WO2016173326A1 - 基于主题的交互系统及方法 - Google Patents

基于主题的交互系统及方法 Download PDF

Info

Publication number
WO2016173326A1
WO2016173326A1 PCT/CN2016/076137 CN2016076137W WO2016173326A1 WO 2016173326 A1 WO2016173326 A1 WO 2016173326A1 CN 2016076137 W CN2016076137 W CN 2016076137W WO 2016173326 A1 WO2016173326 A1 WO 2016173326A1
Authority
WO
WIPO (PCT)
Prior art keywords
topic
conversation
input
current
theme
Prior art date
Application number
PCT/CN2016/076137
Other languages
English (en)
French (fr)
Inventor
聂华闻
Original Assignee
北京贝虎机器人技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京贝虎机器人技术有限公司 filed Critical 北京贝虎机器人技术有限公司
Publication of WO2016173326A1 publication Critical patent/WO2016173326A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor

Definitions

  • the present invention relates to the field of artificial intelligence technologies, and in particular, to a topic-based interaction system and method.
  • various aspects of the subject matter described herein relate to a topic-based interaction system and method, which can generate a conversation output based on a topic input by a conversation, and improve intelligence of conversation content matching to at least solve the low intelligence of chat conversation in related art. technical problem.
  • a subject-based interaction system can include: a topic matching device, configured to respond to an entity's current conversation input, based on a current conversation input and a degree of matching with each topic in the theme library, to obtain a current The dialog input matches the selected theme; and the dialog output generating means is configured to generate a dialog output corresponding to the current dialog input based on the corpus of the at least one dialog pair associated with the matching topic input by the current dialog.
  • the system may further include: a topic screening device, configured to analyze the theme of the current conversation input of the entity based on the feature of the entity, to obtain a topic corresponding to the current conversation input.
  • the dialog output generating device is further configured to generate, according to the current conversation input, a corpus containing at least one conversation pair associated with the topic, and generate a dialog output corresponding to the current conversation input.
  • features of the above entities may include, but are not limited to, at least one of age, occupation, location, education, gender, interests, historical conversation topics, or any combination.
  • the historical conversation topic may input one or more topics corresponding to the previous one or more conversations input with respect to the current conversation.
  • the above system may further include: a topic transfer monitoring means for determining whether the current dialog input transfers the conversation topic based on the pre-configured topic transfer rule in response to the current conversation input of the entity, when determining the current conversation input transfer
  • the dialog output generating means may be configured to generate a dialog output corresponding to the current dialog input based on the corpus containing the at least one conversation pair associated with the topic of the current conversation input matching.
  • the topic screening device may be instructed to analyze the theme of the current conversation input of the entity based on the feature of the entity to obtain the topic corresponding to the current conversation input.
  • the topic transfer rules described above can be configured to determine whether the dialog input transfers the conversation topic based on one or more preset keywords indicating the topic transfer.
  • the theme transfer monitoring device is further configured to determine a current dialog input transfer dialog topic when the current dialog input includes a preset keyword indicating the topic transfer.
  • the historical conversation topic of the entity may include a topic corresponding to the previous conversation input of the entity relative to the current conversation input.
  • the above system may further include: a historical data recording device for recording entity interaction history data including a historical conversation topic of the entity.
  • each topic is configured to include at least one topic word and/or topic sentence characterizing the topic.
  • the corpus may include at least one conversation pair, for each conversation pair, consisting of conversation input and dialog output.
  • Another aspect relates to a topic-based interaction method, the method comprising: in response to a current conversation input of an entity, inputting a matching degree of each topic in the theme library based on a current conversation input, obtaining a current conversation input matching theme; and, based on The current dialog inputs a corpus containing at least one conversation pair associated with the matching topic, producing a dialog output corresponding to the current conversation input.
  • the method may further include: analyzing, according to an entity's feature, an entity's current conversation input matching topic, obtaining a topic corresponding to the current conversation input; wherein, according to the current conversation input, the corresponding topic-related inclusion is included A corpus of at least one conversation pair that produces a dialog output corresponding to the current conversation input.
  • the characteristics of the entity include at least one of age, occupation, location, education, gender, interest, historical conversation theme, or any combination.
  • the method may further include: determining, in response to the current conversation input of the entity, whether the current conversation input transfers the conversation topic based on the pre-configured topic transfer rule; and, when determining the current conversation input to transfer the conversation topic , indicating a corpus containing at least one conversation pair associated with the topic based on the current conversation input, generating a conversation output corresponding to the current conversation input.
  • the entity-based feature analysis entity may be instructed to input the matching topic by the current conversation, and the topic corresponding to the current conversation input is obtained.
  • the topic transfer rule can be configured to determine whether the dialog input transfers the conversation topic based on one or more preset keywords indicating the topic transition, wherein the current conversation input includes a preset indication topic transition When the keyword is used, the current conversation input transition conversation topic is determined.
  • the invention relates to a computer program product comprising a computer readable medium having computer program logic recorded thereon, comprising computer program logic means for enabling a processor to perform any of the methods described above.
  • a topic-based interaction system and method are proposed, and a dialogue output is generated based on a theme, so that the matching of the conversation content is more reasonable and accurate, and at least the conversation content matching is not very accurate, and the intelligence is low.
  • the technical problems have achieved the technical effect of effectively improving the intelligence of interaction.
  • FIG. 1 is a block diagram of a communication system 100 capable of generating a dialog output based on a topic entered by a dialog;
  • FIG. 2 is a block diagram of an example of a topic service system 118
  • FIG. 3 is a block diagram of an example of a topic matching device 204
  • FIG. 4 is a flow chart of an example of a topic-based interaction method
  • FIG. 5 is a flow chart of another example of a topic-based interaction method
  • FIG. 6 is a schematic diagram of a topic-based interaction method
  • FIG. 7 is a structural block diagram of a topic-based interactive system
  • FIG. 8 is a block diagram of still another structure of a topic-based interactive system.
  • FIG. 1 is a block diagram of a communication system 100 that is capable of generating a dialog output based on a topic of conversation input.
  • system 100 includes first through nth computing devices 102a-102n (data devices 102a, 102b, and 102n are explicitly shown in FIG. 1), first server 104, storage system 106, second server 108 and network 110.
  • computing devices 102a, 102b through 102n can include an application 112, a mobile application 124, and a web application 116, respectively.
  • the first server 104 includes a topic service system 118 and the second server 108 includes a conversation output system 120.
  • Each of computing devices 102a-102n can be any type of fixed or mobile computing device, including a desktop computer (eg, a personal computer, etc.), a mobile computer, or a computing device (eg, a personal digital assistant (PDA), laptop, notebook computer , tablet computers such as Apple iPad/Microsoft Surface, netbooks, etc.), mobile phones (for example, cellular phones, smart phones such as Microsoft Windows Phone, Apple iPhone, Google Android Phone), robots with interactive dialogue capabilities, or Other types of other mobile or fixed computing devices.
  • PDA personal digital assistant
  • laptop notebook computer
  • tablet computers such as Apple iPad/Microsoft Surface, netbooks, etc.
  • mobile phones for example, cellular phones, smart phones such as Microsoft Windows Phone, Apple iPhone, Google Android Phone
  • robots with interactive dialogue capabilities or Other types of other mobile or fixed computing devices.
  • Each of the first server 104 and the second server 108 can be implemented in one or more computer systems, including any computing device or server.
  • Network 110 includes one or more communication links and/or communication networks, such as a personal area network (PAN), a local area network (LAN), a wide area network (WAN), or a collection of networks, such as the Internet.
  • PAN personal area network
  • LAN local area network
  • WAN wide area network
  • Computing devices 102a, 102b, and 102n and first server 104, second server 108 can be communicatively coupled to network 110 using various links, including wired and/or wireless links, such as IEEE 802.11 wireless links, global microwave connections Incoming interoperability (wi-max) links, cellular network links, Ethernet links, USB links, and the like.
  • Wi-max Incoming interoperability
  • Each of computing devices 102a-102n can be associated with one or more entities (e.g., users) that interact with the computing device.
  • N computing devices 102a-102n are shown in FIG. 1 for purposes of illustration. There may be any number of computing devices in system 100, including one, tens, hundreds, thousands, and a greater number of computing devices. Each computing device can operate one or more corresponding applications.
  • the subject service system 118 is disposed at the first server 104 and the dialog output system 120 is disposed at the second server 108.
  • the topic service system 118 and/or the conversation output system 120 in the system 100 can be located at any server or in any computing device.
  • storage system 106 is coupled to first server 104 and second server 108.
  • Storage system 106 can be coupled to first server 104 and second server 108 via network 110.
  • Storage system 106 can have the format of a database or other format, and can include one or more of any type of storage mechanism to store theme library 122 and corpus 124, including disks (eg, in a hard drive) or any other type of Storage medium.
  • the theme library 122 and/or corpus 124 can be stored on any computing device.
  • Any of computing devices 102a through 102n can interact with an entity including, but not limited to, conversational interactions, such as voice conversations, text conversations, and the like.
  • An entity may interact with a user interface displayed by an application at its computing device (eg, a web page displayed by a web browser or a user interface provided by another form of application), such as by application 112 at computing device 102a.
  • the entity can interact with a user interface provided by an application at its computing device, such as speaking to a computing device, and the like. As shown in FIG.
  • a dialog request 114 can be sent from one of the computing devices 102a-102n as the entity interacts with a user interface corresponding to one of the computing devices 102a-102n.
  • computing device 102a may send a conversation request 114 in a communication signal over network 110 to be received by topic service system 118 at first server 104, which may carry the entity's current conversation input 126.
  • Dialogue input 126 can include text or voice.
  • the mobile application 124 at the computing device 102b can enable an entity to input speech or enable an entity to enter text, as well as input text or the like by voice.
  • the topic service system 118 at the first server 104 can receive the conversation request 114. Responding to the connection Upon receipt of the dialog request 114, the subject service system 118 determines the subject of the entity's current conversation input match.
  • the topic service system 114 can determine the degree of matching of the current conversation input 126 with each topic in the theme library 122 to get a topic that the current conversation input 126 matches. For example, the topic service system 118 can determine the degree of matching of the current conversation input 126 with each topic such that the topic with the highest degree of matching is the topic that the current conversation input 126 matches.
  • the subject service system 118 at the first server 104 sends a conversation request 128 in the communication signal over the network 110 for receipt by the conversation output system 118 at the second server 108.
