the translation shown (low quality) Rejected

As shown in Table 2, when the user submits an accurate and improved translation, the user submission must be approved, and the user can receive appropriate rewards. However, when a user submits a poor or fraudulent translation (eg, a random message), the user submission should be rejected and no reward should be given to the user. The system and method preferably approve or reject the example as shown in the “Status” column of Table 2.

In certain embodiments, the translation of the original message is classified according to whether the translation is appropriate for the original message. Classification can be handled when features are extracted from translation and original messages in a binomial classification task. The classification technique can be used to ensure that the translation correction submitted by the user is correct. For example, in some examples, the majority-based validation described herein is not suitable because the number of responses that can be received per incorrect translation is small (eg, one to three). Classification techniques can also be used to identify and / or process hash collisions that appear in cached translation data. For example, over 10% of translation entries in a data table can be contaminated due to hash collisions.

Referring to FIG. 38, in various implementations, the accuracy of translation uses a translation accuracy module 3800 that includes a word-based feature module 3802, a language-based feature module 3804, and a word placement module 3806. Is evaluated. The word based feature module 3802 is used to evaluate word based features, such as word count, character count, emoji, numbers, and / or punctuation. For example, when the translation is correct, the number of words in the original message and the number of words in the translation are generally the same. Accordingly, if the number of words in the two messages differs by more than a threshold amount (eg, double), the translation may be considered to be inaccurate or possibly inaccurate. In one example, if the number of words in one of the messages (eg, translation) is 1/2 (or less than 1/2) of the number of words in another message (eg, the original message), then the word-based feature module ( 3802) may conclude that the translation may be inaccurate or inaccurate.

Another word-based feature that can be used to evaluate the accuracy of the translation is the number of characters (eg, letters and numbers) in the original message and translation. In general, when the number of characters in the original message is equal to the number of characters in the translation, the translation can be regarded as correct. The threshold amount can be used to determine when the character counts in the two messages are too different. For example, if the translation has more than 3/2 of the characters in the original message, the word-based feature module 3802 can conclude that the translation is likely to be incorrect or inaccurate.

Another word-based feature that can be used to evaluate the accuracy of translations is the count and order of emoji (e.g., ideograms and smileys used in Japanese electronic messages), which are usually translated from the original message. It remains unchanged between. Emoji tends to be located within a specific range of Unicode text that can be used to detect emoji in a particular sentence. Regular expressions can be used to identify or extract emoji from two messages using this Unicode range. For example, if the input contains three emoji in succession and the output contains only one emoji, this represents a difference between the input and output. If the count and / or order of the emoji is different between the two messages, the word-based feature module 3802 can conclude that the translation is likely to be incorrect or inaccurate.

An additional word-based feature that can be used to evaluate translation accuracy is the count of numbers and punctuation marks in both messages. For example, numbers and punctuation marks can be identified and extracted from the original message and translation, and the length of the Longest Common Susequence (LCS) can be found between them after sorting. This length is divided by the maximum value of the lengths of the two messages, providing the actual numeric value for this word-based feature. In general, the actual numeric value provides an indication of the percentage of the number and punctuation in the two overlapping messages. Experimental results show that for this particular feature, better results are obtained using actual numeric values than binary values. For example, an input sentence of “I am going to meet you at 4:30 Cya !!” in English may have an equivalent output of “Je vais vous recontrer a 4:30 Au revoir !!”. By extracting punctuation and numbers, you can get “4:30 !!” for both English and French versions. In this case, the LCS is 6 (by letter), and the maximum length of the English and French versions is 36 (by letter). The actual numeric value resulting from this word-based feature is 6/36 = 0.167.

Reliance on word-based features alone may be insufficient to assess translation accuracy. For example, a user may submit at least a translation correction by replacing each word in the original message with a dummy word (eg, “xxx”) to produce a fraudulent correction with the same number of words and characters appearing in the original message. Some word-based features can be deceived.

