KR20230108432A - Translation platform - Google Patents

Abstract

In one embodiment of the present disclosure for solving the above problems, a method for providing a professional translation service platform, performed in one or more processors of a computing device, is disclosed. The method includes generating one or more translation models corresponding to one or more domains, and when subscription request information is received from a user terminal, corresponding to the user terminal among the one or more translation models based on the received subscription request information. Determining an optimal translation model to generate a translated sentence corresponding to the received input sentence by utilizing the optimal translation model when translation request information including an input sentence is received from the user terminal, the user When re-translation request information related to the translated sentence is received from a terminal, determining one or more requesting translator terminals to request re-translation based on the received re-translation request information, sending the re-translation request information to the one or more requesting translator terminals. and determining to transmit a first retranslated sentence among the received one or more retranslated sentences to the user terminal when one or more retranslated sentences are received from the one or more requesting translator terminals.

Description

Translation platform {Translation platform}

The present disclosure relates to translation technology using artificial intelligence, and more specifically, to provide a computing device that provides a more professional translation service according to user needs.

As the amount of information that can be processed increases with the development of computer technology, artificial intelligence is evolving at a rapid pace, and interest in machine translation is increasing according to the development of artificial intelligence. As a server platform for such machine translation, there are various translators including Google Translator and Naver Translator.

Recently, as the level of machine translation has rapidly increased through deep learning algorithms, high-quality translation services are currently being supplied. Such machine translation has the advantage of being able to quickly provide a translation of the original text in real time and being inexpensive.

However, conventional machine translation has limitations in outputting a natural translation considering the context of an original sentence related to knowledge in a special specialized field. Specifically, since conventional machine translation is a general-purpose translation applied to multiple fields, there is a concern that words used in a specific field may not be properly translated or the meaning may not be saved. For example, in the IT field, words such as master/slave can be interpreted as a 'main system' that is the subject of an operation in performing a task, and a 'subordinate system' that plays a subordinate role. As the machine translation service platform of 's simply recognizes the words as master/slave, etc., there is a concern that the accuracy of the generated sentences may be lacking.

Therefore, there may be a demand in the industry for a platform that provides more professional translation services in response to the needs of individual users who want to use translation services.

The problem to be solved by the present disclosure is to solve the above problems, and to provide a computing device that provides a more professional translation service in response to the needs of individual users.

The problems to be solved by the present disclosure are not limited to the problems mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the description below.

In order to solve the above problems, a method for providing a professional translation service platform performed in one or more processors of a computing device according to various embodiments of the present disclosure is disclosed. The method includes generating one or more translation models corresponding to one or more domains, and when subscription request information is received from a user terminal, corresponding to the user terminal among the one or more translation models based on the received subscription request information. Determining an optimal translation model to generate a translated sentence corresponding to the received input sentence by utilizing the optimal translation model when translation request information including an input sentence is received from the user terminal; When re-translation request information related to the translated sentence is received from a terminal, determining one or more requesting translator terminals to request re-translation based on the received re-translation request information, sending the re-translation request information to the one or more requesting translator terminals. and determining to transmit a first retranslated sentence among the received one or more retranslated sentences to the user terminal when one or more retranslated sentences are received from the one or more requesting translator terminals.

In an alternative embodiment, the generating of one or more translation models corresponding to one or more domains may include obtaining a training data set comprising a plurality of training data, based on the plurality of training data for one or more domains. The method may include constructing one or more training data subsets by classifying and generating the one or more translation models by performing learning on each of one or more neural network models through each of the one or more training data subsets.

In an alternative embodiment, the step of classifying the plurality of learning data based on one or more domains and constructing one or more subsets of the learning data may include, as a result of analyzing the meaning of each of the plurality of learning data, each of the plurality of learning data Identifying a domain of each of the plurality of learning data based on at least one of the number of repetitions for each word included in the semantic analysis result and a predetermined subject, and the one or more training data sub-domains based on the domain of each of the plurality of training data. It may include building a set.

In an alternative embodiment, the subscription request information is information that is a criterion for determining the optimal translation model corresponding to the user terminal, and the translation request information includes user language identification information related to a starting language and a target language, and It includes user domain identification information, and the retranslation request information is information about a retranslation request for the translated sentence, and may be generated based on the translation request information.

In an alternative embodiment, generating a first input sentence by performing reverse translation on the translated sentence, and calculating a translation reliability of the translated sentence based on a comparison between the generated first input sentence and the input sentence. , determining whether to transmit the translation reliability to the user terminal, and determining whether to transmit the retranslation request information to the one or more requested translator terminals based on the translation reliability.

In an alternative embodiment, the method further includes receiving translation participation information from each of a plurality of translator terminals and constructing a translator database based on the translation participation information of each of the plurality of translator terminals, wherein the translation participation information includes: Translator language identification information and translator domain identification information may be included.

In an alternative embodiment, the step of determining one or more requesting translator terminals to request re-translation based on the received re-translation request information includes identifying one or more candidate translator terminals corresponding to the re-translation request information by referring to the translator database. and determining the one or more requesting translator terminals based on the identified translation evaluation scores of the one or more candidate translator terminals, wherein the translation evaluation scores are determined by the translation performed by each translator terminal corresponding to a previous point in time. It may be an evaluation score for

In an alternative embodiment, the step of determining one or more translator terminals to request re-translation based on the received re-translation request information includes one or more candidate translator terminals corresponding to the domain identification information of the user terminal by referring to the translator database. determining to transmit a translation service participation request to the one or more candidate translator terminals, and determining the one or more requesting translator terminals based on participation responses of the one or more candidate translator terminals to the translation service participation request. steps may be included.

In an alternative embodiment, when one or more retranslated sentences are received from the one or more requesting translator terminals, the step of determining to transmit a first retranslated sentence among the received one or more retranslated sentences to the user terminal may include: and determining the first retranslated sentence from among the one or more retranslated sentences based on evaluation score information of each translator terminal.

In an alternative embodiment, determining to transmit the received one or more retranslated sentences to the user terminal, receiving selection information for at least one retranslated sentence among the one or more retranslated sentences from the user terminal, and receiving the received one or more retranslated sentences. The method may further include identifying the first retranslated sentence from among the one or more retranslated sentences based on the selection information.

In an alternative embodiment, generating additional training data by matching and labeling the input sentence and the first retranslated sentence, and performing additional learning on the optimal translation model through the additional training data to perform the optimal translation model. It may further include the step of updating.

According to another embodiment of the present disclosure, a computing device for providing a professional translation service platform. The computing device includes a processor including one or more cores, a memory for storing program codes executable in the processor, and a network unit for transmitting and receiving data to and from a user terminal and a translator terminal, wherein the processor includes one or more domains corresponding to one or more domains. The above translation models are generated, and when subscription request information is received from the user terminal, an optimal translation model corresponding to the user terminal is determined from among the one or more translation models based on the received subscription request information, and from the user terminal When translation request information including an input sentence is received, a translated sentence corresponding to the received input sentence is generated using the optimal translation model, and retranslation request information related to the translated sentence is received from the user terminal , Based on the received re-translation request information, one or more commissioned translator terminals to be requested for re-translation are determined, the re-translation request information is transmitted to the one or more commissioned translator terminals, and one or more re-translation requests are sent from the one or more commissioned translator terminals. When a sentence is received, it may be determined to transmit a first retranslated sentence among the one or more received retranslated sentences to the user terminal.

According to various embodiments of the present disclosure, it is possible to provide a computing device that provides a more professional translation service in response to the needs of individual users.

The effects of the present disclosure are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

Various aspects are now described with reference to the drawings, wherein like reference numbers are used to collectively refer to like elements. In the following embodiments, for explanation purposes, numerous specific details are set forth in order to provide a thorough understanding of one or more aspects. However, it will be apparent that such aspect(s) may be practiced without these specific details.
1 shows various aspects of a computing device for providing a professional translation service platform related to an embodiment of the present disclosure.
It shows a conceptual diagram showing a system that can be implemented.
2 illustrates a block diagram of an embodiment of the present disclosure and a computing device for providing a professional translation service platform.
3 illustrates an exemplary diagram of one or more translation models related to an embodiment of the present disclosure.
4 shows an exemplary flow chart of a method for providing a professional translation services platform related to one embodiment of the present disclosure.
5 is a flowchart illustratively illustrating steps for performing a method for providing a professional translation service platform related to an embodiment of the present disclosure.
6 is a schematic diagram illustrating one or more network functions related to one embodiment of the present disclosure.

Various embodiments are now described with reference to the drawings. In this specification, various descriptions are presented to provide an understanding of the present disclosure. However, it is apparent that these embodiments may be practiced without these specific details.

The terms “component,” “module,” “system,” and the like, as used herein, refer to a computer-related entity, hardware, firmware, software, a combination of software and hardware, or an execution of software. For example, a component may be, but is not limited to, a procedure, processor, object, thread of execution, program, and/or computer running on a processor. For example, both an application running on a computing device and a computing device may be components. One or more components may reside within a processor and/or thread of execution. A component can be localized within a single computer. A component may be distributed between two or more computers. In addition, these components can execute from various computer readable media having various data structures stored thereon. Components may be connected, for example, via signals with one or more packets of data (e.g., data and/or signals from one component interacting with another component in a local system, distributed system) to other systems and over a network such as the Internet. data being transmitted) may communicate via local and/or remote processes.