  • the conversation request 128 can carry the subject of the current conversation input 126 and the current conversation input 126.
  • the dialog output system 118 at the second server 108 can receive the dialog request 128. In response to the received dialog request 128, the dialog output system 118 generates a dialog output 130 corresponding to the current dialog input based on the currently associated corpus 124 associated with the topic input 126. Corpus 124 contains at least one conversation pair, each conversation pair being comprised of dialog input and dialog output. For example, the dialog output system 118 can determine the degree of match between the current dialog input 126 and the dialog input in the conversation pair, resulting in a conversation pair that is most similar to the current conversation input 126, and outputting the conversation in the conversation pair as the current conversation input. 126 dialogue output.
  • the dialog output system 118 at the second server 108 can transmit the generated dialog output 130 in the communication signal over the network 110 for the application at the computing device (here, the mobile application 114 at the computing device 102b, for example, A dialog output 130 is received at any of the devices 102a-102n.
  • the dialog output 130 can include text or speech.
  • the mobile application 114 at the computing device 102b can receive the conversation output 130.
  • the mobile application 114 at the computing device 102b can cause the dialog output 130 to be presented to the entity (user).
  • the mobile application 114 can display the dialog output 130 (text) to the entity via a graphical user interface (GUI), or synthesize the speech data of the dialog output 130 (text) via a speech synthesis function, and play the speech of the dialog output 130 via the audio playback device. Or play the dialog output 130 (voice) through the audio playback device.
  • GUI graphical user interface
  • Each of the computing devices 102a-102n can interact with the entity multiple times.
  • each of computing devices 102a-102n can have multiple rounds of conversations with an entity.
  • the entity may enter a dialog input through a user interaction interface provided by the computing device, and the computing device may send a dialog input in the communication signal over the network 110 for receipt at the subject service system 118 at the first server 104.
  • the subject service system 118 can determine the subject of the dialog input and send the topic of the dialog input in the communication signal over the network 110 for receipt at the dialog output system 118 at the second server 108.
  • the dialog output system 118 can determine the dialog output based on the dialog input and its subject matter.
  • the dialog output system 118 can transmit the determined dialog output in the communication signal over the network 110 for receipt at the corresponding computing device.
  • the topic service system 118 at the first server 104 can determine the topic corresponding to the current conversation input 126 of the entity based on the entity history dialog input corresponding topic and the entity's current conversation input matching topic.
  • the theme service The system 118 may determine the topic that the current conversation input 126 matches based on the current conversation input of the entity, for example, based on the current conversation input and the degree of matching of the topics in the theme library to determine the topic that the current conversation input 126 matches, the current conversation input 126 matches.
  • the subject can be one or more.
  • the topic service system 118 may select the topic with the highest similarity to the historical conversation topic of the entity from among the plurality of topics that match the current conversation input 126 as the topic corresponding to the current conversation input 126.
  • the topic service system 118 is configured to treat topics having a higher degree of matching than a preset value as a plurality of topics that match the current conversation input 126, and/or a theme configured to match the matching degree from high to low. Multiple topics that match 126 as the current conversation input.
  • the topic service system 118 at the first server 104 can be configured to input the corresponding topic relative to the previous conversation input 126 of the current conversation as a historical conversation topic for determining the topic corresponding to the current conversation input 126.
  • the topic service system 118 at the first server 104 may, after determining the topic corresponding to the current conversation input 126, store entity conversation history data including the topic corresponding to the current conversation input 126 to receive the entity
  • the corresponding topic is input using the stored dialog to determine the topic corresponding to the received next dialog input.
  • the topic service system 118 at the first server 104 can determine the topic corresponding to the entity's current conversation input 126 based on features other than the historical conversation topic.
  • features of an entity may include, but are not limited to, at least one of age, occupation, location, education, gender, interest, or any combination.
  • the topic service system 118 at the first server 104 can be configured to determine age, occupation, location, education, gender, interest based on the correspondence between the topics in the theme library and age, occupation, location, education, gender, interests, and the like. Corresponding theme.
  • the topic service system 118 at the first server 104 determines a plurality of topics that match the entity's current conversation input 126, and determines a topic that matches the features of the entity based on the characteristics of the entity, resulting in a topic corresponding to the current conversation input 126.
  • the topic service system 118 at the first server 104 can determine whether the current conversation input 126 of the entity transfers the conversation topic based on the topic transfer rules.
  • the topic transfer rule can be configured to determine whether the dialog input transfers the conversation topic based on one or more preset keywords indicating the topic transfer.
  • the topic service system 118 at the first server 104 can determine the topic that the current conversation input 126 matches when monitoring the transition conversation topic. As described above, the topic with the highest similarity of the current conversation input 126 can be selected as the conversation output.
  • the subject; the subject corresponding to the current conversation input 126 is determined based on the characteristics of the entity when the transition conversation topic is not monitored.
  • FIG. 2 is a block diagram of an example of a topic service system 118.
  • the topic service system 118 can include a topic transfer monitoring device 202, a topic matching device 204, a topic screening device 206, and an output interface 208.
  • the topic transfer monitoring device 202 is configured to respond to the dialog request 201, and the dialog request 201 can carry the current dialog input 126 of the entity, and based on the topic transfer rule, determine whether the current dialog input 126 of the entity transfers the conversation topic.
  • the subject matching device 204 is instructed to determine the topic to which the current conversation input 126 matches.
  • the topic matching device 204 is configured to determine a topic in the theme library 212 that matches the current conversation input 126 based on the indication of the topic transition monitoring device 202, and the topic with the highest matching degree may be the subject corresponding to the current conversation input 126, and
  • the determined subject is sent to output interface 208.
  • the output interface 208 transmits the determined subject in the communication signal over the network 110 (shown in FIG. 1) for receipt at the dialog output system 118 at the second server 108 as shown in FIG.
  • the topic transfer monitoring device 202 is further configured to, when the entity's current dialog input 126 has not transferred the conversation topic, instruct the topic matching device 204 to determine a topic for which the current conversation input 126 matches, the topic screening device 206 matches based on the current conversation input 126.
  • the topic determines the topic corresponding to the current conversation input 126.
  • the topic matching device 204 is configured to determine topics in the theme library 212 that match the current conversation input 126, and may send the determined topics that match the current conversation input 126 to the topic screening device 206.
  • the topic screening device 206 is configured to filter out topics corresponding to the current conversation input 126 from topics that match the current conversation input 126 based on the characteristics of the entity.
  • the topic screening device 206 can send the topic corresponding to the current dialog input 126 to the output interface 208.
  • Output interface 208 can transmit a topic corresponding to current dialog input 126 in the communication signal over network 110 (shown in FIG. 1) for receipt at dialog output system 118 at second server 108 as shown in FIG. .
  • the topic transfer monitoring device 202 can determine whether the current conversation input 126 of the entity transfers the conversation topic in accordance with the pre-configured topic transfer rules 210.
  • the topic transfer rule 210 may include, but is not limited to, keywords and/or key sentences indicating a topic transfer, such as "changing a topic", "not wanting to talk about this", and the like. In some examples, any phrase that can represent a conversation topic transfer can be used for topic transfer rule 210.
  • the theme library 212 can be configured to include a plurality of topics, each of the plurality of topics being comprised of one or more keywords and/or key sentences, the one or more keywords and/or key sentences A subject vocabulary that can form a topic. There may be a concept inclusion and an inclusion relationship between the keywords, which may be presented in a tree structure to constitute a keyword tree. Each topic may correspond to one or more corpora, and the corpus may consist of a conversation pair that includes dialog input and dialog output.
  • the topic matching device 204 can perform word segmentation on the entity's current conversation input 126 to obtain one or more dialog phrases for topic matching. For each topic, the topic matching device 204 may determine the distance between at least part of the words of the dialog phrase and the keywords in the topic vocabulary, obtain the similarity between the dialog phrase and each topic, and measure the current conversation input 126 of the entity by the similarity degree. The degree of matching of the theme. In one example, similar The highest-level topic is the subject corresponding to the current conversation input 126 of the entity, or one or more topics whose similarity is higher than the preset value is used as the theme of the entity's current conversation input 126, or the preset with high similarity is selected. The number of topics is entered as the subject of the entity's current conversation input 126.
  • the topic matching device 204 can filter the dialog phrases, such as a noun as a word for topic matching.
  • the topic matching device 204 can also determine the weights of the words of the dialog phrase, and sum the distances of at least some of the words in the dialog phrase with the keywords in the keyword pool according to the determined weights to obtain the current conversation input 126 and each topic of the entity. Similarity.
  • FIG. 3 is a block diagram of an example of the topic matching device 204.
  • the topic matching device 204 can include a word segmentation module 302, a screening module 304, and a similarity determination module 306.
  • the word segmentation module 302 can perform word segmentation on the dialog input to obtain a dialog phrase input by the dialog.
  • the screening module 304 may filter the dialog phrases obtained by the word segmentation module 302 to obtain one or more words for performing topic matching, wherein the determined words may be part or all of the dialogue phrases.
  • the similarity determination module 306 can determine the distance between the word selected by the screening module 304 and the keyword in the subject vocabulary, and then obtain the similarity between the word and the subject selected by the screening module 304 based on the determined distance.
  • the topic matching device 204 can also include a weight determination module 308.
  • the weight determination module 308 can determine the weights of the individual words selected by the screening module 304, for example, assign different weights to the verbs and nouns, assign different weights to the real words and the virtual words, and assign different weights to different part of speech in the real words.
  • the similarity determination module 306 can sum the distances of the filtered words and the keywords in the subject vocabulary according to the determined weights to obtain the similarity between the dialogue input and each topic.
  • the weight determination module 308 can directly determine the weight of each word in the phrase, and the screening module 304 can not filter.
  • the similarity determination module 306 can sum the distances of each word in the dialog phrase with the keywords in the theme library according to the determined weights, and obtain the similarity between the dialogue input and the topic.
