WO2017041370A1 - 基于人工智能的人机聊天方法和装置 - Google Patents

基于人工智能的人机聊天方法和装置 Download PDF

Info

Publication number
WO2017041370A1
WO2017041370A1 PCT/CN2015/096342 CN2015096342W WO2017041370A1 WO 2017041370 A1 WO2017041370 A1 WO 2017041370A1 CN 2015096342 W CN2015096342 W CN 2015096342W WO 2017041370 A1 WO2017041370 A1 WO 2017041370A1
Authority
WO
WIPO (PCT)
Prior art keywords
chat
module
user
input information
information
Prior art date
Application number
PCT/CN2015/096342
Other languages
English (en)
French (fr)
Inventor
赵世奇
亓超
吴华
忻舟
�田�浩
周湘阳
陈洪亮
温泉
张晓庆
许心诺
严睿
Original Assignee
百度在线网络技术(北京)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 百度在线网络技术(北京)有限公司 filed Critical 百度在线网络技术(北京)有限公司
Publication of WO2017041370A1 publication Critical patent/WO2017041370A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/903Querying
    • G06F16/90335Query processing
    • G06F16/90348Query processing by searching ordered data, e.g. alpha-numerically ordered data
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/953Querying, e.g. by the use of web search engines
    • G06F16/9535Search customisation based on user profiles and personalisation

Definitions

  • the invention relates to the field of artificial intelligence technology, in particular to a human-machine chat method and device based on artificial intelligence.
  • AI Artificial Intelligence is a branch of computer science, abbreviated as AI. It is a new technical science that studies and develops theories, methods, techniques and application systems for simulating, extending and extending human intelligence.
  • Human-machine chat refers to the process of interactive chat between a person and a computer. Human-machine chat can be applied to entertainment and emotional companionship, intelligent service anthropomorphization and so on. For example, through the human-machine chat system, conversations can be made anytime and anywhere, alleviating people's life stress, and children can also help children improve their language skills.
  • each chat sentence pair contains a chat sentence P (post) and a P sentence for the next sentence R (Response).
  • chat sentence Q query
  • the traditional human-machine chat system has the following shortcomings: lack of multiple rounds of chat ability, that is, the user puts up a chat sentence, the machine replies to the chat sentence, lacks initiative, is not real and natural; for time-sensitive dialogues such as news, It is impossible to answer accurately; in addition, the lack of personalization of the chat content, the same or similar questions raised by different users, can only respond to the same answer, lack of personalization and intelligence.
  • an object of the present invention is to provide a human-machine chat method based on artificial intelligence, which can perform multiple rounds of chats with a user, which is real and natural, and has the initiative to return a reply corresponding to the user's style for different users, and more personalized. And intelligent.
  • a second object of the present invention is to provide a human-machine chat device based on artificial intelligence.
  • the first aspect of the present invention provides a human-machine chat method based on artificial intelligence.
  • the method includes: receiving input information input by a user; distributing the input information to a chat service module; receiving a candidate reply returned by the plurality of chat service modules, wherein the candidate reply has a corresponding confidence; based on the confidence Sorting the to-be-selected responses, and generating chat information according to the sorting result, and providing the chat information to the user.
  • the artificial intelligence-based human-machine chat method of the embodiment of the present invention receives the input information input by the user, and distributes the input information to the chat service module, and then receives the candidate reply returned by the plurality of chat service modules, and selects the candidate based on the confidence degree.
  • the reply is sorted, and the chat information is generated according to the sorting result, and the chat information is provided to the user, and the user can perform multiple rounds of chat, which is real and natural, and has the initiative to return a reply corresponding to the user style for different users, and is more personalized, Intelligent.
  • the second aspect of the present invention provides a human-machine chat device based on artificial intelligence, comprising: a first receiving module, configured to receive input information input by a user; and a distribution module, configured to distribute the input information to a chat service a second receiving module, configured to receive a candidate reply returned by the multiple chat service modules, where the candidate reply has a corresponding confidence level; and a providing module, configured to send the candidate reply based on the confidence level Sorting is performed, and chat information is generated according to the sorting result, and the chat information is provided to the user.