To avoid this problem, the translation accuracy module 3800 can use a language based feature module 3804 to evaluate language based features in addition to or instead of word based features. For example, in one embodiment, the words appearing in the original message and the translation may be tagged to identify a part of speech (POS) (eg, verb, noun, adjective, etc.) in both messages (eg, open source. POS tagger). Each word in the messages can be tagged according to part of speech, with a different number of tags, using a different set of tags for each language. For example, the sample sentence “The device is easy to use” may be tagged as “The_DT device_NP is_VBZ easy_JJ to_TO use_VB” by a POS tagger indicating the part of speech of each word in the sentence. In this case, the tags are: delimiter (DT), noun phrase (NP), singular present tense verb (VBZ), adjective (JJ), TO (TO), simple verb (VB) : Simple verb). The tags of primary interest in this proposal are verbs, typically followed by adjectives and adverbs.

In a particular example, the original message and translation are tagged separately (eg, using POS taggers), and the resulting tag of each message is counted to identify the number of verbs, adjectives, adverbs, etc. in each message. Simplified verb tags VB due to different types of verbs used in each language (modal verb, infinite verb, past tense verb, future tense verb, etc.) Can be obtained by using maps for all types of verbs in each language. For example, an English verb part-of-speech tag may be mapped to a single verb tag VB as follows; 'VBD' (verb, past tense) => 'VB', 'VBG' (verb, gerund) => 'VB', 'VBN' (verb) , Past participle) => 'VB', 'VBP' (Verb, non third person singular present) => 'VB', and 'VBZ' (verb ), Third person singular present => 'VB'. The POS tag of the tagged message can be replaced with a simplified set of POS tags.

After simplifying the POS tag, the number of verb tags VB can be counted in the original message and translation. Ideally, the number of verbs in each message should be the same, with some exceptions. For example, “was sleeping” in English is translated as “dormais” in French. The English POS tagger can tag two different verbs, “was” and “sleeping”. On the other hand, the French POS tagger can tag a single verb “dormais”. Verbs such as “is”, “was”, and “can” are known as auxiliary verbs in English. Other languages may not have verbs equivalent to these auxiliary verbs, but instead use a single verb as a replacement. To account for these differences in verb use among languages, systems and methods can use thresholds (eg, 2 or 3) for differences in the number of verbs between the original message and the translation. For example, when the difference between the number of verbs in the two messages is greater than 2, the language-based feature module 3804 may consider that the translation may or may not be accurate. This threshold 2 was found through trial and error to produce reasonable results. Other parts of speech (eg, adjectives and adverbs) can be counted and compared between two messages using one or more additional thresholds.

However, in some examples, the user can deceive this translation accuracy check by simply copying and submitting the existing translation as a correction to the translation. In this case, the submission may be classified as a valid correction, but the user may not be compensated for the submission. In some cases, the user can also simply change the case of some words in the existing translation to produce and submit a valid correction. Accordingly, in some embodiments, the systems and methods determine whether existing translations and user submissions are the same. If the existing translations and user submissions are the same (eg, case and capitalization), no compensation may be given to the user.