  • the theme library 212 can be configured to include a plurality of topics, each of the plurality of topics being comprised of one or more key sentences that can form a subject vocabulary for the topic.
  • the topic matching device 204 can determine the similarity between the dialog input and the key sentence in the topic vocabulary, and obtain the similarity between the dialog input and the topic. For example, the topic matching device 204 can segment the dialogue input and the key sentence, calculate the word frequency, obtain the word frequency vector of the dialogue input and the key sentence, calculate the vector cosine value of the speech input vector and the key word vector of the key sentence, and then obtain the dialogue input and the key.
  • the similarity of sentences can be configured to include a plurality of topics, each of the plurality of topics being comprised of one or more key sentences that can form a subject vocabulary for the topic.
  • the topic matching device 204 can determine the similarity between the dialog input and the key sentence in the topic vocabulary, and obtain the similarity between the dialog input and the topic. For example, the topic matching device 204 can segment the dialogue input and the
  • FIG. 4 is a flow chart of an example of a topic-based interaction method.
  • the method can be applied in the environment as shown in FIGS. 1, 2, and 3, but is not limited thereto.
  • the method can include steps 402 through 404.
  • the current conversation input matching topic is obtained (step 402).
  • a corpus containing at least one conversation pair associated with the matching topic of the current conversation input generates a dialog output corresponding to the current conversation input (step 404).
  • the topic of the entity's current conversation input matching may be analyzed based on the feature of the entity to obtain a topic corresponding to the current conversation input (step 406).
  • the conversation output corresponding to the current conversation input may be generated based on the corpus of the at least one conversation pair associated with the corresponding topic input.
  • the characteristics of the entity may include, but are not limited to, at least one of age, occupation, location, education, gender, interest, historical conversation theme, or any combination.
  • the current conversation input may be determined to transfer the conversation topic based on the pre-configured topic transfer rules in response to the current conversation input of the entity (step 406); and, when determining the current conversation input to transfer the conversation topic, A corpus containing at least one conversation pair associated with the topic based on the current conversation input is instructed to generate a dialog output corresponding to the current conversation input (step 402).
  • the entity-based feature analysis entity may be instructed to input the matching topic, and the current conversation input corresponding topic is obtained (step 406).
  • the topic transfer rule may be configured to determine whether the dialog input transfers the conversation topic based on one or more preset keywords indicating the topic transition, wherein when the current conversation input includes a preset keyword indicating the topic transition, determining The current conversation input shifts the conversation topic.
  • FIG. 5 is a flow chart of another example of a topic-based interaction method.
  • the topic of the dialog input matching/correspondence is determined by keywords in this example.
  • the method can include steps 502 through 506.
  • the user's dialog input is received through the user interface, and the dialog input is voice-recognized to obtain voice data input by the dialog.
  • the content of the voice may be recognized when the user hears the sound.
  • the voice data is received and keywords are extracted from the voice data (step 502).
  • the recognized voice data is: I want to cook, but I don't know what to cook. Analysis of this sentence, you can extract the key words: cooking, or dishes. That is, it may be that the received voice information is decomposed and confirmed, and a plurality of fields in the voice data are confirmed, and a solid word with a clear meaning is found from the plurality of decomposed fields, or a word with a clear pointing is found.
  • the voice data is: recommend a few good-looking movies to me, and the corresponding keywords can be: movies.
  • the voice data is: What is the weather like today, the corresponding keyword can be: weather.
  • the keywords in the sentence are analyzed. For example, if the keyword is a movie, the robot knows. What you want to discuss with me is the content related to the movie. If the keyword is weather, the robot knows. To discuss the weather-related content, you can talk about the conversation theme set to movie, or set to weather.
  • a dialog topic matching the keyword is found in the theme library (step 504).
  • a theme library can be stored.
  • a plurality of conversation topics can be preset. The selection of the conversation topics can be artificially recorded or summarized, for example, the people can be chatted.
  • the topics that are often involved or discussed in the process are the subject of dialogue, or they can be the subject of dialogue.
  • the topic of the conversation corresponding to the keyword can be found in the theme library. For example, if the keyword is weather, it can be matched in the theme library.
  • the theme can be the same word or a similar word, for example, the weather, the corresponding theme can be the weather itself. It can also be a similar word such as climate.
  • the theme in the theme library can also exist in large libraries and small libraries. Take “movie” as an example. Maybe match the theme of the big library: movie drama, small library is: movie, or specifically in the subsequent processing, match out more
  • the theme of the small library for example, specific to a movie. That is, at the time of processing, through the broad to specific theme, it is also possible to achieve more precise setting and matching of conversation topics.
  • a conversation topic can correspond to a corpus. After matching the theme, the corpus corresponding to the conversation topic may be pre-trained, and the human-computer interaction is performed as the dialogue matching material of the human-computer interaction (step 506).
  • a huge database is preset in the system database.
  • This database stores a corpus of multiple conversation topics, that is, each conversation topic can correspond to a corpus.
  • a corpus can consist of multiple question and answer pairs. Because each conversation topic corresponds to a corpus, after determining the conversation topic of the voice data, the data matching can be directly located in the corpus corresponding to the conversation topic. Because the selected conversation content is based on this determined conversational theme, the conversation is made closer to the actual human interaction.
  • a relationship can be set between the question and answer pairs of the conversation topic corpus and the question and answer pair, so that after the robot answers the first question raised by the person, it can also trigger the subsequent dialogue, thereby making the communication Can continue.
  • the party that asks can initiate a follow-up question: Can you tell me how to do it? It can also be the main trigger of the answering party: need me to tell you how to do it? This can be achieved by setting the relationship between the question and answer pair and the question and answer pair.
  • a question and answer pair with a semantic relationship can be associated, and a jump condition can be set, and if the jump condition is satisfied, You can skip to a follow-up question or answer pair or a related question and answer pair for a follow-up conversation.
  • the human-computer interaction in the so-called dialogue is a process of speech recognition and corresponding speech output, that is, a question and answer corresponding to the speech data can be matched from the determined corpus. Yes, and the answer content in the question and answer pair is output as the output content.
  • the dialogue theme jump mechanism that is, in the process of human-computer interaction, it is determined at the moment whether a topic jump is needed, that is, whether the current conversation theme changes, and if a change occurs, the corpus corresponding to the changed conversation theme is used as a person.
  • the interactive dialogue of the machine matches the material for human-computer interaction. There are still many situations that trigger the change of this conversation theme. The following two examples are used as examples:
  • the complexity and completeness of corpus content often have an important impact on the accuracy of human-computer interaction.
  • the webpage page may be crawled from the Internet, and then the content on the crawled webpage page is used as the training data for the training of the corpus, wherein in the training process, the content on the same webpage page is determined to be based on the same The content of the topic.
  • the training library can be greatly enriched, and the resulting corpus is more perfect and specific, and the final human-computer interaction is closer to people and The real interaction between people.
  • the robot can also perform training in real time when talking to a person, or "hear" the conversation of the surrounding people. At the same time, learning is also carried out, so that the corpus of training is more complete and comprehensive.
  • FIG. 7 is a structural block diagram of a topic-based interaction system, as shown in FIG. 7, including: a receiving unit 701, a searching unit 702, and an interaction unit 703. The structure will be described below.
  • the receiving unit 701 is configured to receive voice data, and extract keywords from the voice data;
  • the searching unit 702 is connected to the receiving unit 701, and is configured to find, in the theme library, a dialog topic that matches the keyword;
  • the interaction unit 703 is connected to the search unit 702, and is configured to perform human-computer interaction as a human-computer interaction dialog matching material by using a pre-trained corpus corresponding to the conversation theme, wherein a conversation theme corresponds to a corpus.
  • a corpus consists of multiple question and answer pairs.
  • the interaction unit 703 is specifically configured to match a question and answer pair corresponding to the voice data from the corpus, and output the answer content in the question and answer pair as the output content.
  • the device for implementing the theme-based human-computer interaction engine may further include: a determining unit 801, configured to perform human-computer interaction on the corpus corresponding to the conversation topic obtained by pre-training. In the process of interacting with the material for human-computer interaction, determining whether the current conversation theme changes; the jumping unit 802 is configured to use the corpus corresponding to the changed conversation theme as a human-machine when determining that the conversation theme changes. Interactive conversations match the material for human interaction.
  • the determining unit 801 is specifically configured to determine whether the current conversation topic changes when the current corpus does not match the question and answer pair corresponding to the voice data input by the current human-machine interaction; or, currently received It is determined whether the current conversation topic has changed when the interval between the voice data and the last received voice data is greater than a predetermined time threshold.
  • the apparatus for implementing the theme-based human-computer interaction engine may further include a training unit.
  • the webpage is crawled from the Internet; the content on the crawled webpage is used as the training data for the training of the corpus, wherein in the training process, the content on the same webpage is determined as the content based on the same theme. .
  • a storage medium is further provided, wherein the software includes the above-mentioned software, including but not limited to: an optical disk, a floppy disk, a hard disk, an erasable memory, and the like.
  • the embodiment of the present invention achieves the following technical effects: the corpus on which the interaction is based is based on the conversation theme, and the general conversation is based on the habit of a certain topic when the conversation is usually with people. It is consistent, which makes the matching of this dialogue content more reasonable and accurate, effectively solves the technical problem that the dialogue content matching in the prior art is not very accurate and low in intelligence, and achieves the technical effect of effectively improving the intelligence of human-computer interaction. .
  • modules or steps of the embodiments of the present invention can be implemented by a general computing device, which can be concentrated on a single computing device or distributed in multiple computing devices. Alternatively, they may be implemented by program code executable by the computing device such that they may be stored in the storage device by the computing device and, in some cases, may be different from The steps shown or described are performed sequentially, or they are separately fabricated into individual integrated circuit modules, or a plurality of modules or steps thereof are fabricated into a single integrated circuit module. Thus, embodiments of the invention are not limited to any specific combination of hardware and software.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Machine Translation (AREA)