In addition, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or”. That is, unless otherwise specified or clear from the context, “X employs A or B” is intended to mean one of the natural inclusive substitutions. That is, X uses A; X uses B; Or, if X uses both A and B, "X uses either A or B" may apply to either of these cases. Also, the term "and/or" as used herein should be understood to refer to and include all possible combinations of one or more of the listed related items.

Also, the terms "comprises" and/or "comprising" should be understood to mean that the features and/or components are present. However, it should be understood that the terms "comprises" and/or "comprising" do not exclude the presence or addition of one or more other features, elements, and/or groups thereof. Also, unless otherwise specified or where the context clearly indicates that the singular form is indicated, the singular in this specification and claims should generally be construed to mean "one or more".

Those skilled in the art will further understand that the various illustrative logical blocks, components, modules, circuits, means, logics, and algorithm steps described in connection with the embodiments disclosed herein may be implemented using electronic hardware, computer software, or combinations of both. It should be recognized that it can be implemented as To clearly illustrate the interchangeability of hardware and software, various illustrative components, blocks, configurations, means, logics, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented in hardware or as software depends on the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application. However, such implementation decisions should not be interpreted as causing a departure from the scope of this disclosure.

The description of the presented embodiments is provided to enable any person skilled in the art to use or practice the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art of this disclosure. The general principles defined herein may be applied to other embodiments without departing from the scope of this disclosure. Thus, the present disclosure is not limited to the embodiments presented herein. This disclosure is to be interpreted in the widest light consistent with the principles and novel features presented herein.

In this specification, a computer means any kind of hardware device including at least one processor, and may be understood as encompassing a software configuration operating in a corresponding hardware device according to an embodiment. For example, a computer may be understood as including a smartphone, a tablet PC, a desktop computer, a laptop computer, and user clients and applications running on each device, but is not limited thereto.

Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings.

Although each step described in this specification is described as being performed by a computer, the subject of each step is not limited thereto, and at least a part of each step may be performed in different devices according to embodiments.

1 illustrates a conceptual diagram illustrating a system in which various aspects of a computing device for providing a professional translation service platform related to an embodiment of the present disclosure may be implemented.

A system according to embodiments of the present disclosure may include a computing device 100 , a user terminal 10 , a translator terminal 20 , an external server 30 , and a network. Components shown in FIG. 1 are exemplary, and additional components may exist or some of the components shown in FIG. 1 may be omitted. The computing device 100, the user terminal 10, the translator terminal 20, and the external server 30 according to embodiments of the present disclosure transmit data for the system according to the embodiments of the present disclosure through a network. can transmit and receive each other.

Networks according to embodiments of the present disclosure include Public Switched Telephone Network (PSTN), xDigital Subscriber Line (xDSL), Rate Adaptive DSL (RADSL), Multi Rate DSL (MDSL), and Very High Speed DSL (VDSL). , Various wired communication systems such as Universal Asymmetric DSL (UADSL), High Bit Rate DSL (HDSL), and Local Area Network (LAN) can be used.

In addition, the networks presented herein include Code Division Multi Access (CDMA), Time Division Multi Access (TDMA), Frequency Division Multi Access (FDMA), Orthogonal Frequency Division Multi Access (OFDMA), Single Carrier-FDMA (SCFDMA) and other systems. Various wireless communication systems such as

The network according to the embodiments of the present disclosure may be configured regardless of its communication mode, such as wired and wireless, and is composed of various communication networks such as a personal area network (PAN) and a wide area network (WAN). It can be. In addition, the network may be the known World Wide Web (WWW), or may use a wireless transmission technology used for short-range communication such as Infrared Data Association (IrDA) or Bluetooth. The techniques described herein may be used in the networks mentioned above as well as other networks.

According to an embodiment of the present disclosure, the user terminal 10 is a terminal capable of receiving information related to a translated sentence corresponding to an input sentence through information exchange with the computing device 100, which means a terminal possessed by a user. can do. For example, the user terminal 10 may be a terminal related to a user who wants to translate an input sentence composed of a specific language into another language or a user who wants to perform inspection on the translated sentence. The user may receive translated sentences and retranslated sentences corresponding to input sentences through the user terminal 10 .

The translator terminal 20 of the present disclosure provides a retranslated sentence corresponding to an input sentence related to a translation request of the user terminal 10 through information exchange with the computing device 100, and is a terminal capable of receiving a reward corresponding thereto. , which may refer to a terminal possessed by a user (or translator) performing translation work. For example, the translator terminal 20 transmits a retranslated sentence related to a result of performing translation on the input sentence received from the user terminal 10 to the computing device 100, and corresponds to the result of performing the translation from the computing device 100. Compensation information can be obtained. A user (or translator) performing a translation job may receive an input sentence and compensation information related to the translation job through the translator terminal 20 .

These user terminals 10 and translator terminals 20 may refer to any type of entity(s) in a system having a mechanism for communication with the computing device 100 . For example, the user terminal 10 and the translator terminal 20 may include a personal computer (PC), a note book, a mobile terminal, a smart phone, a tablet PC, and a tablet PC. It may include a wearable device and the like, and may include all types of terminals capable of accessing wired/wireless networks. In addition, the user terminal 10 and the translator terminal 20 may include an arbitrary server implemented by at least one of an agent, an application programming interface (API), and a plug-in. Also, the user terminal 10 and the translator terminal 20 may include an application source and/or a client application. In FIG. 1, only one user terminal and one translator terminal associated with the computing device 100 are illustratively illustrated, but the computing device of the present disclosure is associated with a plurality of user terminals and a plurality of translator terminals to provide a translation service platform. It will be apparent to those skilled in the art that it can be provided.

According to an embodiment of the present disclosure, the external server 30 may be a server that stores learning data related to learning of a translation model. The translation model of the present disclosure may be a neural network model that takes an input sentence as an input and outputs a translated sentence corresponding to the input sentence. For example, the external server 30 may be a server that stores a plurality of input sentences corresponding to various languages and translation information corresponding to each of the plurality of input sentences. Information stored in the external server 30 may be used as training data, verification data, and test data for training the neural network in the present disclosure. That is, the external server 30 may be a server that stores information about a data set for learning the deep learning model of the present disclosure.

The computing device 100 of the present disclosure may build a learning data set based on input sentences composed of various languages from the external server 30 and translated sentence information corresponding to each input sentence, and one through the training data set. By performing learning on a plurality of neural network models including one or more network functions, a plurality of translation models performing translation tasks corresponding to one or more specialized fields (or domains) may be generated. In addition, the computing device 100 of the present disclosure may perform augmentation on learning data by transmitting the retranslated sentence received from the translator terminal 20 to the external server 30 in response to the input sentence.

The external server 30 is a digital device, and may be a digital device equipped with a processor and memory and computing capability, such as a laptop computer, a notebook computer, a desktop computer, a web pad, and a mobile phone. The external server 30 may be a web server that processes services. The types of servers described above are only examples and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may receive an input sentence from the user terminal 10 and may provide a professional translation service for the received input sentence.

Specifically, the computing device 100 may generate one or more translation models corresponding to one or more domains. Here, one or more domains are references for classifying a plurality of learning data, and may be related to various specialized fields. For example, one or more domains may be related to various specialized fields such as medical field, legal field, IT field, or economic field. The computing device 100 may generate one or more translation models corresponding to each of one or more domains related to various specialized fields.

For example, the computing device 100 corresponds to each of one or more domains, a first translation model performing translation related to the medical field, a second translation model performing translation related to the legal field, and translation related to the IT field. A third translation model performing translation and a fourth translation model performing translation related to the economic field may be generated. The detailed description of one or more domains and translation models generated corresponding to each of the domains described above is only an example, and the present disclosure is not limited thereto.

That is, the computing device 100 may generate one or more translation models that perform translation specialized for each of various fields. For example, as shown in FIG. 3 , the computing device 100 corresponds to one or more domains. to generate a legal translation model 210, an IT translation model 220, and an economic translation model 230, and at least one of the legal translation model 210, the IT translation model 220, and the economic translation model 230 may be determined as an optimal translation model corresponding to the user terminal 10. Accordingly, it is possible to provide a translation model that performs professional translation according to the individual needs of each of a plurality of users. Therefore, the user can obtain a more professional translation in response to the input sentence through a translation model that meets his or her needs.

Also, when receiving retranslation request information from the user terminal 10, the computing device 100 may determine one or more requesting translator terminals that have requested retranslation based on the received retranslation request information. The re-translation request information may be information about a re-translation request for the translated sentence.

Specifically, the computing device 100 may provide an optimal translation model that meets the needs of the user of the user terminal 10 . The computing device 100 may generate a translated sentence by processing an input sentence to be translated by the user of the user terminal 10 as an input to the corresponding optimal translation model. As the computing device 100 transmits the output translated sentence to the user terminal 10, the user terminal 10 may receive the translated sentence corresponding to the input sentence. The user of the user terminal 10 may generate retranslation request information in the process of reviewing the corresponding translated sentence and transmit it to the computing device 100. When it is determined that the meaning or expression of the translated sentence is not appropriate, an input may be applied to the user terminal 10 to generate retranslation request information.