In certain embodiments, POS tag checks are used to identify cases when the automated translation system fails to accurately identify the language of the original message. For example, the language of the original message may be incorrectly detected when the user's translation correction passes the word count check but fails the POS tag check. Inaccurate language detection is also likely when the number of verbs is equal to zero, or when all tags are nouns in one message and not in the other. For example, the original Spanish message cites “Pizt te enviA³ tan pronto regrese una marcha.” However, if the language is detected to be English, the English POS tagger may be unable to tag the message, and by default, all words can be tagged as nouns. The output of the English POS tagger, for example, [{“tag”: “NP”, “word”: “Pizt”, “stem”: “<unknown>“}, {“tag”: “NN”, “ word ”:“ te ”,“ stem ”:“ te ”}, {“ tag ”:“ NN ”,“ word ”:“ envi＼u00c3＼u00b3 ”,“ stem ”:“ <unknown> “}, {“ tag ”:“ NN ”,“ word ”:“ tan ”,“ stem ”:“ tan ”}, {“ tag ”:“ RB ”,“ word ”:“ pronto ”,“ stem ”:“ pronto ”}, {“Tag”: “JJ”, “word”: “regrese”, “stem”: “<unknown>“}, {“tag”: “NN”, “word”: “una”, “stem”: “ <unknown> "}, {" tag ":" NN "," word ":" marcha "," stem ":" <unknown> "}]. In comparison, the output of the Spanish tagger of the same original message is [{“tag”: “NP”, “word”: “Pizt”, “stem”: “<unknown>“}, {“tag”: “PPX” , “Word”: “te”, “stem”: “t＼u00fa”}, {“tag”: “VLfin”, “word”: “envi＼u00c3＼u00b3”, “stem”: “<unknown>“ }, {“Tag”: “ADV”, “word”: “tan”, “stem”: “tan”}, {“tag”: “ADV”, “word”: “pronto”, “stem”: “ pronto ”}, {“ tag ”:“ VLfin ”,“ word ”:“ regrese ”,“ stem ”:“ regresar ”}, {“ tag ”:“ ART ”,“ word ”:“ una ”,“ stem ” : “Un”}, {“tag”: “NC”, “word”: “marcha”, “stem”: “marcha”}]. The tags “NN”, “RB”, and “PPX” indicate nouns (singular or non-numerical), adverbs, and possessive pronouns, respectively.

Accordingly, in a particular example, the parts of speech of the original message and the translation are compared to determine whether the language was properly identified in the original message. Generally, a language detection failure is likely to occur when one of the messages (eg, a translation) is tagged as having a non-zero number of verbs, and the other message (eg, the original message) is tagged as having no verbs. have. Language detection failure is also likely when all the words in one message are tagged with nouns, while the other message has several types of POS tags (eg nouns, verbs and adjectives).

In various embodiments, translation accuracy is assessed by identifying and reviewing proper nouns in the original message and translation. Generally, when the translation is correct, the proper noun (eg, the names of people and cities) is the same in the translation and the original message. Thus, comparison of two messages that are left untranslated and filtering of common words can be useful as a feature to identify a true translation. In some examples, the presence of such untranslated proper nouns can help improve translation accuracy, but the absence of untranslated proper nouns may not give any information about the translation accuracy. If the proper noun is identified in the original message but is not present in the translation, the accuracy of the translation may be considered to be inaccurate or inaccurate. Penalties can be added to the actual valued scores returned for these intrinsic noun features, which helps to identify incorrect translations and improve translation accuracy. For example, when the proper noun is different between the two messages, the accuracy score for the translation can be reduced by a penalty.

Alternatively or additionally, translation accuracy can be assessed by analyzing and comparing grammar in the original message and translation. Working in multiple languages can make it difficult to parse a tree for every language to understand the grammar of the sentence. Messages are also often written in a chat language, which leads to other grammars that are compared to the explicit or formal speech of the native language.

Accordingly, in order to recognize patterns among grammars of chat languages, sentences can be tagged with POS tags to establish an N-gram language model for POS tags, thereby providing an approximation of the underlying grammatical structure. An n-gram can be defined as a collection of n-sequential words. This n-gram model can typically be for a particular language, and / or can be used to represent n consecutive words in a particular language. In certain implementations, the method of word-based n-grams is extended to parts of speech-based n-grams. That is, a superficial method of parsing a sentence can be used to tag the word of the sentence with a POS tagger. In one approach, BLUE scores are calculated for POS tags rather than actual text.