Abstract

一种基于主题的交互系统及方法，该方法包括：响应于实体当前的对话输入，基于当前的对话输入与主题库中各个主题的匹配度，得到当前的对话输入匹配的主题；以及，基于当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生当前的对话输入对应的对话输出。该方法解决了现有技术中对话内容匹配不是很准确，智能性低的技术问题，达到了有效提高人机交互智能性的技术效果。

Description

基于主题的交互系统及方法
本申请要求2015年04月30日递交的申请号为201510212167.9、发明名称为“基于主题的人机交互引擎的实现方法和装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本发明涉及人工智能技术领域，特别涉及一种基于主题的交互系统及方法。
背景技术
随着科技的发展和人们生活水平的不断提高，人工智能进入了千家万户。目前，人工智能已经不再局限于以前的智能扫地机器人等这些常规的智能设备，能与人进行人机交互，或者智能聊天的机器人也开始逐渐出现和普及。
既然是智能聊天机器人，人们所期望的是这些机器人与人进行人机交互的流畅性和合理性越高越好，最理想的状态是，让人感受不到是与机器在聊天，而是感觉和一个真实的人聊天一样。
然而，基于现在的智能聊天技术，一般都是有一个很大的数据库，从这个数据库中进行对话内容的匹配，因为数据库中数据的不完备，以及数据组织的散乱等问题的存在，使得现在的智能聊天机器人进行聊天的时候，不够智能化，且一般每一句的交流都需要人进行触发，机器人自身一般不能够有效地进行会话的展开和持续。
针对上述问题，目前尚未提出有效的解决方案。
发明内容
提供本概述以便以简化形式介绍将在以下具体实施方式中进一步描述的一些带表性概念。本概述不旨在表示出所要求保护的主题的关键特征或必要特征，也不旨在以限制所要求保护的主题和范围的任何方式来使用。
简要地，在此描述的主题的各个方面涉及基于主题的交互系统及方法，可基于对话输入的主题产生对话输出，提高对话内容匹配的智能度，以至少解决相关技术中聊天对话智能度低的技术问题。
一个方面，涉及基于主题的交互系统，该系统可包括：主题匹配装置，用于响应于实体当前的对话输入，基于当前的对话输入与主题库中各个主题的匹配度，得到当前的 对话输入匹配的主题；以及，对话输出产生装置，用于基于当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生当前的对话输入对应的对话输出。
在一个示例中，上述系统还可包括：主题筛选装置，用于基于实体的特征分析实体当前的对话输入匹配的主题，得到当前的对话输入对应的主题。其中，对话输出产生装置进一步用于基于当前的对话输入对应的主题相关联的包含至少一个对话对的语料库，产生当前的对话输入对应的对话输出。
在一个示例中，上述实体的特征可包括但不限于年龄、职业、所在地、学历、性别、兴趣、历史对话主题中至少之一或者任意组合。其中，历史对话主题可为相对于当前的对话输入的上一个或多个对话输入对应的一个或多个主题。
在一个示例中，上述系统还可包括：主题转移监控装置，用于响应于实体当前的对话输入，基于预先配置的主题转移规则判断当前的对话输入是否转移对话主题，当确定当前的对话输入转移对话主题时，可指示对话输出产生装置基于当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生当前的对话输入对应的对话输出。当确定当前的对话输入未转移对话主题时，可指示主题筛选装置基于实体的特征分析实体当前的对话输入匹配的主题，得到当前的对话输入对应的主题。
在一个示例中，上述主题转移规则可被配置为基于一个或多个预设的指示主题转移的关键词确定对话输入是否转移对话主题。其中，主题转移监控装置，进一步用于在当前的对话输入中包含预设的指示主题转移的关键词时，确定当前的对话输入转移对话主题。
在一个示例中，实体的历史对话主题可包括实体的相对于当前的对话输入的上一个对话输入对应的主题。
在一个示例中，上述系统还可包括：历史数据记录装置，用于记录包含实体的历史对话主题在内的实体交互历史数据。
在一个示例中，对于所述主题库中的每个主题，每个主题被配置为包括至少一个表征主题的主题词和/或主题句。语料库可包括至少一个对话对，对于每个对话对，可由对话输入和对话输出构成。
另一个方面，涉及基于主题的交互方法，该方法包括：响应于实体当前的对话输入，基于当前的对话输入与主题库中各个主题的匹配度，得到当前的对话输入匹配的主题；以及，基于当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生当前的对话输入对应的对话输出。
在一个示例中，该方法还可包括：基于实体的特征分析实体当前的对话输入匹配的主题，得到所述当前的对话输入对应的主题；其中，基于当前的对话输入对应的主题相关联的包含至少一个对话对的语料库，产生当前的对话输入对应的对话输出。
在一个示例中，实体的特征包括年龄、职业、所在地、学历、性别、兴趣、历史对话主题中至少之一或者任意组合。
在一个示例中，上述方法还可包括：响应于实体当前的对话输入，基于预先配置的主题转移规则判断所述当前的对话输入是否转移对话主题；以及，当确定当前的对话输入转移对话主题时，指示基于当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生当前的对话输入对应的对话输出。当确定当前的对话输入未转移对话主题时，可指示基于实体的特征分析实体当前的对话输入匹配的主题，得到当前的对话输入对应的主题。
在一个示例中，主题转移规则可被配置为基于一个或多个预设的指示主题转移的关键词确定对话输入是否转移对话主题，其中，在当前的对话输入中包含预设的指示主题转移的关键词时，确定当前的对话输入转移对话主题。
再一方面，还涉及包括具有记录在其上的计算机程序逻辑的计算机可读介质的计算机程序产品，包括用于使得处理器能够执行本实施例上述任意方法的计算机程序逻辑构件。
在本发明实施例中，提出了一种基于主题的交互系统及方法，基于主题产生对话输出，使得这种对话内容的匹配更为合理准确，至少能够解决对话内容匹配不是很准确，智能性低的技术问题，达到了有效提高交互智能性的技术效果。
附图说明
此处所说明的附图用来提供对本发明的进一步理解，构成本申请的一部分，并不构成对本发明的限定。在附图中：
图1为能够基于对话输入的主题产生对话输出的通信系统100的框图；
图2为主题服务系统118一个示例的框图；
图3为主题匹配装置204一个示例的框图；
图4为基于主题的交互方法一示例的流程图；
图5为基于主题的交互方法另一示例的流程图；
图6为基于主题的交互方法的一示意图；
图7为基于主题的交互系统的一种结构框图；以及
图8为基于主题的交互系统的又一种结构框图。
具体实施方式
为使本发明的目的、技术方案和优点更加清楚明白，下面结合实施方式和附图，对本发明做进一步详细说明。在此，本发明的示意性实施方式及其说明用于解释本发明，但并不作为对本发明的限定。
应当理解，此处的任何示例都是非限制性的。因此，本发明不限于此处所描述的任何特定实施例、方面、概念、结构、功能或示例。相反，此处所描述的任何一个实施例、方面、结构、功能或示例都是为限制性的。
针对一个实施例或示例描述和/或例示的特征，可以在一个或更多个其它实施例或示例中以相同方式或以类似方式使用，和/或与其他实施例或示例的特征相结合或代替其他实施例或示例的特征。
应当强调的是，词语“包括”、“基于”或“根据”当在本说明书中使用时用来指所引述的特征、要素、步骤或组成部分的存在，但不排除一个或更多个其它特征、要素、步骤、组成部分或它们的组合的存在或增加。
图1为能够基于对话输入的主题产生对话输出的通信系统100的框图。
如图1所示，系统100包括第一至第n计算设备102a-102n(在图1中明确地示出了计算设备102a、102b以及102n)、第一服务器104、存储系统106、第二服务器108以及网络110。如图1进一步所示，计算设备102a、102b至102n可分别包括应用程序112、移动应用程序124以及web应用程序116。更近一步的，第一服务器104包括主题服务系统118，并且第二服务器108包括对话输出系统120。
计算设备102a-102n中每个可以是任何类型的固定或移动计算设备，包括台式计算机(例如个人计算机等)、移动计算机或计算设备(例如，个人数字助理(PDA)、膝上计算机、笔记本计算机、诸如苹果iPad/微软Surface之类的平板计算机、上网本等)、移动电话(例如，蜂窝电话、诸如微软Windows电话、苹果iPhone、谷歌Android电话之类的智能电话)、具有对话交互能力的机器人或者其他类型的其他移动或固定计算设备。第一服务器104和第二服务器108中每个可在一个或多个计算机系统中实施，包括任何计算设备或者服务器。
计算设备102a至102n、第一服务器104、第二服务器108以及存储系统106由网络 110通信地耦合。网络110包括一个或多个通信链接和/或通信网络，诸如个域网(PAN)、局域网(LAN)、广域网(WAN)或者网络的集合，例如因特网(Internet)。计算设备102a、102b和102n以及第一服务器104、第二服务器108可使用各种链路通信地耦合到网络110，包括有线和/或无线链路，诸如IEEE802.11无线链路、全球微波接入互操作(wi-max)链路、蜂窝式网络链路、以太网链路、USB链路等。
计算设备102a-102n中每个可与一个或多个实体(例如用户)相关联，该一个或多个实体与计算设备交互。在图1中出于举例说明的目的示出了n个计算设备102a-102n。在系统100中可存在任何数目的计算设备，包括一个、数十个、数百个、成千上万以及更大数目的计算设备。每个计算设备可操作一个或多个对应的应用程序。
在图1中出于示意性目的，示出了主题服务系统118设置于第一服务器104、并且对话输出系统120设置于第二服务器108的情形。在系统100中主题服务系统118和/或对话输出系统120可设置在任意服务器，或者设置在任意计算设备中。
如图1所示，存储系统106被耦合到第一服务器104和第二服务器108。可将存储系统106通过网络110耦合到第一服务器104和第二服务器108。存储系统106可具有数据库的格式或其他格式，并且可包括任何类型的存储机制中的一个或多个以存储主题库122和语料库124，包括磁盘(例如，在硬盘驱动器中)或者任何其他类型的存储介质。主题库122和/或语料库124可存储在任意计算设备。
计算设备102a至102n中的任一个可与实体进行交互，该交互包括但不限于对话交互，例如语音对话、文字对话等。实体可与由其计算设备处的应用程序显示的用户接口(例如，由web浏览器显示的网页或者由另一形式的应用程序提供的用户接口)交互，诸如由计算设备102a处的应用程序112、计算设备102b处的移动应用程序124、或者计算设备102n处的web应用程序116显示的用户接口。或者实体可与由其计算设备处的应用程序提供的用户接口交互，例如对着计算设备讲话等。如图1所示，由于实体与计算设备102a-102n中之一对应的用户接口交互，可从计算设备102a-102n中的一个发送对话请求114。例如，计算设备102a可通过网络110在通信信号中发送对话请求114，以在第一服务器104处被主题服务系统118接收到，对话请求114中可携带实体当前的对话输入126。