  • the artificial intelligence-based human-machine chat device of the embodiment of the present invention receives the input information input by the user, distributes the input information to the chat service module, and then receives the candidate reply returned by the plurality of chat service modules, and selects the candidate response based on the confidence level.
  • the reply is sorted, and the chat information is generated according to the sorting result, and the chat information is provided to the user, and the user can perform multiple rounds of chat, which is real and natural, and has the initiative to return a reply corresponding to the user style for different users, and is more personalized, Intelligent.
  • a third aspect of the embodiments of the present invention discloses a storage medium for storing an application program, where the application program is used to execute the artificial intelligence based human-machine chat method according to the first aspect of the present invention.
  • a fourth aspect of the embodiments of the present invention discloses an apparatus, including: one or more processors; a memory; one or more modules, the one or more modules being stored in the memory when When multiple processors are executing, do the following:
  • FIG. 1 is a flow chart of an artificial intelligence based human-machine chat method according to an embodiment of the present invention.
  • FIG. 2 is a schematic diagram of the effect of a topic chat map in accordance with one embodiment of the present invention.
  • FIG. 3 is a schematic structural diagram 1 of a human-machine chat device based on artificial intelligence according to an embodiment of the present invention.
  • FIG. 4 is a second schematic structural diagram of a human-machine-based chat device based on artificial intelligence according to an embodiment of the present invention.
  • FIG. 5 is a schematic structural diagram of a search-based chat module according to an embodiment of the present invention.
  • FIG. 6 is a schematic structural diagram of a rich knowledge chat module according to an embodiment of the present invention.
  • FIG. 7 is a schematic structural diagram 1 of a portrait-based chat module according to an embodiment of the present invention.
  • FIG. 8 is a second schematic structural diagram of a portrait-based chat module according to an embodiment of the present invention.
  • FIG. 9 is a third schematic structural diagram of a human-machine chat device based on artificial intelligence according to an embodiment of the present invention.
  • FIG. 10 is a fourth structural diagram of a human-machine-based chat device based on artificial intelligence according to an embodiment of the present invention.
  • FIG. 11 is a schematic structural diagram 5 of a human-machine chat device based on artificial intelligence according to an embodiment of the present invention.
  • FIG. 12 is a schematic structural diagram 6 of a human-machine chat apparatus based on artificial intelligence according to an embodiment of the present invention.
  • FIG. 1 is a flow chart of an artificial intelligence based human-machine chat method according to an embodiment of the present invention.
  • the artificial intelligence based human-machine chat method may include:
  • S1. Receive input information input by the user.
  • the input information may be voice information or text information.
  • the input information After receiving the input information input by the user, the input information may be corrected and/or rewritten, used to correct the typos in the input information, and to rewrite the irregular colloquial expression.
  • the above information that is chatted with the user may also be obtained, and then according to the above information, it is determined whether the dependency of the input information and the above information is greater than a preset relationship threshold. If it is greater than the preset relationship threshold, the input information can be complemented according to the above information, thereby ensuring the smoothness of the human-machine chat.
  • complementing the input information may include referring to the digestion. For example, if the input information is "Has he got married?", he can replace "He” in the input information with "Andy Lau” according to the above information "Andy Lau”.
  • Completing the input information may also include omitting the completion. For example, the above information "Andy Lau's wife is called Zhu Liqian.”, the input information is "I don't know.”, then the input information can be completed as "I don't know Zhu Liqian.”
  • the current topic information of the user can be obtained, so that the subsequent chat service module can chat with the chat message.
  • the problem is guided.
  • domain analysis may be performed on the input information to obtain an area corresponding to the input information.
  • the input information can then be distributed to chat service modules having the same or similar fields based on the fields corresponding to the input information.
  • the chat service module may include one or more of a search-based chat module, a rich knowledge chat module, a portrait-based chat module, and a crowdsourcing-based chat module.
  • the search-based chat module may perform a word-cutting on the input information to generate a plurality of chat phrases, and then query the chat corpus according to the plurality of chat phrases to generate a plurality of chat corpus sentences and multiple chat corpus upper sentences.