A trigram language model can be created for POS tagged sentences for each language. For example, the sentence “The device is easy to use” has the POS tagged output of “The_DT device_NP is_VBZ easy_JJ to_TO use_VB”. The word-based trigrams in this sentence are {The, device, is}, {device, is, easy}, {is, easy, to}, {easy, to, use}. The corresponding POS-based trigram may be {DT, NP, VBZ}, {NP, VBZ, JJ}, {VBZ, JJ, TO}, {JJ, TO, VB}.

Trigrams with higher probability can be used to infer the partial structure of the grammar. For example, a trigram language model built for POS tags may have a probability associated with each trigram. Probability can be calculated as the ratio of the number of times a particular trigram occurred in the same text for all trigrams in the compared text corpus. Grammatic trigrams tend to repeat often, and for this reason they will have a higher probability (also known as language model scores). For this reason, when a message receives a higher score for this language model, the message is more likely to be grammatically correct. This score can be useful for detecting cases when a user enters a spam message to receive a reward. The score can also be useful in determining when language detection has failed. For example, the scores of sentences in the wrong language may be lower, since distinct models can be used for each language. Scores can also be useful for detecting when multiple translations are good. Separated models can be trained for human and machine translation for this purpose.

In certain embodiments, language models are trained using translations that have been proven to be accurate. Trigram models can be built for POS tags.

The inherent problem with n-grams of any size is the lack of all possible hypotheses. In this case, a back-off method follows when n-1-gram and n-2-gram are identified. For example, if an unknown POS trigram is found, such as {DT, NP, JJ}, the model will score the language model scores in which the bigrams {DT, NP} and {NP, JJ} represent grammar. You can check retrospectively to see if you have it. If the bigram has no probability (or if the probability is very low), another backoff can be done to check the unigram language model score for {DT}, {NT}, {JJ}. All three models can also be grouped by linear interpolation with more weight on the trigram model, less weight on the bigram model, and much less weight on the unigram model.

The trigrams where POS tags occur most often for chat English are: <IN> <VBD> <PP> = 1.0; <FW> <NN> <IN> = 1.0; <DT> <WP> <VBP> = 1.0; And <RB> </ S1> </ S2> = 1.0. A score of 1.0 indicates a 100% probability that a particular trigram sequence is always grammatical when it occurs. Generally, the chat language uses more pronouns that start with “Wh” and are followed by verbs, which are captured as <DT> <WP> <VBP>. People also tend to end the chat language as an adverb or adjective, as in “You are cool” or “This is awesome”, which is captured as <RB> </ S1> </ S2> and S1 and S2 Is the end of the sentence tags. Thus, these trigrams can be used to recognize at least some structure of the underlying grammar in each language. Language models can be created for any language of interest, and the models can be stored in JSON format and loaded immediately at any time.

In some implementations, after the message is POS tagged, the system and method can count the number of trigrams in the message matching the existing trigram language model. Each trigram in the model can have a probability score associated with it, and for each trigram in the message, the corresponding score of the trigram is identified in the model. In some examples, if the score is greater than a threshold (eg, 0.05), the trigram is counted as a match. Otherwise, the trigram is not counted as a match. The system and method can calculate the number of matching trigrams for the total number of trigrams in a sentence, and this ratio can be used as an actual valued feature to evaluate translation accuracy. For example, a trigram of {easy, to, use} that can occur frequently in grammatical text may have a high probability score (language model score) of about 0.68. In contrast, {easy, use, to} inscriptional trigrams may have a lower probability of occurrence of about 0.04. When compared to a defined threshold of 0.05, non-grammatical trigrams can be filtered out, and the number of trigrams above the threshold can provide a quantitative value for the grammar of the text.