对话输入126可包括文本或者语音。例如，计算设备102b处的移动应用程序124，可以使得实体能够输入语音，或者使得实体能够输入文本，以及通过语音来输入文本等。
如图1所示，第一服务器104处的主题服务系118可接收对话请求114。响应于接 收到的对话请求114，主题服务系118确定实体当前的对话输入匹配的主题。主题服务系统114可确定当前的对话输入126与主题库122中各个主题的匹配度，以得到当前的对话输入126匹配的主题。例如，主题服务系118可确定当前的对话输入126与各个主题的匹配度，使得匹配度最高的主题为当前的对话输入126匹配的主题。第一服务器104处的主题服务系118通过网络110在通信信号中发送对话请求128，以在第二服务器108处被对话输出系统118接收到。对话请求128中可携带当前的对话输入126匹配的主题以及当前的对话输入126。
第二服务器108处的对话输出系统118可接收对话请求128。响应于接收到的对话请求128，对话输出系统118基于当前的对话输入126匹配的主题相关联的语料库124，产生当前的对话输入对应的对话输出130。语料库124包含至少一个对话对，每个对话对可由对话输入和对话输出构成。例如，对话输出系统118可确定当前的对话输入126和对话对中的对话输入的匹配度，得到与当前的对话输入126最相似的对话对，将该对话对中的对话输出作为当前的对话输入126的对话输出。第二服务器108处的对话输出系统118可通过网络110在通信信号中发送产生的对话输出130，以在计算设备处的应用程序(在此以计算设备102b处的移动应用程序114为例，计算设备102a-102n中任一计算设备也是可行的)处接收对话输出130。对话输出130可包括文本或者语音。
计算设备102b处的移动应用程序114可接收对话输出130。响应于接收到的对话输出130，计算设备102b处的移动应用程序114可使得对话输出130被呈现给实体(用户)。例如，移动应用程序114可通过图形用户接口(GUI)向实体显示对话输出130(文本)，或者通过语音合成功能合成对话输出130(文本)的语音数据，通过音频播放装置播放对话输出130的语音，或者通过音频播放装置播放对话输出130(语音)。
计算设备102a-102n中的每一个可与实体多次交互。例如，计算设备102a-102n中的每一个可与实体连续多轮对话。对于每轮对话，实体可通过计算设备提供的用户交互接口输入对话输入，计算设备可通过网络110在通信信号中发送对话输入，以在第一服务器104处的主题服务系118处被接收。主题服务系118可确定对话输入的主题，并通过网络110在通信信号中发送对话输入的主题，以在第二服务器108处的对话输出系统118处被接收。对话输出系统118可基于对话输入及其主题确定对话输出。对话输出系统118可通过网络110在通信信号中发送确定得到的对话输出，以在对应的计算设备处被接收。
第一服务器104处的主题服务系118可基于实体历史对话输入对应的主题和实体当前的对话输入匹配的主题，确定实体当前的对话输入126对应的主题。例如，主题服务 系118可基于实体当前的对话输入确定当前的对话输入126匹配的主题，例如基于当前的对话输入与主题库中各主题的匹配度确定当前对话输入126匹配的主题，当前的对话输入126匹配的主题可为一个或多个。主题服务系118可从与当前对话输入126匹配的多个主题中，选择与实体的历史对话主题相似度最高的主题作为当前的对话输入126对应的主题。主题服务系118被配置为将匹配度高于预设值的主题作为与当前的对话输入126匹配的多个主题，和/或被配置为将匹配度由高到低的预设个数的主题作为当前对话输入126匹配的多个主题。
第一服务器104处的主题服务系118，可被配置为将相对于当前的对话输入126的上一对话输入对应的主题，作为用于确定当前的对话输入126对应的主题的历史对话主题。在一个示例中，第一服务器104处的主题服务系118可在确定当前的对话输入126对应的主题后，存储包括当前对话输入126对应的主题在内的实体对话历史数据，以在接收到实体的下一对话输入时，使用存储的对话输入对应的主题来确定接收到的下一对话输入对应的主题。
第一服务器104处的主题服务系118，可根据实体除历史对话主题之外的特征确定实体当前的对话输入126对应的主题。例如，实体的特征可包括但不限于年龄、职业、所在地、学历、性别、兴趣中至少之一或者任意组合。第一服务器104处的主题服务系118可被配置为基于主题库中各主题与年龄、职业、所在地、学历、性别、兴趣等的对应关系，确定与年龄、职业、所在地、学历、性别、兴趣对应的主题。例如，第一服务器104处的主题服务系118确定与实体当前的对话输入126匹配的多个主题，并基于实体的特征确定与实体的特征匹配的主题，得到当前的对话输入126对应的主题。
在一个示例中，第一服务器104处的主题服务系118可基于主题转移规则判断实体当前的对话输入126是否转移对话主题。例如，主题转移规则可被配置为基于一个或多个预设的指示主题转移的关键词确定对话输入是否转移对话主题。第一服务器104处的主题服务系118，可在监控到转移对话主题时，确定当前的对话输入126匹配的主题，如前所述可选择当前的对话输入126相似度最高的主题作为产生对话输出的主题；在未监控到转移对话主题时，基于实体的特征确定当前的对话输入126对应的主题。
图2为主题服务系统118一个示例的框图。
如图2所示，主题服务系统118可包括主题转移监控装置202、主题匹配装置204、主题筛选装置206以及输出接口208。
主题转移监控装置202，被配置为响应对话请求201，对话请求201可携带实体当前的对话输入126，基于主题转移规则判断实体当前的对话输入126是否转移对话主题。当实体当前的对话输入126转移对话主题时，指示主题匹配装置204确定当前的对话输入126匹配的主题。主题匹配装置204，被配置为基于主题转移监控装置202的指示，确定主题库212中与当前的对话输入126匹配的主题，可将匹配度最高的主题作为当前的对话输入126对应的主题，并将确定得出的主题发送给输出接口208。输出接口208通过网络110(如图1所示)在通信信号中发送确定得出的主题，以在如图1所示的第二服务器108处的对话输出系统118处被接收。
主题转移监控装置202，还被配置为当实体当前的对话输入126未转移对话主题时，指示主题匹配装置204确定当前的对话输入126匹配的主题，主题筛选装置206基于当前的对话输入126匹配的主题确定当前的对话输入126对应的主题。主题匹配装置204被配置为确定主题库212中与当前的对话输入126匹配的主题，并可将确定得出的与当前的对话输入126匹配的主题发送给主题筛选装置206。主题筛选装置206被配置为基于实体的特征，从与当前的对话输入126匹配的主题中筛选出与当前的对话输入126对应的主题。主题筛选装置206可将与当前的对话输入126对应的主题发送给输出接口208。输出接口208可通过网络110(如图1所示)在通信信号中发送与当前的对话输入126对应的主题，以在如图1所示的第二服务器108处的对话输出系统118处被接收。
在一个示例中，主题转移监控装置202可按照预先配置的主题转移规则210确定实体当前的对话输入126是否转移对话主题。主题转移规则210可以包括但不限于指示主题转移的关键词和/或关键句等，例如“换个话题”、“不想聊这个了”等等。在一些示例中，任何能够表示对话主题转移的词句均可用于主题转移规则210。
在一个示例中，主题库212可被配置为包括多个主题，多个主题中每一个主题可由一个或多个关键词和/或关键句构成，该一个或多个关键词和/或关键句可构成主题的主题词库。关键词之间可存在概念包含和被包含关系，可按照树状结构进行呈现，构成关键词树。每个主题可以对应一个或多个语料库，语料库可由包括对话输入和对话输出的对话对构成。
在一个示例中，主题匹配装置204可对实体当前的对话输入126进行分词处理，得到一个或多个用于进行主题匹配的对话词组。对于每个主题，主题匹配装置204可确定对话词组的至少部分词与主题词库中关键词的距离，得到对话词组与每个主题的相似度，以相似度来衡量实体当前的对话输入126与主题的匹配程度。在一个示例中，可将相似 度最高的主题作为实体当前的对话输入126对应的主题，或者将相似度高于预设值的一个或多个主题作为实体当前的对话输入126匹配的主题，或者将相似度高的预设个数的主题作为实体当前的对话输入126匹配的主题。
在一个示例中，主题匹配装置204可对对话词组进行筛选，例如可将名词作为进行主题匹配的词。主题匹配装置204还可确定对话词组各个词的权重，按照确定的权重来对对话词组中至少部分词与主题词库中的关键词的距离求和，得到实体当前的对话输入126与每个主题的相似性。
图3为主题匹配装置204一个示例的框图。如图3所示，主题匹配装置204可包括分词模块302、筛选模块304，以及相似性确定模块306。分词模块302可对对话输入进行分词处理，得到对话输入的对话词组。筛选模块304可对分词模块302分词得到的对话词组进行筛选，得到用于进行主题匹配的一个或多个词，其中，确定出的词可为对话词组中的部分或者全部。相似性确定模块306可确定筛选模块304筛选出的词与主题词库中的关键词的距离，进而基于确定出的距离得到筛选模块304筛选出的词与主题的相似度。
在一个示例中，主题匹配装置204还可包括权重确定模块308。权重确定模块308可确定筛选模块304筛选出的各个词的权重，例如，为动词和名词分配不同的权重，为实词和虚词分配不同的权重，对于实词中不同的词性可以分配不同的权重。相似性确定模块306可按照确定的权重来对筛选出的词与主题词库中的关键词的距离求和，得到对话输入与每个主题的相似性。在一个示例中，权重确定模块308可直接确定对此词组中各个词的权重，筛选模块304可不进行筛选。相似性确定模块306可按照确定出的权重对对话词组中每个词与主题库中的关键词的距离求和，得到对话输入与主题的相似度。
在一个示例中，主题库212可被配置为包括多个主题，多个主题中每一个主题可由一个或多个关键句构成，该一个或多个关键句可构成主题的主题词库。主题匹配装置204可确定对话输入与主题词库中的关键句的相似度，得到对话输入与主题的相似度。例如，主题匹配装置204可对对话输入和关键句进行分词，并计算词频，得到对话输入和关键句的词频向量，计算对话输入和关键句的词频向量的向量余弦值，进而得到对话输入与关键句的相似度。
图4为基于主题的交互方法一示例的流程图。
参考图4，该方法可应用在如图1、图2以及图3所示的环境中，但不限于此。该方法可包括步骤402至步骤404。响应于实体当前的对话输入，基于当前的对话输入与主题库中各个主题的匹配度，得到当前的对话输入匹配的主题(步骤402)。基于当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生当前的对话输入对应的对话输出(步骤404)。