The computing device 100 may receive retranslation request information related to a corresponding translation sentence (ie, a translation related to an output of a translation model) from the user terminal 10 . In other words, when it is determined that the translation related to the output of the optimal translation model is not appropriate, the user terminal 10 may generate and transmit retranslation request information related to the translation to the computing device 100 . When receiving retranslation request information, the computing device 100 may determine one or more requesting translator terminals to request retranslation. In this case, one or more requested translator terminals to which the computing device 100 requests retranslation of the input sentence of the user terminal 10 may be a translator terminal corresponding to the input sentence received from the user terminal 10 . In detail, the computing device 100 may determine which domain among one or more domains the input sentence is associated with based on a semantic analysis result of the input sentence. The computing device 100 may determine a translator terminal corresponding to a domain related to an input sentence as one or more requested translator terminals to request retranslation of the input sentence. For example, if the input sentence is "If there is no provision in the law regarding civil affairs, the customary law shall prevail, and if there is no customary law, then it shall be governed by the law." , it is possible to determine translator terminals corresponding to the legal field among a plurality of translator terminals as one or more requested translator terminals.

To this end, the computing device 100 may construct a translator database by receiving translation participation information from a plurality of translator terminals. The translation participation information may be generated in relation to a point in time at which translator terminals intend to participate in translation, and may be information for identifying each of a plurality of translator terminals. Also, the translation participation information may include translator language identification information and translator domain identification information. The translator language identification information may be information for identifying a language that can be used by a user of each translator terminal (ie, a language that can perform translation). The translator domain identification information may be information related to one or more domains (or fields) in which each translator terminal can perform translation professionally. For example, the translator language identification information of the first translator terminal may include information that the first user (or the first translator) can use Korean, English, and Japanese, and the translator domain identification information may include information in the IT field. It may include information that the related translation can be professionally performed. The detailed description of the language identification information and domain identification information of the first translator terminal described above is only an example, and the present disclosure is not limited thereto. In other words, the computing device 100 may determine one or more requested translator terminals corresponding to the input sentence received from the user terminal 10 by referring to the translator database. In addition, the computing device 100 may transmit retranslation request information including the input sentence to one or more commissioned translator terminals determined to correspond to the input sentence.

Accordingly, each of the one or more commissioned translator terminals 20 may receive and display the input sentence of the user terminal 10 . That is, a user related to translation may generate a retranslated sentence based on an input sentence displayed through each requesting translator terminal 20 and transmit the generated retranslated sentence to the computing device 100 .

The computing device 100 may receive one or more retranslated sentences from each of the one or more requested translator terminals 20 . Also, the computing device 100 may determine a first retranslated sentence related to an optimally translated sentence among one or more retranslated sentences. For example, when one or more retranslated sentences are received, the computing device 100 may determine a first retranslated sentence based on translation evaluation scores of requesting translator terminals related to generation of each retranslated sentence. The translation evaluation score may be an evaluation score for translation performed by each translator terminal corresponding to this point in time. As a specific example, the translation evaluation scores of the first translator terminal, the second commissioned translator terminal, and the third commissioned translator terminal are 50, 76, and 80, respectively, and the retranslation first subsentence and retranslation from each commissioned translator terminal When receiving the second sub-sentence and the re-translated third sub-sentence, the computing device 100 identifies the one with the highest translation evaluation score of the third requesting translator terminal and transmits the re-translated third sub-sentence generated in response to the corresponding translator terminal. It can be determined by the first retranslated sentence. The detailed numerical description of the translation evaluation score corresponding to each requesting translator terminal described above is only an example, and the present disclosure is not limited thereto.

That is, the computing device 100 may determine a first retranslated sentence more suitable for the input sentence of the user terminal 10 from among one or more retranslated sentences received from each of one or more requesting translator terminals and transmit it to the user terminal 10. In other words, the computing device 100 determines a commissioned translator terminal that can perform more professional translation in response to a translation request from the user terminal 10 and associates it with the user terminal 10 to provide professional translation information. can provide In this case, since the requested translator terminals that the computing device 100 associates with the user terminal 10 are terminals related to the user who translates the corresponding domain through the translation request information of the user terminal 10, the input sentence In response, natural and appropriate translation sentences can be provided. In addition, since the translated sentence is a sentence translated by a professional translator in the field, it is possible to provide a translated sentence for a language expression or language structure that the machine cannot understand, and it is possible to provide a translated sentence reflecting a more appropriate expression. Therefore, it is possible to improve the user's service satisfaction.

Also, the computing device 100 may augment the learning data based on the input sentence and the first retranslated sentence. Specifically, the computing device 100 may generate additional learning data by matching and labeling the input sentence and the first retranslated sentence. In addition, the computing device 100 may perform additional learning on the optimal translation model corresponding to the user terminal 10 through additional learning data and update the optimal translation model. In this case, the updated optimal translation model may be learned to output a first retranslated sentence generated by a professional translator in response to the input sentence related to the translation request of the user terminal 10 as an input. That is, the computing device 100 augments learning data through each of various input sentences received from a plurality of user terminals and various translated sentences of corresponding translator terminals, and updates an optimal translation model through the augmented learning data. can do. Accordingly, the translation model can be advanced over time by the above-described process of the computing device 100 . This may cause an effect of improving reliability and accuracy of the translated sentence output through the translation model.

In one embodiment, the computing device 100 may be a terminal or a server, and may include any type of device. The computing device 100 is a digital device, such as a laptop computer, a notebook computer, a desktop computer, and a web browser. It may be a digital device equipped with a processor and having an arithmetic capability equipped with a memory, such as a pad or a mobile phone. The computing device 100 may be a web server that processes services. The types of computing devices described above are only examples and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may be a server providing a cloud computing service. More specifically, the computing device 100 may be a kind of Internet-based computing and may be a server that provides a cloud computing service in which information is processed by another computer connected to the Internet rather than a user's computer. The cloud computing service may be a service that stores data on the Internet and allows users to use the data stored on the Internet anytime and anywhere through Internet access without installing necessary data or programs on their computers. Easy to share and forward with just a click. In addition, the cloud computing service not only simply stores data in a server on the Internet, but also allows users to perform desired tasks by using the functions of application programs provided on the web without installing a separate program. It may be a service that allows you to work while sharing. In addition, the cloud computing service may be implemented in the form of at least one of Infrastructure as a Service (IaaS), Platform as a Service (PaaS), Software as a Service (SaaS), virtual machine-based cloud server, and container-based cloud server. That is, the computing device 100 of the present disclosure may be implemented in the form of at least one of the aforementioned cloud computing services. The specific description of the cloud computing service described above is only an example, and may include any platform for constructing the cloud computing environment of the present disclosure.

A detailed configuration of a method for providing a professional translation service platform corresponding to an individual user in the present disclosure and effects thereof will be described later with reference to FIG. 2 .

2 illustrates a block diagram of an embodiment of the present disclosure and a computing device for providing a professional translation service platform.

As shown in FIG. 2 , the computing device 100 may include a network unit 110 , a memory 120 and a processor 130 . The components included in the aforementioned computing device 100 are examples, and the scope of the present disclosure is not limited to the aforementioned components. That is, additional components may be included or some of the aforementioned components may be omitted according to an implementation aspect of the embodiments of the present disclosure.

According to an embodiment of the present disclosure, the computing device 100 may include a user terminal 10 , a translator terminal 20 , and a network unit 110 that transmits and receives data to and from an external server 30 . The network unit 110 may transmit and receive data for performing a method for providing a professional translation service platform according to an embodiment of the present disclosure and a learning data set for learning a neural network model to other computing devices, servers, etc. there is. That is, the network unit 110 may provide a communication function between the computing device 100 and the user terminal 10 , the translator terminal 20 and the external server 30 . For example, the network unit 110 may receive translation request information including an input sentence from the user terminal 10 , retranslation request information related to a re-translation request for the translated sentence, and the like. For another example, the network unit 110 may receive a translated sentence from the translator terminal 20 . As another example, the network unit 110 may receive a training data set for training the translation model of the present disclosure from the external server 30 . Additionally, the network unit 110 may permit information transmission between the computing device 100, the user terminal 10, and the external server 30 by calling a procedure to the computing device 100.

The network unit 110 according to an embodiment of the present disclosure includes a Public Switched Telephone Network (PSTN), x Digital Subscriber Line (xDSL), Rate Adaptive DSL (RADSL), Multi Rate DSL (MDSL), and VDSL ( Various wired communication systems such as Very High SpeedDSL), Universal Asymmetric DSL (UADSL), High Bit Rate DSL (HDSL), and Local Area Network (LAN) can be used.

In addition, the network unit 110 presented in this specification is CDMA (Code Division Multi Access), TDMA (Time Division Multi Access), FDMA (Frequency Division Multi Access), OFDMA (Orthogonal Frequency Division MultiAccess), SC-FDMA (Single Carrier) -FDMA) and other systems.

In the present disclosure, the network unit 110 may be configured regardless of its communication mode, such as wired and wireless, and may be configured with various communication networks such as a personal area network (PAN) and a wide area network (WAN). there is. In addition, the network may be the known World Wide Web (WWW), or may use a wireless transmission technology used for short-range communication such as Infrared Data Association (IrDA) or Bluetooth. The techniques described herein may be used in the networks mentioned above as well as other networks.