After obtaining a POS tagged message, the probability of a sentence can be calculated using a trigram language model. In some embodiments, the log probability of each trigram in the sentence is determined, and the sum of the log probabilities is calculated. Then, this sum is divided by the number of words in the sentence to get a score for the sentence. The score can be used as a real valued feature to evaluate translation accuracy. The logarithmic probability of the trigram is preferably calculated by linear interpolation of trigram, bigram and unigram probabilities. For example, in the sentence “The device is easy to use”, the POS tagged output is “The_DT device_NP is_VBZ easy_JJ to_TO use_VB”. The POS-based trigrams for the sentence are {DT, NP, VBZ}, {NP, VBZ, JJ}, {VBZ, JJ, TO}, and {JJ, TO, VB}. Each of these trigrams has a probability of occurrence in a particular language corpus. Assuming that the probabilities for each trigram are 0.12, 0.44, 0.32, and 0.89, the combined score for the grammar of the sentence can be calculated as a logarithmic probability. In this case, the log probability is equal to -1.82, calculated as log (0.12) + log (0.44) + log (0.32) + log (0.89). On a relative scale, the logarithmic probability provides a numerical representation of the grammar of a sentence.

In addition to or instead of the word-based and language-based features described above, the translation accuracy module 3800 can use the word placement module 3806 to evaluate the placement of words between the original message and the translation. To generate a list of potential word placement pairs, the translation accuracy check algorithm is run with word-based features alone in the database, and the parallel corpus is extracted as translation pairs with probability> 0.90. This indicates that only good quality translation message pairs are filtered to create a parallel corpus. 100K sentences are collected for English and French pairs and English and Spanish pairs. The parallel corpus for these 100K sentences is supplied to a statistical machine translation toolkit (ie GIZA ++) to extract word placement. The toolkit extracts approximately 25-30K word placement pairs and associated probability scores for these pairs.

Considering that there are generally multiple word alignments per word, it may be beneficial to select a placement with a probability score greater than a certain threshold. However, despite the threshold, more than one batch per word can be obtained, most of which are typically due to spelling errors or the same word in different tenses (eg past tense or future tense). Sample word assignments between English and French are shown in Table 3. Separate word placement can be obtained for two possible sequences of two languages (eg English vs. French and French vs. English). Thus, two word batch files can be extracted for each language pair. The word placement for the source language (i.e., the language of the original message) versus the target language (i.e., the language of translation) can be referred to as source placement, while the word placement for the target language vs. source language is referred to as target placement Can be.

Example of word placement between English and French English French Troops troupes troupe Attacks attaques attaques Smoke fumee fume fumez fumee

Source batch and target batch can be loaded into two separate files. For each word in the original message that is also present in the source batch, the system and method determine whether at least one corresponding word is present in the translation and also in the target batch. Then, the same process is applied to the translated sentences. Finally, the percentage of words found in the word batch file is returned as a feature to evaluate translation accuracy.

In certain implementations, word placement is extracted for language pairs that include English as one of the languages. For example, the language pair can be English combined with one of Spanish, French, Polish, Portuguese, Dutch, German, Danish, Swedish, Turkish, Italian, and Russian, with a total of 11 language pairs and 11 x 2 = 22 Word batch file. For language pairs that do not contain English (eg, French to Russian translation), English can be used as an intermediate language to extract batches. For example, when a French-to-Russian translation is verified, French-to-English can be used to extract word placement from the original message, and Russian-to-English can be used to extract word placement from the translation. The intersection of these two sets gives the percentage of matches in word placement in both messages. This method can be extended to any language pair.

In various embodiments, user confidence is calculated to provide an indication of whether the user's translation submission is trustworthy. In addition to the other approaches described herein, the user can be almost trusted based on the user's history of translation corrections. The number of corrected translations and the number of compensated translations can be stored in a data table for various users. This allows the percentage of corrections made by the player to be calculated, and translation corrections can be approved, at least to a certain extent, based on a particular user's confidence threshold. This makes it easy to distinguish spammers from true players who correct translations.