在一个示例中，可基于实体的特征分析实体当前的对话输入匹配的主题，得到当前的对话输入对应的主题(步骤406)。其中，可基于当前的对话输入对应的主题相关联的包含至少一个对话对的语料库，产生当前的对话输入对应的对话输出。实体的特征可包括但不限于年龄、职业、所在地、学历、性别、兴趣、历史对话主题中至少之一或者任意组合。
在一个示例中，可响应于实体当前的对话输入，基于预先配置的主题转移规则判断所述当前的对话输入是否转移对话主题(步骤406)；以及，当确定当前的对话输入转移对话主题时，指示基于当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生当前的对话输入对应的对话输出(步骤402)。当确定当前的对话输入未转移对话主题时，可指示基于实体的特征分析实体当前的对话输入匹配的主题，得到当前的对话输入对应的主题(步骤406)。主题转移规则可被配置为基于一个或多个预设的指示主题转移的关键词确定对话输入是否转移对话主题，其中，在当前的对话输入中包含预设的指示主题转移的关键词时，确定当前的对话输入转移对话主题。
图5为基于主题的交互方法另一示例的流程图。
如图5所示，在该示例中以关键词来确定对话输入匹配/对应的主题。该方法可包括步骤502至步骤506。
在该示例中，通过用户接口接收用户的对话输入，对对话输入进行语音识别，得到对话输入的语音数据，例如，可以是在“听到”有声音的时候，识别声音所说的内容。接收语音数据，从语音数据中提取关键字(步骤502)。例如，识别出的语音数据为：我想做菜，但是不知道有什么好做的菜。对这句话进行分析，可以提取出其关键字为：做菜，或者菜。即，可以是对接收到的语音信息进行分解确认，确认出这个语音数据中的多个字段，从分解出的多个字段中找出有明确含义的实词，或者是找出有明确指向的词。例如，语音数据为：向我推荐几部好看的电影吧，其对应的关键字就可以是：电影。语音数据为：今天天气怎么样啊，其对应的关键字就可以是：天气。
通过对语音数据的解析，分析得到语句中的关键字，例如，关键字为电影，机器人就知道了，你要和我讨论的是与电影相关的内容，关键字为天气，机器人就知道了，要讨论的是与天气相关的内容，相应的就可以讲对话主题设定为电影，或者设定为天气。
在主题库中查找与关键字匹配的对话主题(步骤504)。在系统数据可中可以存储有一个主题库，这个主题库中可以预置多个对话主题，这些对话主题的选取可以是人为记录的，也可以是经验总结的，例如，可以是将人们在聊天过程中经常涉及或者谈论到的主题作为对话主题，也可以是将人们普遍关心的内容作为对话主题。这样在获取到关键字后，就可以在主题库中查找出与该关键字对应的对话主题。例如，关键字为天气，就可以在主题库中进行匹配的，为了提高匹配的有效性，主题可以是相同的词，也可以是相近的词，例如：天气，对应的主题可以是天气本身，也可以是气候等相近的词。
主题库中的主题也可以存在大库和小库，以“电影”为例，也许匹配出的大库的主题：影视剧，小库就是：电影，或者具体在后续的处理中，匹配出更小库的主题，例如具体到某一部电影。即，在处理的时候，通过主题的由宽泛到具体，也可以实现更为精确的对话主题的设定和匹配。
一个对话主题可对应一个语料库。在匹配到主题后，可以预先训练得到的与对话主题对应的语料库，作为人机交互的对话匹配素材进行人机交互(步骤506)。
在系统数据库中预置了一个庞大的数据库，这个数据库中存储有多个对话主题的语料库，即每个对话主题可以对应一个语料库。在一个示例中，语料库可以由多个问答对组成。因为每个对话主题都对应一个语料库，这样，在确定语音数据的对话主题后，就可以直接定位到该会话主题对应的语料库中进行数据匹配。因为选取的对话内容是基于这个确定的会话主题的，所以使得会话更为贴近实际的人与人之间的交互。
在一个示例中，可以在对话主题的语料库的问答对与问答对之间设置关联关系，这样，在机器人回答完人所提出的第一个问题后，还可以自己触发后续的对话，从而使得交流可以持续下去。例如，如图6所示，人问：我想做菜，但是不知道有什么好做的菜，机器人回答：酸辣土豆丝比较容易。在这之后，可以是问的一方主动发起后续问话：那你可以告诉我如何做吗？也可以是答的一方主要触发：需要我告诉你如何做吗？这种可以通过对问答对与问答对之间的关系设定来实现，例如，可以将具备语义上的关联关系的问答对进行关联，设定跳转条件，在满足跳转条件的情况下，可以跳至后续问答对或者关联问答对进行后续对话。其实，所谓的对话上的人机交互，也就是一个语音识别和对应语音输出的过程，即，可以是从确定的语料库中匹配出与所述语音数据对应的问答 对，并将问答对中的答复内容作为输出内容进行语音输出。
考虑到在实际交流的时候，存在对话主题跳转的情况，例如：可能当前还在讨论一部电影，突然谈到了这个电影中出现的某道菜比较好吃，这个时候对话人之间可能就开始围绕如何做这道菜，或者去哪可以吃这道菜进行讨论。在人机交互的过程中，如果遇到这种情况，仍旧在原本的对话主题(电影)对应的语料库中进行匹配，显然的得到的结果是不准确的，为了克服这种问题，可以设定对话主题跳转机制，即，在人机交互的过程中，时刻确定是否需要主题跳转，即判断当前的对话主题是否发生变化，如果发生变化，则以变化后的对话主题对应的语料库作为人机交互的对话匹配素材进行人机交互。触发这种对话主题变化的情况还是比较多的，下面以两种为例进行说明：
1)在当前的语料库中匹配不出与当前人机交互输入的语音数据对应的问答对时，则需要确定当前的对话主题是否发生变化；或者，
2)在当前接收到的语音数据与上一次接收到的语音数据间隔时间大于预定时间阈值时，则确定当前的对话主题是否发生变化。
对于上述第一种方式主要是考虑到如果一直在当前的对话主题对应的语料库中进行匹配，匹配不出合适的问答对，则就存在主题需要跳转的可能，这个时候就需要确认一个新的会话主题，以实现交流的持续性。对于上述第二种方式主要是考虑到如果对话的一方长时间为给予相应，也许对前一个话题的讨论已经完毕，在经过这段时间后，对方已经开始另外的主题讨论，这个时候可以实现主题的跳转，即确定当前接收到的语音数据的关键字，以确定该对话是否已分属于别的主题。
语料库内容的复杂性和完备性，往往对人机交互的准确性有着很重要的影响，考虑到现在互联网技术的迅速发展和云计算技术的发展，在训练得到对话主题对应的语料库的时候，可以采取从互联网上抓取页面，从中提取数据以进行训练的方式。具体的，可以是从互联网上抓取网页页面，然后以抓取的网页页面上的内容作为训练数据进行语料库的训练，其中，在训练的过程中，将同一网页页面上的内容确定为基于同一主题的内容。即，因为出现在同一页面上的内容往往是基于一个主题的，例如，到贴吧中参与讨论，或者在知道中参与问答，往往一个页面中所涉及的大方向的问题是基于一个主题，这样进行筛选和训练得到的内容或者是得到的问答对一般也是对应同一主题的。因为互联网数据是非常之多的，因此，通过这种取材训练的方式，可以极大的丰富训练库，使得得到的语料库更为完善和具体，也使得最终的人机交互更为贴近于人与人之间的真实交互。
在一个示例中，如果将上述的基于主题的人机交互引擎的实现方法应用到聊天机器人，那么这个机器人可以再与人进行对话的时候也实时进行训练，或者是“听到”周围人的对话的时候，也进行学习，从而使得训练得到的语料库更为完善和全面。
基于同一发明构思，本发明实施例中还提供了一种基于主题的人机交互引擎的实现装置，如下面的实施例所述。由于基于主题的人机交互引擎的实现装置解决问题的原理与基于主题的人机交互引擎的实现方法相似，因此基于主题的人机交互引擎的实现装置的实施可以参见基于主题的人机交互引擎的实现方法的实施，重复之处不再赘述。以下所使用的，术语“单元”或者“模块”可以实现预定功能的软件和/或硬件的组合。尽管以下实施例所描述的装置较佳地以软件来实现，但是硬件，或者软件和硬件的组合的实现也是可能并被构想的。图7为基于主题的交互系统的一种结构框图，如图7所示，包括：接收单元701、查找单元702和交互单元703，下面对该结构进行说明。
接收单元701，用于接收语音数据，并从所述语音数据中提取关键字；
查找单元702，与接收单元701相连，用于在主题库中查找出与所述关键字匹配的对话主题；
交互单元703，与查找单元702相连，用于以预先训练得到的与所述对话主题对应的语料库，作为人机交互的对话匹配素材进行人机交互，其中，一个对话主题对应一个语料库。
在一个示例中，语料库由多个问答对组成。
在一个实施方式中，交互单元703具体可以用于从语料库中匹配出与语音数据对应的问答对，并将问答对中的答复内容作为输出内容进行语音输出。
在一个实施方式中，如图8所示，上述基于主题的人机交互引擎的实现装置还可以包括：判断单元801，用于在以预先训练得到的与对话主题对应的语料库，作为人机交互的对话匹配素材进行人机交互的过程中，确定当前的对话主题是否发生变化；跳转单元802，用于在确定对话主题发生变化的情况下，以变化后的对话主题对应的语料库作为人机交互的对话匹配素材进行人机交互。
在一个实施方式中，判断单元801具体用于在当前的语料库中匹配不出与当前人机交互输入的语音数据对应的问答对时，确定当前的对话主题是否发生变化；或者，在当前接收到的语音数据与上一次接收到的语音数据间隔时间大于预定时间阈值时，确定当前的对话主题是否发生变化。
在一个实施方式中，上述基于主题的人机交互引擎的实现装置还可以包括训练单元， 具体用于从互联网上抓取网页页面；以抓取的网页页面上的内容作为训练数据进行语料库的训练，其中，在训练的过程中，将同一网页页面上的内容确定为基于同一主题的内容。
在另外一个实施例中，还提供了一种软件，该软件用于执行上述实施例及优选实施方式中描述的技术方案。
在另外一个实施例中，还提供了一种存储介质，该存储介质中存储有上述软件，该存储介质包括但不限于：光盘、软盘、硬盘、可擦写存储器等。
从以上的描述中，可以看出，本发明实施例实现了如下技术效果：交互时所依据的语料库是基于对话主题的，这与人们平时对话交流时候一般的会话都是基于某个主题的习惯是一致的，从而使得这种对话内容的匹配更为合理准确，有效解决了现有技术中对话内容匹配不是很准确，智能性低的技术问题，达到了有效提高人机交互智能性的技术效果。
显然，本领域的技术人员应该明白，上述的本发明实施例的各模块或各步骤可以用通用的计算装置来实现，它们可以集中在单个的计算装置上，或者分布在多个计算装置所组成的网络上，可选地，它们可以用计算装置可执行的程序代码来实现，从而，可以将它们存储在存储装置中由计算装置来执行，并且在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤，或者将它们分别制作成各个集成电路模块，或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。这样，本发明实施例不限制于任何特定的硬件和软件结合。
以上所述仅为本发明的优选实施例而已，并不用于限制本发明，对于本领域的技术人员来说，本发明实施例可以有各种更改和变化。凡在本发明的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。