According to an embodiment of the present disclosure, the memory 120 may store a computer program for performing a method for providing a professional translation service platform according to an embodiment of the present disclosure, and the stored computer program may be stored in the processor 130 It can be read and driven by In addition, the memory 120 may store any type of information generated or determined by the processor 130 and any type of information received by the network unit 110 . Also, the memory 120 may store information related to an input sentence and a translated sentence corresponding to the input sentence. For example, the memory 120 stores input/output data (eg, translation request information including an input sentence, translation request information corresponding to the input sentence, and retranslation request information related to a re-translation request of the translated sentence, etc.) can be stored temporarily or permanently.

According to an embodiment of the present disclosure, the memory 120 is a flash memory type, a harddisk type, a multimedia card micro type, or a card type memory (eg, SD or XD memory, etc.), RAM (Random Access Memory, RAM), SRAM (Static Random Access Memory), ROM (Read-Only Memory, ROM), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory) ), a magnetic memory, a magnetic disk, and an optical disk may include at least one type of storage medium. The computing device 100 may operate in relation to a web storage that performs a storage function of the memory 120 on the Internet. The above description of the memory is only an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the processor 130 may include one or more cores, and includes a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), It may include a processor for data analysis and deep learning, such as a tensor processing unit (TPU).

The processor 130 may read the computer program stored in the memory 120 and process data for deep learning according to an embodiment of the present disclosure. According to an embodiment of the present disclosure, the processor 130 may perform an operation for learning a neural network. The processor 130 is used for neural network learning, such as processing input data for learning in deep learning (DL), extracting features from input data, calculating errors, and updating neural network weights using backpropagation. calculations can be performed.

Also, in the processor 130, at least one of CPU, GPGPU, and TPU may process learning of a network function. For example, the CPU and GPGPU can process learning of network functions and data classification using network functions. In addition, in one embodiment of the present disclosure, the learning of a network function and data classification using a network function may be processed by using processors of a plurality of computing devices together. In addition, a computer program executed in a computing device according to an embodiment of the present disclosure may be a CPU, GPGPU or TPU executable program.

In this specification, network functions may be used interchangeably with artificial neural networks and neural networks. In this specification, a network function may include one or more neural networks, and in this case, an output of the network function may be an ensemble of outputs of one or more neural networks.

The processor 130 may read the computer program stored in the memory 120 and provide one or more translation models according to an embodiment of the present disclosure. According to an embodiment of the present disclosure, the processor 130 may perform a calculation to obtain a translated sentence corresponding to an input sentence. According to an embodiment of the present disclosure, the processor 130 may perform calculations to train one or more translation models.

According to one embodiment of the present disclosure, the processor 130 may typically process overall operations of the computing device 100 . The processor 130 may provide or process appropriate information or functions to the user terminal by processing signals, data, information, etc. input or output through the components described above or by driving an application program stored in the memory 120. there is.

According to an embodiment of the present disclosure, the processor 130 may generate one or more translation models corresponding to one or more domains. Each of the one or more translation models in the present disclosure may be, for example, a model that performs Neural Machine Translation (NMT) based on an artificial neural network. Each of the one or more translation models may convert an input sentence into a sentence vector form and generate a sentence of an output language based on the converted sentence vector form. In this case, a part that converts an input sentence into a sentence vector may be an encoder, and a part that generates an output sentence may be a decoder. The language translation model can automatically learn numerous parameter values constituting an artificial neural network of an encoder and a decoder through training data composed of learning input data and learning result data.

For another example, each of the one or more translation models may be a neural network model that performs corpus-based machine translation (CMT). Corpus-based machine translation may include Example-Based Machine Translation (EBMT) and Statistical Machine Translation (SMT). Example-based machine translation may be a machine translation method in which information of a translated input sentence and a translated sentence is stored, and a translation result is output using the previously stored information in response to a request for translation of the same sentence. Statistical-based machine translation may be a type of machine translation that analyzes the frequency of linguistic correlation between the original text and the translated text, statisticizes it, and uses it to calculate the translated text.

That is, each of the one or more translation models may be a neural network model trained to generate a translated sentence by taking an input sentence as an input through a plurality of training data composed of a plurality of learning input data and a plurality of learning output data.

For generating one or more translation models, the processor 130 may obtain a training data set including a plurality of training data. Specifically, the processor 130 may acquire a learning data set through the network unit 110 . Acquisition of a learning data set according to an embodiment of the present disclosure may be receiving or loading a learning data set stored in the memory 120 . For example, acquiring a learning data set may be receiving a learning data set stored in the external server 30 . Reception of learning data may be receiving or loading original text data from another storage medium, another computing device, or a separate processing module within the same computing device, based on a wired/wireless communication means.

In detail, the processor 130 may construct one or more sub-data of training data by classifying a plurality of training data based on one or more domains. One or more domains are criteria for classifying a plurality of learning data, and may be related to various specialized fields. For example, one or more domains may relate to various specialized fields such as medical field, legal field, IT field, or economic field.

The plurality of learning data may include a plurality of learning input data composed of the original text and a plurality of learning output data composed of translations of the original text. For example, the first learning input data may be a sentence composed of Korean, and the first learning output data may be a sentence composed of English, which is an English translation result of the corresponding sentence. In this case, the first learning input data and the first learning output data may be labeled learning data. That is, in neural network learning, the first learning input data may be data related to the input of the neural network. The first learning output data may be data compared with an output of the neural network for the first learning input data.

The processor 130 identifies a domain of each of the plurality of learning data based on at least one of the semantic analysis result of each of the plurality of learning data, the repetition count for each word included in the semantic analysis result of each of the plurality of training data, and a predetermined subject. can do.

Specifically, the processor 130 may determine one or more domains based on a semantic analysis result for each of a plurality of learning data. For example, if each of the words included in the semantic analysis result is 'copyright', 'in copyright', 'exclusive', 'exclusive', 'guaranteeing rights', 'is a system', the processor (130 ) can determine 'copyright' or 'rights' as a domain by determining that the meaning of 'copyright' or 'rights' is relatively important. The detailed description of one or more words included in the above semantic analysis result and the importance of each of the one or more words is only an example, and the present disclosure is not limited thereto.

In addition, the processor 130 may determine one or more domains for the conversation transcript based on the number of repetitions for each word included in the semantic analysis result for each learning input data. The processor 130 may identify repetition counts of each of one or more words included in the semantic analysis result, and determine a word having a high repetition count as a domain for each learning input data.

Also, the processor 130 may determine a domain based on a predetermined subject. The processor 130 may determine a subject of the learning input data through a semantic analysis result, and may determine one or more domains corresponding to the determined subject. In this case, each of a plurality of domains corresponding to each of a plurality of subjects may be previously stored in the memory 120 . For example, domains related to the field of law may be matched and stored in the memory 120 in advance in response to a subject related to 'copyright'. The foregoing subject and specific description of the domain corresponding to the subject are only examples, and the present disclosure is not limited thereto.

The processor 130 may build one or more subsets of training data based on each domain of the plurality of training data. In addition, the processor 130 may generate one or more translation models by performing training on each of one or more neural network models through each of one or more training data subsets.

For a specific example, a training data set related to learning of a neural network may include a plurality of training data related to various domains. In this case, the computing device 100 may generate one or more subsets of learning data by classifying a plurality of pieces of learning data based on each of one or more domains. For example, one or more training data subsets may be a combination of training data subsets, and each training data subset may be a combination of a plurality of training data based on different domains. For example, one or more training data subsets built by the processor 130 may include a first training data subset and a second training data subset. In this case, the first learning data subset may include a plurality of learning data related to the medical field, and the second learning data subset may include a plurality of learning data related to the legal field. The specific description of one or more specialized fields described above is only an example, and the present disclosure is not limited thereto.

More specifically, each learning data subset may include learning input sub-data composed of a starting language and second learning output sub-data composed of a target language with respect to each field. In this case, the processor 130 may generate a plurality of translation models by performing learning on each neural network model through each training data subset.

As a specific example, one or more subsets of learning data may include a first subset of learning data related to the field of law, and the first subset of learning data includes first learning input sub-data composed of Korean and English. It may include first learning output sub data. The processor 130 may output first learning result data by inputting the first learning input sub-data included in the first training data subset to the first neural network model. The processor 130 may derive an error between the output first learning result data and the first learning output data labeled in the first learning input sub-data. In addition, the processor 130 may perform learning on the first neural network model by backpropagating the corresponding error and adjusting the weight of the neural network so that the first learning result data and the first learning output sub-data are similar. As the learning of the first neural network model as described above is completed, the first translation model may be generated. That is, the first translation model may output a translated sentence composed of English by inputting an input sentence composed of Korean in relation to the field of law through the above-described learning. In this case, since the learning data related to the learning of the first translation model is data related to the legal field by the processor 130, the first translation model can output a more suitable translation sentence in response to an input sentence in the corresponding legal field. there is.

That is, the processor 130 classifies a plurality of learning data related to learning of the neural network based on one or more domains related to various specialized fields to generate learning data subsets, and utilizes the generated learning data subsets to classify each of the various fields. One or more translation models that perform specialized translations can be created. Accordingly, it is possible to provide a suitable translation model according to individual needs of each of a plurality of users. Accordingly, the user can obtain more specialized translation-related information in response to an input sentence to be translated through a translation model suitable for the user's needs.