39 is a flow diagram of a method 3900 for correcting translation errors, according to certain embodiments. Method 3900 includes providing a text message chat system to a plurality of users of the online game (step 3902). The original text message is received in the first language from the first user of the online game (step 3904). An initial translation is created in the second language of the original text message (step 3906). The original text message and initial translation are provided to the second user of the online game (step 3908). For example, the second user, on the display of the client device, can view the two translations together or separately. Translation correction to raise an error in the initial translation is received from the second user (step 3910). The method 3900 can also include identifying the most accurate translation correction (step 3912) from a plurality of translation corrections (including translation corrections from a second user). Alternatively or additionally, the method 3900 includes evaluating the accuracy of translation correction from a second user (step 3914) using at least one of word-based features, language-based features, and word placement features.

The above-described word-based and language-based features (eg, four word-based features and four language-based features) can be fitted using a linear regression model. After training, the model preferably returns the actual valued number for each translation pair, and a threshold can be used to classify whether each translation pair is good or bad. For example, if the features return a numeric value of x1, x2, x3, ... x8, then the regression equation is y = a1 * x1 + a2 * x2 +. . . + a8 * x8, where a1, a2,. . . a8 is a coefficient obtained by modeling a linear regression equation, and y is an output value. The preferred value for the threshold y was found to be 0.65 after the experiment with precision and recall using the ROC curve.

After adding the word placement-based features described above and re-running linear regression for word-based and language-based features, the preferred threshold is changed to 0.76. The addition of word placement based features also increases the AUC value from 0.853 to 0.976.

Table 4 shows the regression coefficients obtained by training 13k sentences for each of the word-based features, language-based features, and word placement features. The results in the table indicate that the emoji feature and character count, after normalization, have a small regression coefficient, which means that these features have been found to make a small contribution to the accuracy of the translation.

Coefficient of regression for translation accuracy features Features (Featur) Coefficient Numbers 5.39E-01 Emojis 6.99E-20 Character Counts 1.76E-17 Word Counts 2.07E-01 Count <VB> 9.55E-02 Trigram Match 1.67E-02 Trigram Probs -2.24E-02 Untranslated nouns 4.18E-02 Word Alignment Match 4.48E-01

The regression model can be evaluated by 10-fold cross validation for 13k sentences of French vs. English and Spanish vs. English pairs. The gold standard label for the data is calculated using some human supervised BING translation. Table 5 shows the precision, recall, accuracy, and F-measure values expressed as percentages in the evaluation. Precision represents the ratio of True Positive (TP) translation pairs to all Translation Pairs marked as true by our method. Recall is the ratio of True Positive (TR) translation pairs to all Translation pairs known as true pairs. Accuracy is the ratio of the sum of the results marked as True Positive (TP) and True Negative (TN) to the total test set size. F-measure is the harmonic mean of precision and recall. These metrics can be used in classification tasks to measure system performance and reliability. 13k sentences can be used as a training set, and 400 hand-annotated sentences are used for testing. Since the test set is small, the number for testing is smaller than the number for 13k sentences.

Translation accuracy study results in 13,000 sentences Measurement 13k sentences
(10 fold cv) 400 sentences 13k sentences
(Word-based features) Precision 90 77 88.2 Recall 79.3 72 79.3 F-Measure 84.31 74.41 83.56 Accuracy 97.3 96 97

Table 6 shows the results of the 10-fold cross validation for the 13k sentence data set. At this time, W means the use of word-based features, L means the use of language-based features, and A means the use of word-based features. The result is that word-based features help improve precision, and word-position features help significantly improve recall. The word-based feature gives only a small elongation for both precision and recall. Recall, in general, indicates how accurately bad decisions are detected in the entire data set. The results in the table show that adding word placement based features improves recall. Precision indicates how accurately a good translation is predicted in the entire translation marked as correct by the system. The results in the table show that word-based features improve precision.