Claims (23)

  1. 一种基于主题的交互系统，其特征在于，包括：
    主题匹配装置，用于响应于实体当前的对话输入，基于所述当前的对话输入与主题库中各个主题的匹配度，得到所述当前的对话输入匹配的主题；以及
    对话输出产生装置，用于基于所述当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生所述当前的对话输入对应的对话输出。
  2. 根据权利要求1所述的交互系统，其特征在于，所述系统还包括：主题筛选装置，用于基于实体的特征分析所述实体当前的对话输入匹配的主题，得到所述当前的对话输入对应的主题；
    其中，所述对话输出产生装置进一步用于基于所述当前的对话输入对应的主题相关联的包含至少一个对话对的语料库，产生所述当前的对话输入对应的对话输出。
  3. 根据权利要求2所述的交互系统，其特征在于，所述实体的特征包括年龄、职业、所在地、学历、性别、兴趣、历史对话主题中至少之一或者任意组合。
  4. 根据权利要求1至3中任一项所述的交互系统，其特征在于，所述系统还包括：
    主题转移监控装置，用于响应于所述实体当前的对话输入，基于预先配置的主题转移规则判断所述当前的对话输入是否转移对话主题，当确定所述当前的对话输入转移对话主题时，指示所述对话输出产生装置基于所述当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生所述当前的对话输入对应的对话输出。
  5. 根据权利要求4所述的交互系统，其特征在于，所述主题转移规则被配置为基于一个或多个预设的指示主题转移的关键词确定对话输入是否转移对话主题，其中，所述主题转移监控装置进一步用于在所述当前的对话输入中包含所述预设的指示主题转移的关键词时，确定所述当前的对话输入转移对话主题。
  6. 根据权利要求3所述的交互系统，其特征在于，所述实体的历史对话主题包括所述实体的相对于所述当前的对话输入的上一个对话输入对应的主题。
  7. 根据权利要求1至5中任一项所述的交互系统，其特征在于，所述系统还包括：历史数据记录装置，用于记录包含所述实体的历史对话主题在内的实体交互历史数据。
  8. 根据权利要求1至6中任一项所述的交互系统，其特征在于，对于所述主题库中的每个主题，所述每个主题被配置为包括至少一个表征主题的主题词和/或主题句；和/或，所述语料库包括至少一个对话对。
  9. 一种基于主题的交互方法，其特征在于，包括：
    响应于实体当前的对话输入，基于所述当前的对话输入与主题库中各个主题的匹配度，得到所述当前的对话输入匹配的主题；以及
    基于所述当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生所述当前的对话输入对应的对话输出。
  10. 根据权利要求9所述的方法，其特征在于，
    所述方法还包括：基于实体的特征分析所述实体当前的对话输入匹配的主题，得到所述当前的对话输入对应的主题；
    其中，基于所述当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生所述当前的对话输入对应的对话输出，包括：基于所述当前的对话输入对应的主题相关联的包含至少一个对话对的语料库，产生所述当前的对话输入对应的对话输出。
  11. 根据权利要求10所述的方法，其特征在于，所述实体的特征包括年龄、职业、所在地、学历、性别、兴趣、历史对话主题中至少之一或者任意组合。
  12. 根据权利要求9至11中任一项所述的方法，其特征在于，所述方法还包括：
    响应于所述实体当前的对话输入，基于预先配置的主题转移规则判断所述当前的对话输入是否转移对话主题；以及
    当确定所述当前的对话输入转移对话主题时，指示基于所述当前的对话输入匹配的主题相关联的包含至少一个对话对的语料库，产生所述当前的对话输入对应的对话输出。
  13. 根据权利要求12所述的方法，其特征在于，所述主题转移规则被配置为基于一个或多个预设的指示主题转移的关键词确定对话输入是否转移对话主题，其中，在所述当前的对话输入中包含所述预设的指示主题转移的关键词时，确定所述当前的对话输入转移对话主题。
  14. 一种基于主题的人机交互引擎的实现方法，其特征在于，包括：
    接收语音数据，从所述语音数据中提取关键字；
    在主题库中查找出与所述关键字匹配的对话主题；
    以预先训练得到的与所述对话主题对应的语料库，作为人机交互的对话匹配素材进行人机交互，其中，一个对话主题对应一个语料库。
  15. 根据权利要求14所述的方法，其特征在于，所述语料库由多个问答对组成。
  16. 根据权利要求15所述的方法，其特征在于，以预先训练得到的与所述对话主题对应的语料库，作为人机交互的对话匹配素材进行人机交互，包括：
    从所述语料库中匹配出与所述语音数据对应的问答对，并将问答对中的答复内容作为输出内容进行语音输出。
  17. 根据权利要求14所述的方法，其特征在于，在以预先训练得到的与所述对话主题对应的语料库，作为人机交互的对话匹配素材进行人机交互的过程中，还包括：
    确定当前的对话主题是否发生变化；
    在确定对话主题发生变化的情况下，以变化后的对话主题对应的语料库作为人机交互的对话匹配素材进行人机交互。
  18. 根据权利要求17所述的方法，其特征在于，确定当前的对话主题是否发生变化，包括：
    在当前的语料库中匹配不出与当前人机交互输入的语音数据对应的问答对时，则确定当前的对话主题是否发生变化；或者，
    在当前接收到的语音数据与上一次接收到的语音数据间隔时间大于预定时间阈值时，则确定当前的对话主题是否发生变化。
  19. 根据权利要求14至18中任一项所述的方法，其特征在于，与所述对话主题对应的语料库是按照以下方式训练得到的：
    从互联网上抓取网页页面；
    以抓取的网页页面上的内容作为训练数据进行语料库的训练，其中，在训练的过程中，将同一网页页面上的内容确定为基于同一主题的内容。
  20. 一种基于主题的人机交互引擎的实现装置，其特征在于，包括：
    接收单元，用于接收语音数据，并从所述语音数据中提取关键字；
    查找单元，用于在主题库中查找出与所述关键字匹配的对话主题；
    交互单元，用于以预先训练得到的与所述对话主题对应的语料库，作为人机交互的对话匹配素材进行人机交互，其中，一个对话主题对应一个语料库。
  21. 根据权利要求20所述的装置，其特征在于，所述语料库由多个问答对组成。
  22. 根据权利要求21所述的装置，其特征在于，所述交互单元具体用于从所述语料库中匹配出与所述语音数据对应的问答对，并将问答对中的答复内容作为输出内容进行语音输出。
  23. 根据权利要求20所述的装置，其特征在于，还包括：
    判断单元，用于在以预先训练得到的与所述对话主题对应的语料库，作为人机交互的对话匹配素材进行人机交互的过程中，确定当前的对话主题是否发生变化；
    跳转单元，用于在确定对话主题发生变化的情况下，以变化后的对话主题对应的语料库作为人机交互的对话匹配素材进行人机交互。
PCT/CN2016/076137 2015-04-30 2016-03-11 基于主题的交互系统及方法 WO2016173326A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201510212167 2015-04-30
CN201510212167.9 2015-04-30