According to an embodiment of the present disclosure, the processor 130 may determine an optimal translation model corresponding to a user terminal from among one or more translation models. Specifically, when subscription request information is received from the user terminal, the processor 130 may determine an optimal translation model corresponding to the corresponding user terminal 10 from among one or more translation models based on the received subscription request information. The optimal translation model may be a translation model for outputting more appropriate translation information in a field related to the user terminal 10 . The subscription request information may be information serving as a criterion for determining an optimal translation model corresponding to a user terminal. The subscription request information may include at least one of user language identification information and user domain identification information. The user language identification information may include information about a start language desired for translation by the user of the user terminal 10 and a destination language corresponding to a translation result. For example, the user language identification information may include information indicating that the input sentence to be translated by the user of the first user terminal is Japanese (ie, the starting language) and the translated sentence to be obtained as a translation result is English (ie, the target language). can The specific description of the language related to the above-described input sentence and translated sentence is only an example, and the present disclosure is not limited thereto. The user domain identification information may be information for identifying which field an input sentence to be translated by the user of the user terminal 10 is related to. For example, the user domain identification information may include information indicating that an input sentence to be translated by the user of the first user terminal is related to the IT field. For another example, the user domain identification information may include information indicating that an input sentence to be translated by the user of the second user terminal is related to the legal field. The detailed description of the aforementioned user language identification information and domain identification information is only an example, and the present disclosure is not limited thereto.

That is, the processor 130 may determine an optimal translation model corresponding to the corresponding user terminal from among one or more translation models generated in correspondence with various professional fields based on the user language identification information and the user domain identification information included in the subscription request information. there is. For example, the subscription request information received from the first user terminal 10 is user language identification information indicating that the input sentence the user wants to translate is Korean and the translated sentence to be obtained as a result of the translation is English, and the input the user wants to translate When the sentence includes user domain identification information related to the legal field, the processor 130 performs learning including learning input sub-data composed of Korean and learning output sub-data composed of English related to the legal field among one or more translation models. A translation model generated through learning using a data subset may be determined as an optimal translation model corresponding to the first user terminal 10 .

In a further embodiment, subscription request information received from the user terminal 10 may include a customized learning data set. Here, the customized learning data set is information related to the unique translation style of the user of the user terminal 10 and may be used as learning data for generating an optimal translation model more suitable for the user. For example, the customized learning data set may include information related to a translation job performed by a corresponding user at a time in the past. That is, the customized learning data set may be learning data in which a specific user's translation style is reflected. Accordingly, the processor 130 may identify whether or not the customized learning data set is included in the subscription request information received from the user terminal 10 . In addition, when it is identified that the subscription request information includes a custom training data set corresponding to the corresponding user, the processor 130 trains a neural network model composed of one or more network functions through the custom training data set to obtain an optimal translation model. can create In other words, when the processor 130 receives a customized learning data set, which is a set of information about the user's own translation style, from the user terminal 10, the processor 130 may generate an optimal translation model through the customized learning data set. there is. In this case, since the optimal translation model generated by the processor 130 is a neural network model learned through learning data related to the user's existing translation style (ie, a customized learning data set), it corresponds to the user's needs in response to the input sentence. It is possible to output a translated sentence that more closely matches the .

That is, the processor 130 may generate a plurality of translation models in relation to various domains, and select and provide an optimal translation model expected to be more suitable for the user terminal 10 from among the generated plurality of translation models. . In addition, when the processor 130 receives a custom training data set in which the user's unique translation characteristics are reflected from the user terminal 10, the processor 130 may generate and provide an optimal translation model specialized for the user through the corresponding training data. there is.

According to an embodiment of the present disclosure, the processor 130 may receive translation request information including an input sentence from a user terminal. The input sentence may mean a sentence composed of a starting language in which the user of the user terminal 10 intends to perform translation.

When translation request information is received from the user terminal 10, the processor 130 may generate a translated sentence corresponding to the translation request information by utilizing an optimal translation model. In this case, the optimal translation model, as described above, is determined based on the user language identification information and the user domain identification information corresponding to the user terminal 10, and corresponds to the input sentence received from the user terminal 10. It may be a translation model that generates more suitable translation sentences. Specifically, the processor 130 may identify an input sentence from the received translation request information, process the identified input sentence as an input of an optimal translation model, and generate a translated sentence corresponding to the input sentence.

According to an embodiment of the present disclosure, the processor 130 may calculate the translation reliability of the translated sentence. More specifically, the processor 130 may generate a back-translated sentence by performing backtranslation on a translated sentence generated (or translated) corresponding to an input sentence. In this case, the reverse translation performed by the processor 130 may mean translating the translated sentence into a language corresponding to the input sentence. The processor 130 may calculate the translation reliability of the translated sentence based on the comparison between the reverse translated sentence and the input sentence.

For example, when it is desired to translate an input sentence composed of Korean into a translated sentence composed of English, the processor 130 may generate a translated sentence by processing the input sentence as an input of a translation model. In this case, the processor 130 may calculate translation reliability corresponding to the translated sentence in order to determine the translation accuracy of the translated sentence. The processor 130 processes the translated sentence written in English as an input of another translation model (eg, a translation model that is related to the same domain and outputs a translated sentence composed of Korean by taking an input sentence composed of English as an input) It is possible to generate a reverse translation sentence composed of Korean.

Also, the processor 130 may calculate translation reliability based on a comparison between an input sentence composed of the same Korean language and a reverse-translated sentence. For example, as a result of comparing the first input sentence and the first reverse-translated sentence corresponding to the first input sentence, when it is determined that the structure or meaning of each sentence is similar, the processor 130 determines the first translation sentence for the first translated sentence. Higher translation reliability can be calculated by determining that the translation accuracy is high. For another example, as a result of comparing the second input sentence and the second reverse-translated sentence corresponding to the second input sentence, when it is determined that the structure or meaning of each sentence is not similar, the processor 130 performs the second translation. A lower translation reliability may be calculated by determining that the translation accuracy of the sentence is low. The specific descriptions of the above-described input sentence, translated sentence, and reverse-translated sentence are only examples, and the present disclosure is not limited thereto.

That is, the processor 130 may calculate the translation reliability of the translated sentence by performing reverse translation on the translated sentence generated corresponding to the input sentence. The translation reliability corresponding to such a translated sentence indicates the accuracy of the translation, and can be used as information for recognizing the accuracy of the translation for the user in the inspection process.

According to an embodiment of the present disclosure, the processor 130 may receive translation participation information from each of a plurality of translator terminals. The translation participation information may be generated in relation to a point in time at which translator terminals intend to participate in translation, and may be information for identifying each of a plurality of translator terminals. In addition, the translation participation information may include translator language identification information and translator domain identification information. The translator language identification information may be information for identifying a language that can be utilized by a user of each translator terminal (ie, a language that can perform translation). The translator domain identification information may be information related to one or more fields in which each translator terminal can perform translation professionally. For example, the translator language identification information of the first translator terminal may include information that the first user (or the first translator) can use Korean, English, and Japanese, and the translator domain identification information may include information in the IT field. It may include information that the related translation can be professionally performed. The detailed description of the language identification information and domain identification information of the first translator terminal described above is only an example, and the present disclosure is not limited thereto.

Also, the processor 130 may build a translator database based on translation participation information of each of a plurality of translator terminals. When translation participation information is received from each of a plurality of translator terminals, the processor 130 may build a translator database based on the corresponding translation participation information. Such a translator database may be referred to by the computing device 100 when the user terminal 10 transmits translation request information corresponding to a specific input sentence to the computing device 100 .

According to an embodiment of the present disclosure, the processor 130 may receive replay* request information from the user terminal. The re-translation request information is information about a re-translation request for a translated sentence, and may be information generated based on the translation request information. Accordingly, the retranslation request information may include an input sentence corresponding to the translation request information and domain identification information. Specifically, the retranslation request information may include an input sentence to be translated, user language identification information included in the translation request information, and user domain identification information.

When retranslation request information is received from the user terminal 10, the processor 130 may determine one or more requesting translator terminals to request retranslation based on the received retranslation request information. Specifically, the processor 130 may identify one or more translator terminals corresponding to the retranslation request information by referring to the translator database. For example, the processor 130 may determine one or more candidate translator terminals corresponding to user language identification information and user domain identification information included in retranslation request information from among a plurality of translator terminals included in the translator database.

As a specific example, the retranslation request information of the user terminal 10 indicates that the input sentence to be translated by the user is Korean, and the user language identification information indicating that the translated sentence to be obtained as a result of the translation is English, and the domain that the input sentence is related to the legal field. In the case of including the identification information, the processor 130 refers to a plurality of translator databases including translator language identification information and translator domain identification information of each of a plurality of translator terminals, and the corresponding user language identification information and user domain identification information, respectively. One or more candidate translator terminals may be identified. In this case, one or more candidate translator terminals identified by the processor 130 may be terminals related to translators who are fluent in Korean and English and who specialize in translation in the field of law.