  • Multiple chat corpora The chat corpus is pre-established, and the source of the chat can include, but is not limited to, “posting-replying” in the forum data such as posting bar, “blog-reply” in the microblog, and “question-answer” in the question and answer community.
  • chat corpus sentences can be filtered. Specifically, the similarity between the input information and the sentences of the plurality of chat corpora can be calculated. If the similarity is less than the first preset similarity threshold, the corresponding chat corpus can be filtered; if the similarity is greater than or equal to the first preset similarity threshold, the corresponding chat corpus can be retained.
  • the chat corpus corresponding to the sentence in the chat corpus after filtering may be classified. Specifically, the similarity between the input information and the plurality of chat corpus sentences is calculated, and the machine based on the similarity is based on GBDT (Gradient Boost Decision Tree), SVM (Support Vector Machine), and the like.
  • the learning model classifies multiple chat corpora sentences.
  • the similarity between the input information and the sentences of the plurality of chat corpora may be a literal similarity between the input information and the sentence of the chat corpus, or may be similar to the input information and the chat corpus based on the deep neural network training.
  • Degree can also be the similarity between the input information and the chat corpus based on the machine translation model. It should be understood that the similarity between the input information and the plurality of chat corpus in the present embodiment and the machine learning model such as GBDT and SVM are well-known technologies, and are not described herein.
  • the search-based chat module can reorder the chat corpus under the classification, and generate a candidate reply according to the sort result.
  • the chat attribute of the user may be obtained according to the above information of the user chat, and then the chat quotation sentence after the classification is reordered based on the learning order model (Learning-To-Rank) according to the chat attribute.
  • the chat attribute may include a chat occasion such as a time and place, a chat fun, a chat style, and the like.
  • the chat attribute is not limited to the above information from the user chat, but also can be obtained according to the user's long-term historical chat record. It should be understood that the learning ordering model in this embodiment is a well-known technology, and details are not described herein.
  • the rich knowledge chat module may generate a search word according to the input information, and perform a search according to the search word to generate a plurality of search results, and then perform sentence extraction on the plurality of search results to obtain a similarity with the search word greater than the second preset similarity.
  • a set of candidate sentences for a sentence of a threshold After that, the sentences in the set of candidate sentences can be rewritten to generate candidate responses.
  • sentences in the candidate sentence set may be reordered according to the user's chat attribute. For example, the input information is “I hope to have the opportunity to travel to Mount Fuji”, and the input information can be parsed and the corresponding search words generated. "Mt.
  • Fuji, Tourism then obtain multiple search results based on the search term, and extract sentences with high similarity to the search term.
  • some sentences may include obvious texts such as “journalists”, so these sentences need to be rewritten to make them more fluid, more like the sentences of natural language chat, and the final candidate for the response is “Mt. Fuji due to Due to the weather, there is only a certain period of summer in the year to climb the mountain. Compared with the traditional reply, "I also want to go to Mount Fuji, let's go together.”, with certain knowledge and timeliness, can enable users to Get useful knowledge during the chat.
  • the human-machine chat system can set its own attributes, status, interests, etc., that is, the system portrait model. It is also possible to set the user's attributes, status, interests, etc., that is, the user portrait model.
  • the system portrait model used may be the same, or a system image model corresponding to each user may be set. Both the system portrait model and the user portrait model are based on the image knowledge map.
  • the portrait knowledge map is a hierarchical knowledge system. For example, the "family member" node may include two child nodes "brothers" and "parents", and the "parent" child nodes include two child nodes "father” and "mother”.
  • Each node corresponds to a plurality of input information template clusters, such as "Who is your father”, “Who is your father”, and "What is your father's name” belongs to the same input information template cluster.
  • Each input information template cluster corresponds to one or more candidate responses.
  • the input information template cluster and the candidate reply may include variables, such as interest, hobbies, and preferences corresponding to the same attribute "INTEREST", and the attribute values of "INTEREST” may include climbing, music, reading, sports, and the like.