Publications (1)

Publication Number Publication Date
WO2016173326A1 true WO2016173326A1 (zh) 2016-11-03

Family

ID=57199007

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/076137 WO2016173326A1 (zh) 2015-04-30 2016-03-11 基于主题的交互系统及方法

Country Status (1)

Country Link
WO (1) WO2016173326A1 (zh)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108920639A (zh) * 2018-07-02 2018-11-30 北京百度网讯科技有限公司 基于语音交互的上下文获取方法及设备
CN108986825A (zh) * 2018-07-02 2018-12-11 北京百度网讯科技有限公司 基于语音交互的上下文获取方法及设备
CN111062213A (zh) * 2019-11-19 2020-04-24 竹间智能科技（上海）有限公司 命名实体识别方法、装置、设备及介质
CN111191034A (zh) * 2019-12-30 2020-05-22 科大讯飞股份有限公司 人机交互方法、相关设备及可读存储介质
CN111198823A (zh) * 2020-01-10 2020-05-26 北京声智科技有限公司 一种多轮会话的测试方法、装置、设备和介质
US10943070B2 (en) 2019-02-01 2021-03-09 International Business Machines Corporation Interactively building a topic model employing semantic similarity in a spoken dialog system
WO2023011296A1 (zh) * 2021-08-04 2023-02-09 北京字跳网络技术有限公司 交互方法、电子设备、存储介质和程序产品

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1953056A (zh) * 2005-10-21 2007-04-25 阿鲁策株式会社 会话控制器
US20070094004A1 (en) * 2005-10-21 2007-04-26 Aruze Corp. Conversation controller
CN101075435A (zh) * 2007-04-19 2007-11-21 深圳先进技术研究院 一种智能聊天系统及其实现方法
CN104350541A (zh) * 2012-04-04 2015-02-11 奥尔德巴伦机器人公司 能够将与用户的自然对话合并到其行为中的机器人,以及编程和使用所述机器人的方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1953056A (zh) * 2005-10-21 2007-04-25 阿鲁策株式会社 会话控制器
US20070094004A1 (en) * 2005-10-21 2007-04-26 Aruze Corp. Conversation controller
CN101075435A (zh) * 2007-04-19 2007-11-21 深圳先进技术研究院 一种智能聊天系统及其实现方法
CN104350541A (zh) * 2012-04-04 2015-02-11 奥尔德巴伦机器人公司 能够将与用户的自然对话合并到其行为中的机器人,以及编程和使用所述机器人的方法

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108920639A (zh) * 2018-07-02 2018-11-30 北京百度网讯科技有限公司 基于语音交互的上下文获取方法及设备
CN108986825A (zh) * 2018-07-02 2018-12-11 北京百度网讯科技有限公司 基于语音交互的上下文获取方法及设备
CN108920639B (zh) * 2018-07-02 2022-01-18 北京百度网讯科技有限公司 基于语音交互的上下文获取方法及设备
US10943070B2 (en) 2019-02-01 2021-03-09 International Business Machines Corporation Interactively building a topic model employing semantic similarity in a spoken dialog system
CN111062213A (zh) * 2019-11-19 2020-04-24 竹间智能科技（上海）有限公司 命名实体识别方法、装置、设备及介质
CN111062213B (zh) * 2019-11-19 2024-01-12 竹间智能科技（上海）有限公司 命名实体识别方法、装置、设备及介质
CN111191034A (zh) * 2019-12-30 2020-05-22 科大讯飞股份有限公司 人机交互方法、相关设备及可读存储介质
CN111191034B (zh) * 2019-12-30 2023-01-17 科大讯飞股份有限公司 人机交互方法、相关设备及可读存储介质
CN111198823A (zh) * 2020-01-10 2020-05-26 北京声智科技有限公司 一种多轮会话的测试方法、装置、设备和介质
WO2023011296A1 (zh) * 2021-08-04 2023-02-09 北京字跳网络技术有限公司 交互方法、电子设备、存储介质和程序产品

Similar Documents

Publication Publication Date Title
TWI732271B (zh) 人機對話方法、裝置、電子設備及電腦可讀媒體
WO2016173326A1 (zh) 基于主题的交互系统及方法
US10740564B2 (en) Dialog generation method, apparatus, and device, and storage medium
US11599729B2 (en) Method and apparatus for intelligent automated chatting
US10311895B2 (en) Assessing the structural quality of conversations
JP2021018797A (ja) 対話の交互方法、装置、コンピュータ可読記憶媒体、及びプログラム
US11729120B2 (en) Generating responses in automated chatting
JP2022551788A (ja) 補助システムのためのプロアクティブコンテンツを生成すること
US9483582B2 (en) Identification and verification of factual assertions in natural language
JP2019102063A (ja) ページ制御方法および装置
RU2711153C2 (ru) Способы и электронные устройства для определения намерения, связанного с произнесенным высказыванием пользователя
WO2019174450A1 (zh) 一种对话生成的方法和装置
US11233756B2 (en) Voice forwarding in automated chatting
RU2711104C2 (ru) Способ и компьютерное устройство для определения намерения, связанного с запросом для создания зависящего от намерения ответа
US11043215B2 (en) Method and system for generating textual representation of user spoken utterance
US20230088445A1 (en) Conversational recommendation method, method of training model, device and medium
US10770072B2 (en) Cognitive triggering of human interaction strategies to facilitate collaboration, productivity, and learning
CN111832308A (zh) 语音识别文本连贯性处理方法和装置
US11115410B1 (en) Secure authentication for assistant systems
Shen et al. Kwickchat: A multi-turn dialogue system for aac using context-aware sentence generation by bag-of-keywords
WO2022271331A1 (en) Machine-learning-model based name pronunciation
CN114064943A (zh) 会议管理方法、装置、存储介质及电子设备
CN110895656A (zh) 一种文本相似度计算方法、装置、电子设备及存储介质
KR102222637B1 (ko) 감성 분석 장치, 이를 포함하는 대화형 에이전트 시스템, 감성 분석을 수행하기 위한 단말 장치 및 감성 분석 방법
KR102226427B1 (ko) 호칭 결정 장치, 이를 포함하는 대화 서비스 제공 시스템, 호칭 결정을 위한 단말 장치 및 호칭 결정 방법

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16785757

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 16785757

Country of ref document: EP

Kind code of ref document: A1