Also, the processor 130 may determine one or more requesting translator terminals based on the translation evaluation scores of the identified one or more candidate translator terminals. The translation evaluation score may be an evaluation score for translation performed by each translator terminal corresponding to a previous point in time. For example, the translation evaluation score may be a translation evaluation score received from a user terminal or other translator terminals. For example, if the first translator terminal has participated in generating translated sentences three times at this point in time, and the translation evaluation sub-scores corresponding to each round are 5, 7, and 9, the processor 130 may configure each of the first translator terminals The translation evaluation score of the first translator terminal may be calculated as 7 through the average of the translation evaluation sub scores per session. The specific numerical description of the number of previous translation participations, translation evaluation sub-scores, and translation evaluation scores in the translator terminal is only an example, and the present disclosure is not limited thereto.

Specifically, the processor 130 may determine the top three candidate translator terminals having the highest translation evaluation scores among the translation evaluation scores of each of the one or more candidate translator terminals as one or more requesting translator terminals. For another example, the processor 130 may determine one or more requesting translator terminals based on whether the translation evaluation scores of each of the one or more candidate translator terminals exceed a predetermined threshold evaluation score. The detailed description of the method for determining the requested translator terminal described above is only an example, and the present disclosure is not limited thereto.

That is, the processor 130 may transmit retranslation request information to verified translators by determining one or more requesting translator terminals based on the translation evaluation scores of each candidate translator terminal among one or more candidate translator terminals. In other words, a virtuous cycle system in which the quality of service is improved can be formed by first providing a service opportunity to a translator terminal that has been highly evaluated by the user.

According to another embodiment of the present disclosure, the processor 130 may determine to generate and transmit candidate translator list information including one or more candidate translator terminals to the user terminal. The candidate translator list information may be information listing translator terminals related to one or more translators capable of performing the translation desired by the user of the user terminal.

Specifically, the processor 130 may determine to transmit a translation service participation request to one or more candidate translator terminals. In this case, each of the one or more candidate translator terminals may transmit a participation signal in response to a participation request regarding whether or not to perform a translation service for an input sentence. As the participation response transmission process exists, the translator terminal can select whether to perform a service related to the translation service participation request. Accordingly, the translator terminal does not respond to a translation service in which there is not much performance experience, thereby avoiding assignment of the translation service.

Correspondingly, the processor 130 may provide an incentive equal to the number of times the translation service is performed, so that the translator terminal may increase the participation response transmission rate for the service. Through the above process, since the translator will immediately respond to the service request with experience and confidence, the quality and efficiency of the service can be improved compared to the system that unconditionally assigns the translation service to the translator. In addition, the processor 130 may receive a participation signal corresponding to a translation service participation request from each of one or more candidate translator terminals, and may determine to generate and transmit translation user list information to the user terminal based on the received participation signal. .

Also, the processor 130 may determine a translator terminal to perform a translation service desired by the user based on a response of the user terminal to the translation user list information.

According to another embodiment of the present disclosure, when receiving retranslation request information from the user terminal 10, the processor 130 generates translator terminal list information based on service history information of the user terminal 10, and It may be determined to transmit the created translator terminal list information to the user terminal 10 . In this case, the service history information may include information about one or more translation terminals that performed translation services in relation to the user terminal 10 at a previous time.

According to an additional embodiment, when receiving retranslation request information from the user terminal 10, the processor 130 identifies candidate translator terminals that have been highly evaluated by users, generates translator terminal list information, and generates translator terminal list information. It may be determined to transmit list information to the user terminal 10 .

In a further embodiment, the processor 130 may determine whether to transmit retranslation request information to one or more requesting translator terminals based on the translation reliability of the translated sentence. The translation reliability of the translated sentence indicates the accuracy of the translation related to the output of the translation model, and can be used as information for recognizing the accuracy of the translation of the translated sentence to the user in the inspection process.

Specifically, the processor 130 may determine to transmit retranslation request information to one or more commissioned translator terminals when the translation reliability of the translated sentence is less than a predetermined threshold reliability. In addition, the processor 130 may determine not to transmit retranslation request information to one or more requested translator terminals when the translation reliability of the translated sentence is equal to or greater than a predetermined threshold reliability.

That is, the processor 130 determines whether the translated sentence has been translated at a reliable level based on the translation reliability of the translated sentence, and the input sentence of the user terminal is automatically transmitted to one or more requesting translator terminals (ie, requesting a translation). decide whether or not to do so). In other words, without a user's separate inspection process for the translated sentence related to the output of the optimal translation model, linkage with the terminal of the requesting translator can be automatically performed based on the calculated translation reliability. Accordingly, a separate inspection process for translated sentences is omitted, and user convenience can be improved.

According to an embodiment of the present disclosure, the processor 130 may determine to transmit retranslation request information to one or more requested translator terminals. Here, the retranslation request information may include an input sentence, and the input sentence may mean a sentence that the user of the user terminal 10 wants to translate. Specifically, the processor 130 may determine to transmit retranslation request information corresponding to the translation request information received from the user terminal 10 to one or more commissioned translator terminals determined through the translation evaluation information. Accordingly, one or more commissioned translator terminals may generate retranslated sentences corresponding to the input sentences. In addition, the retranslated sentences generated through one or more commissioned translator terminals may be transmitted to the processor 130 .

In an additional embodiment, the processor 130 may determine to transmit the translated sentence to one or more requesting translator terminals based on the translation reliability of the translated sentence output through the optimal translation model. Specifically, the processor 130 may determine to transmit the translated sentence to one or more requested translator terminals when the translation reliability of the translated sentence corresponding to the input sentence is equal to or greater than a predetermined threshold reliability. For example, translation reliability may be information indicating the adequacy of translation for a translated sentence. Accordingly, the processor 130 may determine to transmit the translated sentence to one or more requesting translator terminals when it is determined that the translated sentence has adequacy equal to or higher than a predetermined standard through the translation reliability of the translated sentence. In this case, one or more requested translator terminals may be translator terminals that receive the input sentences in order to generate translated sentences corresponding to the input sentences. That is, when the translation reliability of the translated sentence related to the output of the translation model is equal to or greater than a predetermined reference value (ie, greater than or equal to a predetermined threshold reliability), the corresponding translated sentence may be transmitted to each of one or more requesting translator terminals together with the input sentence.

In this case, the requesting translator terminal may simultaneously receive an input sentence and a translated sentence output by an optimal translation model corresponding to the input sentence. Accordingly, the user (or translator) of the requested translator terminal may refer to the translated sentence in the process of generating a retranslated sentence corresponding to the input sentence. The corresponding translated sentence is a translated sentence having a translation reliability higher than a certain standard, and may be a sentence with relatively few corrections or corrections. Therefore, by providing a translated sentence that guarantees a certain level or higher reliability in the process of generating a retranslated sentence for an input sentence, the efficiency of the translator's translation work can be improved.

According to an embodiment of the present disclosure, when one or more retranslated sentences are received from one or more requesting translator terminals, the processor 130 may determine to transmit a first retranslated sentence among the received one or more retranslated sentences to the user terminal. . Specifically, the processor 130 may determine a first retranslated sentence among one or more retranslated sentences based on the translation evaluation scores of each of one or more requesting translator terminals. The first retranslated sentence may refer to a sentence having the most appropriate expression or structure among the retranslated sentences received by the requesting translator terminals. For example, the processor 130 may determine a retranslated sentence received from a requesting translator terminal having the highest translation evaluation score among one or more retranslated sentences as a first retranslated sentence.

According to another embodiment of the present disclosure, when one or more retranslated sentences are received from one or more requested translator terminals, the processor 130 may determine to transmit the received one or more retranslated sentences to the user terminal. Specifically, the processor 130 may determine to generate and transmit candidate re-translation list information including one or more re-translation sentences to the user terminal. The candidate retranslation list information may be information listing one or more retranslation sentences received from each of one or more requesting translator terminals.

Also, the processor 130 may receive selection information about at least one retranslated sentence from among one or more retranslated sentences from the user terminal 10 . Specifically, the processor 130 may transmit candidate re-translation list information including one or more re-translation list information to the user terminal 10, and the user terminal 10 may transmit at least one of the one or more re-translation list information included in the candidate re-translation list information. A user's selection input for one sentence may be received. In this case, the user terminal 10 may generate selection information for a specific retranslated sentence in response to a user's selection input and transmit it to the processor 130 . Accordingly, the processor 130 may identify a first retranslated sentence among one or more retranslated sentences based on the selection information received from the user terminal 10 .

That is, the processor 130 may determine a retranslated sentence related to the user's selection among one or more retranslated sentences from one or more requesting translator terminals as the first retranslated sentence. In other words, the processor 130 may present various retranslated sentences generated in response to the input sentence so that the user of the user terminal 10 can determine his/her desired translated sentence according to his/her preference.

According to an embodiment of the present disclosure, the processor 130 may perform augmentation on training data based on the input sentence and the first retranslated sentence. Specifically, the processor 130 may generate additional training data by matching and labeling the input sentence and the first retranslated sentence. In addition, the processor 130 may perform additional learning on the optimal translation model through additional training data to update the optimal translation model. In this case, the optimal translation model is a translation model determined based on the subscription request information of the user terminal 10, and may be a translation model that outputs a translation sentence more appropriate in the field related to the input sentence to be translated by the user. . In addition, the updated optimal translation model is an optimal translation model additionally trained to output a first retranslated sentence generated by a professional translator in response to the input sentence related to the translation request of the user terminal 10 as an input can

That is, the processor 130 augments learning data through each of various input sentences received from a plurality of user terminals and various translated sentences of the corresponding translator terminals 20, and builds a translation model through the augmented learning data. can be updated Accordingly, the translation model can be advanced over time by the above-described steps of the processor 130 . This may cause an effect of improving reliability and accuracy of the translated sentence in relation to the output of the translation model. In addition, a translation model corresponding to each user terminal is provided, and as the corresponding translation model is continuously updated in relation to each input sentence of the user terminal 10, it can be optimized to a translation model more suitable for each individual user. .