Results of 10-fold cross validation on 13,000 sentence data sets Way Precision Recall F-Measure Accuracy W 88.2 79.3 83.56 97 L 51.9 26.6 35.17 92.5 A 55 96.7 70.11 93.9 WL 90 79.3 84.56 97.3 WA 80.5 96.6 87.81 97.7 LA 57.8 96.6 72.32 94.2 WLA 80.7 96.8 88.01 97.7

Table 7 shows the results of adjusting various word-based, language-based, and word-positioning features with different machine algorithms, according to certain embodiments. The results so far have demonstrated the use of a linear regression technique to group different features together. Machine learning algorithms exist so that they can be used to group together variables (features in context) to produce better ensemble results than individual parts. Linear regression presents a one-dimensional method for combining variables. Two-dimensional and multidimensional methods for combining variables exist in the machine learning algorithm literature. This algorithm is employed to find a more optimal way to combine the features used in the task of predicting good translation pairs.

The results in Table 7 are obtained by combining features with various machine learning algorithms. The data set used contains 13k sentences, and the parameters are adjusted using the gridSearch algorithm. From the methods listed in the table, the gradient boosting classifier and the random forest method are ensemble-based methods and explain why these methods provide better results. Gradient Boosting Machines (GBMs) and random forests, although GBMs take longer to train, provide very good results. However, since the model only needs to be trained once, the training time is largely irrelevant.

Fitting results of translation accuracy features with machine algorithms ML algorithm parameter
(Params) Precision
(Precision) Reproducibility
(Recall) accuracy
(Accuracy) Linear regression
(Linear Regression) Least squares
(Least Squares) 80.7 96.8 97.8 Perceptron
(Perceptron) 98.8 74.3 97.8 Leeds sorter
(Ridge Classifier) 96.1 71.6 97.3 Gaussian Naive Bayes 80.7 90.3 97.4 Crystal tree
(Decision Tree) 91.3 90.9 98.2 Logistic regression
(logistic Regression) L1 norm, C = 100, tol = 0.001 92.1 91.4 98.4 SVM C = 1000, gamma = 0.001 91.3 92.8 98.5 Gradient Boosting Machines n_estimators = 100 92.4 93 98.8 Random forest
(Random Forests) n_estimators = 10 93.3 92.8 98.8

The final translation accuracy check algorithm is tested against a sustained test set of 3045 English-French sentence pairs. Results are shown in Table 8. WLAplain features perform poorly due to the fact that word placement is extracted from a clear language database. Messages are modified after a series of transformations, so a drop in results can be found. The word placement is extracted from the chat term database, and fewer sentences are done with some smoothing to avoid zero scores in the features. WLA shows the results for the chat term database. Bad precision with WLA features indicates that many exact translations are rejected. The features are estimated with linear regression, as the ransom forest overfits the results. Generally, these results are used as the basis for selecting the final set used in the algorithm. Higher accuracy is generally desirable, but higher precision compensates for more true pairs of translations entered by the user, and high recall ensures fewer unsorted and incorrect entries are correct. The WLA feature set is preferred, according to the results in the table. A threshold of 0.75 can be selected for a higher recall (e.g., to get the least possible malicious entries marked as correct), and a threshold of 0.68, hopes for higher precision. It can be selected in some cases.

Results of the translation accuracy checking algorithm Characteristic Critical
(Threshold) Precision
(precision) Reproducibility
(recall) F-measure accuracy
(Accuracy) Area under ROC WL 0.75 85 59 69.65 96 0.723 WLAplain 0.75 19 79 30.63 80 0.631 WLA 0.75 65 93 76.51 95 0.883 WLA 0.68 76 85 80.24 96 0.853

While the present invention has been specifically shown and described with respect to certain preferred embodiments, details may be made within the scope and spirit of the invention as defined by the appended claims, and various modifications may be made to those skilled in the art. Can be understood.