  • the portrait-based chat module may acquire the chat context of the user, and determine whether the collection condition is satisfied according to the chat context. If it is determined that the collection conditions are met, the question can be sent to the user. After that, the user's answer information according to the question can be received, and the user portrait model is updated according to the answer information. For example, when talking to a user about a movie-related topic, you can send the question "What movie do you like?" or the user asks the human-machine chat system "What do you like to eat?", the human-machine chat system can ask the user "You like to eat What? After the user answers, the user portrait model can be updated based on the user's answer information, which is more in line with the user's personalized needs.
  • the portrait-based chat module may also acquire the chat content of the user, extract user image data according to the chat content, and then update the user portrait model according to the extracted user portrait data.
  • the user said in the chat process, "I like to climb the mountain and catch fishing when I am fine.”
  • I can extract the user portrait data "hobby climbing, hobby fishing" to update the user portrait model.
  • an appropriate answer can be extracted based on the user's portrait data, and the appropriate answer information can be returned to the user.
  • Crowdsourcing is a method of outsourcing specific tasks to non-specific users on the Internet.
  • human-machine chat the problem that the machine is difficult to answer can be distributed to the performer to manually respond in real time to meet the user's needs. Actual demand.
  • the crowdsourcing-based chat module can determine whether the input information is suitable for crowdsourcing completion, for example, if the user's mood is low and needs comfort, etc., it is suitable for crowdsourcing completion.
  • the user's input information includes personally identifiable information, password, and telephone number. Such as private information, it is not suitable for crowdsourcing.
  • the input information can be distributed to the corresponding performer.
  • the above information can also be sent to the performer together, and the performer can reply according to the above information and input information.
  • the crowd-based chat module can then receive the reply information of the performer and make a quality judgment on the reply information. If the quality requirement is met, the reply message is used as a candidate reply. For example, if the reply message contains vulgar, reactionary, or pornographic content, the quality is not enough. Or if the executor replies for more than the predetermined duration, the replies of the executor will not be used, and the reply information can be saved to the chat corpus.
  • the topic chat map is a directed graph with a topic as a node. For example, as shown in FIG.
  • the node "leisure” can point to the node “watching movie” and the node “listening to the song", indicating that the topic "leisure” can be guided to the topic "watching movie” or the topic “listening to the song”.
  • the topic "watching movies” and the topic “listening songs” all have a certain guiding probability, and the guiding of the topic can be realized according to the guiding probability, thereby ensuring the diversity of guiding topics.
  • a candidate reply can then be generated based on the guided topic.
  • the template of the candidate reply may be generated based on the Natural Language Generation, and the guide topic may be filled into the template to generate a candidate reply; or the sentence including the guide topic may be selected as a candidate reply, thereby implementing the user.
  • Actively chat topic guide may be generated based on the Natural Language Generation.
  • the candidate reply has a corresponding confidence.
  • the characteristics of the user's input information may be acquired, and the selected responses are sorted based on the characteristics and confidence of the input information.
  • the characteristics of the input information may include classification features, literal features, topic features, and the like. The higher the confidence, the better the quality of the candidate reply is.
  • the responses can be sorted according to the order of confidence, and finally the chat information that meets the user's needs is provided to the user.
  • the Reinforcement Learning can also be updated according to the feedback information of the user, so that the user can be provided with more satisfactory chat information. For example, adding a comment button such as “Like” or “Tread” to the user's chat information to collect feedback information of the user; or analyzing the input information of the user in the chat based on the sentiment analysis technology, thereby obtaining the user's evaluation, For example: "You are really smart", etc.; or judge the user's satisfaction by recording the number of interactions with the user.
  • a comment button such as “Like” or “Tread”
  • the artificial intelligence-based human-machine chat method of the embodiment of the present invention receives the input information input by the user, and distributes the input information to the chat service module, and then receives the candidate reply returned by the plurality of chat service modules, and is based on The reliability is sorted by the selected responses, and the chat information is generated according to the sorting result, and the chat information is provided to the user, and the user can perform multiple rounds of chats, which is real and natural, and has the initiative to return a reply corresponding to the user style for different users. More personalized and intelligent.