4 shows an exemplary flow chart of a method for providing a professional translation services platform related to one embodiment of the present disclosure.

According to an embodiment of the present disclosure, the user terminal 10 may transmit subscription request information to a computing device (301). The subscription request information may include at least one of user language identification information and user domain identification information. The user language identification information may include information about a start language desired for translation by the user of the user terminal 10 and a target language corresponding to a translation result.

According to an embodiment of the present disclosure, the computing device 100 may determine an optimal translation model corresponding to the user terminal 10 (303). Specifically, when subscription request information is received from the user terminal, the computing device 100 may determine an optimal translation model corresponding to the corresponding user terminal 10 from among one or more translation models based on the received subscription request information. The optimal translation model may be a translation model for outputting more appropriate translation information in a field related to the user terminal 10 . The one or more translation models may refer to each translation model performing specialized translation corresponding to each of the one or more domains.

According to an embodiment of the present disclosure, the user terminal 10 may transmit translation request information to the computing device 100 (305). In this case, the translation request information is composed of a starting language for which the user intends to translate. Can contain input text.

According to an embodiment of the present disclosure, the computing device 100 may generate a translated sentence corresponding to the translation request information by utilizing an optimal translation model and transmit the generated translated sentence to the user terminal 10 (307). . In this case, the optimal translation model is determined based on the user language identification information and the user domain identification information corresponding to the user terminal 10, and provides a more appropriate translation sentence in response to the input sentence received from the user terminal 10. It may be a translation model to generate. Specifically, the computing device 100 may identify an input sentence from the received translation request information, process the identified input sentence as an input of an optimal translation model, and generate a translation sentence corresponding to the input sentence.

According to an embodiment of the present disclosure, the translator terminal 20 may transmit translation participation information to the computing device 100 (309). The translation participation information is generated in relation to the time when the translator terminals intend to participate in translation. and may be information for identifying each of a plurality of translator terminals. Also, translation participation information may include translator language identification information and translator domain identification information. The translator language identification information may be information for identifying a language that can be utilized by a user of each translator terminal (ie, a language that can perform translation). The translator domain identification information may be information related to one or more fields in which each translator terminal can perform translation professionally.

According to one embodiment of the present disclosure, the computing device 100 may build a translator database (311). When translation participation information is received from each of a plurality of translator terminals, the computing device 100 may build a translator database based on the translation participation information. Such a translator database may be referred to by the computing device 100 when the user terminal 10 transmits translation request information corresponding to a specific input sentence to the computing device 100 .

According to an embodiment of the present disclosure, the user terminal 10 may transmit retranslation request information to the computing device 100 (313). The re-translation request information is information about a re-translation request for a translated sentence, and may be information generated based on the translation request information. Accordingly, the retranslation request information may include an input sentence corresponding to the translation request information and domain identification information. For example, the user may apply an input to the user terminal 10 to generate retranslation request information when it is determined that the meaning or expression of the translated sentence is not appropriate in the process of inspecting the translated sentence output through the optimal translation model. can

Specifically, the retranslation request information may include an input sentence to be translated, user language identification information included in the translation request information, and user domain identification information. When retranslation request information is received from the user terminal 10, the computing device 100 may determine one or more requesting translator terminals to request retranslation based on the received retranslation request information. Specifically, the computing device 100 refers to the translator database and corresponds to the retranslation request information.

One or more translator terminals may be identified. For example, the computing device 100 may determine one or more candidate translator terminals corresponding to user language identification information and user domain identification information included in retranslation request information from among a plurality of translator terminals included in the translator database.

According to one embodiment of the present disclosure, the translator terminal 20 may transmit one or more retranslated sentences to the computing device 100 (315). When one or more retranslated sentences are received from one or more requesting translator terminals, the computing device 100 may determine to transmit a first retranslated sentence among the received one or more retranslated sentences to the user terminal. Specifically, the computing device 100 may determine a first retranslated sentence among one or more retranslated sentences based on the translation evaluation scores of each of one or more requesting translator terminals. The first retranslated sentence may refer to a sentence having the most appropriate expression or structure among the retranslated sentences received by the requesting translator terminals. For example, the computing device 100 may determine a retranslated sentence received from a requesting translator terminal having the highest translation evaluation score among one or more retranslated sentences as a first retranslated sentence.

According to an embodiment of the present disclosure, the computing device 100 may transmit the first retranslated sentence to the user terminal 10 (317).

According to an embodiment of the present disclosure, the computing device 100 may update the optimal translation model through the input sentence and the first translation sentence (319). The computing device 100 may generate additional learning data by matching and labeling the input sentence and the first retranslated sentence. In addition, the computing device 100 may perform additional learning on the optimal translation model through additional training data to update the optimal translation model. In this case, the optimal translation model is a translation model determined based on the subscription request information of the user terminal 10, and may be a translation model that outputs a translation sentence more appropriate in the field related to the input sentence to be translated by the user. . In addition, the updated optimal translation model is an optimal translation model additionally trained to output a first retranslated sentence generated by a professional translator in response to the input sentence related to the translation request of the user terminal 10 as an input. can

That is, the computing device 100 augments learning data through each of various input sentences received from a plurality of user terminals and various translated sentences of the corresponding translator terminals 20, and the translation model through the augmented learning data. can be updated.

5 is a flowchart illustratively illustrating steps for performing a method for providing a professional translation service platform related to an embodiment of the present disclosure.

According to one embodiment of the present disclosure, the computing device 100 may generate one or more translation models corresponding to one or more domains (410).

According to an embodiment of the present disclosure, when subscription request information is received from a user terminal, the computing device 100 may determine an optimal translation model corresponding to the user terminal from among one or more translation models based on the received subscription request information. Yes (420).

According to an embodiment of the present disclosure, when translation request information is received from a user terminal, the computing device 100 may generate a translated sentence corresponding to the received input sentence by utilizing the first translation model (430). .

According to an embodiment of the present disclosure, when retranslation request information related to a translated sentence is received from a user terminal, the computing device 100 may determine one or more requesting translator terminals to request retranslation based on the received retranslation request information. Yes (440).

According to an embodiment of the present disclosure, the computing device 100 may determine to transmit retranslation request information to one or more commissioned translator terminals (450).

According to an embodiment of the present disclosure, when one or more retranslated sentences are received from one or more requesting translator terminals, the computing device 100 may determine a first retranslated sentence among the one or more received retranslated sentences as a user terminal (460). ).

The order of the steps shown in FIG. 5 described above may be changed as needed, and at least one or more steps may be omitted or added. That is, the foregoing steps are merely one embodiment of the present disclosure, and the scope of rights of the present disclosure is not limited thereto.

6 is a schematic diagram illustrating one or more network functions related to one embodiment of the present disclosure.

Throughout this specification, computational model, neural network, network function, and neural network may be used interchangeably. A neural network may consist of a set of interconnected computational units, which may be generally referred to as “nodes”. These “nodes” may also be referred to as “neurons”. A neural network includes one or more nodes. Nodes (or neurons) constituting neural networks may be interconnected by one or more “links”.

In a neural network, one or more nodes connected through a link may form a relative relationship of an input node and an output node. The concept of an input node and an output node is relative, and any node that has an output node relationship to one node may have an input node relationship to another node, and vice versa. As described above, an input node to output node relationship may be created around a link. More than one output node can be connected to one input node through a link, and vice versa.

In a relationship between an input node and an output node connected through one link, the value of the output node may be determined based on data input to the input node. Here, a node interconnecting an input node and an output node may have a weight. The weight may be variable, and may be changed by a user or an algorithm in order to perform a function desired by the neural network. For example, when one or more input nodes are interconnected by respective links to one output node, the output node is set to a link corresponding to values input to input nodes connected to the output node and respective input nodes. An output node value may be determined based on the weight.

As described above, in the neural network, one or more nodes are interconnected through one or more links to form an input node and output node relationship in the neural network. Characteristics of the neural network may be determined according to the number of nodes and links in the neural network, an association between the nodes and links, and a weight value assigned to each link. For example, if there are two neural networks having the same number of nodes and links and different weight values between the links, the two neural networks may be recognized as different from each other.

A neural network may include one or more nodes. Some of the nodes constituting the neural network may configure one layer based on distances from the first input node, for example, a set of nodes having a distance n from the first input node, n layers can be configured. The distance from the first input node may be defined by the minimum number of links that must be passed through to reach the corresponding node from the first input node. However, the definition of such a layer is arbitrary for explanation, and the order of a layer in a neural network may be defined in a method different from the above. For example, a layer of nodes may be defined by a distance from a final output node.