102: client
104: chat client system
106: computer network
108: chat server
112: chat host system
114: Communication Conversion and Translation (CTT) System
110: translation server
116: translation module

Claims (21)

A method executed by one or more computer processors, the method comprising:
Obtaining an original text message in a first language written by the first user;
Obtaining a translated text message in a second language of the original text message;
Comparing the plurality of features associated with the original text message with each of the plurality of features associated with the translated text message, wherein the plurality of features include (i) word count, character count, emoji, Comprising at least one word-based feature selected from the group consisting of numbers and punctuation marks, and (ii) at least one language-based feature comprising parts of speech selected from the group consisting of verbs, nouns, adverbs and adjectives; And
Calculating a metric based on the comparison of the features, and determining the accuracy of the translated text message based on the calculated metric. According to claim 1,
The translated text message is written by a second user. According to claim 1,
The translated text message includes a correction of a previous translated text message of the original text message. According to claim 1,
Comparing the plurality of features,
And determining a difference in the at least one word-based feature between the original text message and the translated text message. According to claim 1,
Comparing the plurality of features,
And determining a difference in the number of parts of speech between the original text message and the translated text message. According to claim 1,
The step of calculating the metric,
And determining the placement of at least one word in the original text message and at least one respective word in the translated text message. According to claim 1,
The metric is,
A method further based on the grammar of the translated text message. According to claim 1,
The step of calculating the metric,
Generating a part-of-speech n-gram representation of the translated text message; And
And calculating a probability of an n-gram representation for the second language. According to claim 1,
The step of calculating the metric,
And fitting the features into a linear regression model. According to claim 1,
The step of calculating the metric,
And fitting the features with one or more machine learning algorithms. In the system,
A non-transitory computer-readable medium on which instructions are stored; And
At least one processor configured to execute the instructions,
The at least one processor,
Obtain an original text message in a first language written by the first user;
Obtain a translated text message in a second language of the original text message;
Compare the plurality of features associated with the original text message with each of the plurality of features associated with the translated text message, wherein the plurality of features include: (i) word count, character count, emoji, number and At least one word-based feature selected from the group consisting of punctuation marks, and (ii) at least one language-based feature comprising parts of speech selected from the group consisting of verbs, nouns, adverbs and adjectives; And
A system for performing an operation that calculates a metric based on a comparison of the features and determines the accuracy of the translated text message based on the calculated metric. The method of claim 11,
The translated text message is written by a second user. The method of claim 11,
And the translated text message comprises correction of a previous translated text message of the original text message. The method of claim 11,
Comparing the plurality of features,
And determining a difference in the at least one word-based feature between the original text message and the translated text message. The method of claim 11,
Comparing the plurality of features,
And determining a difference in the number of parts of speech between the original text message and the translated text message. The method of claim 11,
The operation of calculating the metric,
And determining the placement of at least one word in the original text message and at least one respective word in the translated text message. The method of claim 11,
The metric is,
A system further based on the grammar of the translated text message. The method of claim 11,
The operation of calculating the metric,
Generate a part-of-speech n-gram representation of the translated text message; And
And calculating a probability of an n-gram representation for the second language. The method of claim 11,
The operation of calculating the metric,
A system comprising fitting the features into a linear regression model. The method of claim 11,
The operation of calculating the metric,
A system comprising fitting the features with one or more machine learning algorithms. A non-transitory computer-readable recording medium comprising executable instructions, comprising:
The executable instructions are executable by one or more processors and cause the one or more processors to perform operations,
The operation is,
Obtaining an original text message in a first language written by the first user;
Obtaining a translated text message in a second language of the original text message;
Comparing the plurality of features associated with the original text message with each of the plurality of features associated with the translated text message, wherein the plurality of features include (i) word count, character count, emoji, Comprising at least one word-based feature selected from the group consisting of numbers and punctuation marks, and (ii) at least one language-based feature comprising parts of speech selected from the group consisting of verbs, nouns, adverbs and adjectives; And
And calculating a metric based on the comparison of the features and determining the accuracy of the translated text message based on the calculated metric.

Images