An initial input node may refer to one or more nodes to which data is directly input without going through a link in relation to other nodes among nodes in the neural network. Alternatively, in a relationship between nodes based on a link in a neural network, it may mean nodes that do not have other input nodes connected by a link. Similarly, the final output node may mean one or more nodes that do not have an output node in relation to other nodes among nodes in the neural network. Also, the hidden node may refer to nodes constituting the neural network other than the first input node and the last output node. In the neural network according to an embodiment of the present disclosure, the number of nodes of the input layer may be the same as the number of nodes of the output layer, and the number of nodes decreases and then increases again as the number of nodes progresses from the input layer to the hidden layer. can In addition, the neural network according to another embodiment of the present disclosure may be a neural network in which the number of nodes of the input layer may be less than the number of nodes of the output layer and the number of nodes decreases as the number of nodes increases from the input layer to the hidden layer. there is. In addition, the neural network according to another embodiment of the present disclosure is a neural network in which the number of nodes in the input layer may be greater than the number of nodes in the output layer, and the number of nodes increases as the number of nodes increases from the input layer to the hidden layer. can A neural network according to another embodiment of the present disclosure may be a neural network in the form of a combination of the aforementioned neural networks.

A deep neural network (DNN) may refer to a neural network including a plurality of hidden layers in addition to an input layer and an output layer. Deep neural networks allow you to identify latent structures in data. In other words, it can identify the latent structure of a photo, text, video, sound, or music (e.g., what objects are in the photo, what the content and emotion of the text are, what the content and emotion of the audio are, etc.). . Deep neural networks include convolutional neural networks (CNNs), recurrent neural networks (RNNs), auto encoders, generative adversarial networks (GANs), and restricted boltzmann machines (RBMs). machine), deep belief network (DBN), Q network, U network, Siamese network, and the like. The description of the deep neural network described above is only an example, and the present disclosure is not limited thereto.

The neural network may be trained using at least one of supervised learning, unsupervised learning, and semisupervised learning. The learning of neural networks is to minimize errors in the output. In learning the neural network, iteratively inputs training data to the neural network, calculates the output of the neural network and the error of the target for the training data, and transfers the error of the neural network from the output layer of the neural network to the input layer in the direction to reduce the error. It is a process of updating the weight of each node of the neural network by backpropagating in the same direction. In the case of teacher learning, each learning data is labeled with a correct answer (ie, labeled learning data), and in the case of comparative teacher learning, each learning data may not be labeled as a correct answer. That is, for example, learning data in the case of teacher learning regarding data classification may be data in which each learning data is labeled with a category. Labeled training data is input to the neural network, and an error may be calculated by comparing an output (category) of the neural network and a label of the training data. As another example, in the case of comparative history learning for data classification, an error may be calculated by comparing input learning data with a neural network output. The calculated error is back-propagated in a reverse direction (ie, from the output layer to the input layer) in the neural network, and the connection weight of each node of each layer of the neural network may be updated according to the back-propagation. The amount of change in the connection weight of each updated node may be determined according to a learning rate. The neural network's calculation of input data and backpropagation of errors can constitute a learning cycle (epoch). The learning rate may be applied differently according to the number of iterations of the learning cycle of the neural network. For example, a high learning rate may be used in the early stages of neural network training to increase efficiency by allowing the neural network to quickly achieve a certain level of performance, and a low learning rate may be used in the later stage to increase accuracy.

In neural network learning, generally, training data can be a subset of real data (ie, data to be processed using the trained neural network), and therefore, errors for training data are reduced, but errors for real data are reduced. There may be incremental learning cycles. Overfitting is a phenomenon in which errors for real data increase due to excessive learning on training data. For example, a phenomenon in which a neural network that learned a cat by showing a yellow cat does not recognize that it is a cat when it sees a cat other than yellow can be a type of overfitting. Overfitting can act as a cause of increasing error in machine learning algorithms. can Various optimization methods can be used to prevent such overfitting. In order to prevent overfitting, methods such as increasing training data, regularization, and omitting some nodes of a network in the process of learning may be applied.

Steps of a method or algorithm described in connection with an embodiment of the present disclosure may be implemented directly in hardware, implemented in a software module executed by hardware, or a combination thereof. A software module may include RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), Flash Memory, hard disk, removable disk, CD-ROM, or It may reside in any form of computer readable recording medium well known in the art to which this disclosure pertains.

Components of the present disclosure may be implemented as a program (or application) to be executed in combination with a computer, which is hardware, and stored in a medium. Components of the present disclosure may be implemented as software programming or software elements, and similarly, embodiments may include various algorithms implemented as data structures, processes, routines, or combinations of other programming constructs, such as C, C++ , Java (Java), can be implemented in a programming or scripting language such as assembler (assembler). Functional aspects may be implemented as an algorithm running on one or more processors.

Those skilled in the art will understand that the various illustrative logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein are electronic hardware, (for convenience) For purposes of this document, it will be understood that various forms of program or design code (referred to herein as "software") or a combination of both may be implemented. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the particular application and the design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

Various embodiments presented herein may be implemented as a method, apparatus, or article using standard programming and/or engineering techniques. The term "article of manufacture" includes a computer program, carrier, or media accessible from any computer-readable device. For example, computer-readable media include magnetic storage devices (eg, hard disks, floppy disks, magnetic strips, etc.), optical disks (eg, CDs, DVDs, etc.), smart cards, and flash memory. device (eg, EEPROM, card, stick, key drive, etc.), but is not limited thereto. Additionally, various storage media presented herein include one or more devices and/or other machine-readable media for storing information. The term “machine-readable medium” includes, but is not limited to, wireless channels and various other media that can store, hold, and/or convey instruction(s) and/or data.

It is to be understood that the specific order or hierarchy of steps in the processes presented is an example of exemplary approaches. Based upon design priorities, it is to be understood that the specific order or hierarchy of steps in the processes may be rearranged within the scope of this disclosure. The accompanying method claims present elements of the various steps in a sample order, but are not meant to be limited to the specific order or hierarchy presented.

The description of the presented embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art of this disclosure, and the general principles defined herein may be applied to other embodiments without departing from the scope of this disclosure. Thus, the present disclosure is not to be limited to the embodiments presented herein, but is to be interpreted in the widest scope consistent with the principles and novel features presented herein.

Claims (5)

generating one or more translation models corresponding to one or more domains;
determining an optimal translation model corresponding to the user terminal from among the one or more translation models based on the received subscription request information when subscription request information is received from the user terminal;
generating a translated sentence corresponding to the received input sentence by utilizing the optimal translation model when translation request information including an input sentence is received from the user terminal;
determining one or more requesting translator terminals to request re-translation based on the received re-translation request information when re-translation request information related to the translated sentence is received from the user terminal;
generating a first input sentence by performing reverse translation on the translated sentence, and calculating translation reliability of the translated sentence based on a comparison between the generated first input sentence and the input sentence;
determining to transmit the translation reliability to the user terminal;
determining whether to transmit the translated sentence to the one or more requested translator terminals based on the translation reliability;
determining whether to transmit the retranslation request information to the one or more requested translator terminals based on the translation reliability;
determining to transmit the retranslation request information to the at least one requested translator terminal; and
determining to transmit a first retranslated sentence among the received one or more retranslated sentences to the user terminal when one or more retranslated sentences are received from the one or more requested translator terminals;
Including,
The subscription request information,
Information that is reference for determining the optimal translation model corresponding to the user terminal, and includes a customized learning data set including information related to a translation job performed at a previous time related to a user corresponding to the user terminal,
The translation request information,
Includes user language identification information and user domain identification information related to the source language and destination language;
The retranslation request information,
Information about a re-translation request for the translated sentence, which is generated based on the translation request information;
Determining the optimal translation model,
identifying whether the subscription request information includes the customized learning data set; and
Based on the identification result, performing learning on one or more network functions through the custom learning data set to generate the optimal translation model; including,
A method for providing a professional translation services platform, performed on one or more processors of a computing device. According to claim 1,
When one or more retranslated sentences are received from the one or more requested translator terminals, the step of determining to transmit a first retranslated sentence among the received one or more retranslated sentences to the user terminal,
determining to generate and transmit candidate retranslation list information including the one or more retranslated sentences to the user terminal; and
Identifying the first retranslated sentence among the one or more retranslated sentences based on selection information received from the user terminal in relation to the candidate retranslation list information;
A method for providing a professional translation services platform, performed on one or more processors of a computing device. According to claim 1,
Generating one or more translation models corresponding to the one or more domains,
obtaining a learning data set including a plurality of learning data;
constructing one or more training data subsets by classifying the plurality of training data based on one or more domains; and
Generating the one or more translation models by performing learning on each of the one or more neural network models through each of the one or more subsets of training data;
A method for providing a professional translation services platform, performed on one or more processors of a computing device. According to claim 3,
The step of constructing one or more learning data subsets by classifying the plurality of learning data based on one or more domains,
Identifying a domain of each of the plurality of learning data based on at least one of a semantic analysis result of each of the plurality of learning data, a repetition count for each word included in the semantic analysis result of each of the plurality of learning data, and a predetermined subject. ; and
Constructing the one or more subsets of learning data based on each domain of the plurality of learning data;
A method for providing a professional translation services platform, performed on one or more processors of a computing device. According to claim 1,
determining to transmit the received one or more retranslated sentences to the user terminal;
Receiving selection information for at least one retranslated sentence among the one or more retranslated sentences from the user terminal; and
Further comprising: identifying the first retranslated sentence from among the one or more retranslated sentences based on the received selection information;
A method for providing a professional translation services platform, performed on one or more processors of a computing device.