  • the present invention also provides a human-machine chat device based on artificial intelligence.
  • FIG. 3 is a schematic structural diagram 1 of a human-machine chat device based on artificial intelligence according to an embodiment of the present invention.
  • the artificial intelligence-based human-machine chat device may include: a first receiving module 1000, a distribution module 2000, a chat service module 3000, a second receiving module 4000, and a providing module 5000.
  • the first receiving module 1000 is configured to receive input information input by a user.
  • the input information may be voice information or text information.
  • the distribution module 2000 is for distributing input information to the chat service module 3000.
  • the second receiving module 4000 is configured to receive candidate responses returned by the plurality of chat service modules 3000. Among them, the candidate reply has a corresponding confidence.
  • the providing module 5000 is configured to sort the responses according to the confidence degree, generate chat information according to the sorting result, and provide the chat information to the user.
  • the providing module 5000 may acquire features of the user's input information, and sort the responses to be selected based on the characteristics and confidence of the input information.
  • the characteristics of the input information may include classification features, literal features, topic features, and the like. The higher the confidence, the better the quality of the candidate reply is.
  • the responses can be sorted according to the order of confidence, and finally the chat information that meets the user's needs is provided to the user.
  • the chat service module 3000 can include a search-based chat module 3100, a rich knowledge chat module 3200, a portrait-based chat module 3300, and a crowdsourcing-based chat module 3400.
  • the search-based chat module 3100 may include a word-cutting sub-module 3110, a generating sub-module 3120, a filtering sub-module 3130, a sorting sub-module 3140, and a sorting sub-module 3150.
  • the filtering sub-module 3130 may include a calculating unit 3131, a filtering unit 3132, and a retaining unit 3133.
  • the sorting sub-module 3140 may include a calculating unit 3141, a sorting unit 3142, and the sorting sub-module 3150 may include an obtaining unit 3151 and a sorting unit 3152.
  • the word-cutting sub-module 3110 can cut the input information to generate a plurality of chat phrases, and then the generating sub-module 3120 can query the chat corpus according to the plurality of chat phrases to generate a plurality of chat corpus and multiple chats.
  • the corpus corresponds to multiple chat corpora sentences.
  • the chat corpus is pre-established, and the source of the chat can include, but is not limited to, “posting-replying” in the forum data such as posting bar, “blog-reply” in the microblog, and “question-answer” in the question and answer community.
  • the filtering sub-module 3130 can filter a plurality of chat corpus sentences.
  • the calculation unit 3131 can calculate the similarity between the input information and the sentences of the plurality of chat corpora. If the similarity is less than the first preset similarity threshold, the filtering unit 3132 may filter the corresponding chat corpus upper sentence; if the similarity is greater than or equal to the first preset similarity threshold The value, the retention unit 3133 may retain the corresponding chat corpus.
  • the categorization sub-module 3140 may classify the chat corpus corresponding to the sentence in the chat corpus after the filtering.
  • the calculating unit 3141 may calculate the similarity between the input information and the plurality of chat corpus sentences, and the classifying unit 3142 is based on the similarity based on the GBDT (Gradient Boost Decision Tree), SVM (Support Vector Machine, A machine learning model such as Support Vector Machine classifies multiple chat corpora sentences.
  • the similarity between the input information and the sentences of the plurality of chat corpora may be a literal similarity between the input information and the sentence of the chat corpus, or may be similar to the input information and the chat corpus based on the deep neural network training.
  • Degree can also be the similarity between the input information and the chat corpus based on the machine translation model. It should be understood that the similarity between the input information and the plurality of chat corpus in the present embodiment and the machine learning model such as GBDT and SVM are well-known technologies, and are not described herein.
  • the sorting sub-module 3150 can then reorder the chat corpus in the classified category and generate a candidate reply according to the sorting result.
  • the obtaining unit 3151 may acquire the chat attribute of the user according to the above information of the user chat, and the sorting unit 3152 reorders the chat corpus after the classification according to the learning ordering model (Learning-To-Rank) according to the chat attribute.
  • the chat attribute may include a chat occasion such as a time and place, a chat fun, a chat style, and the like.
  • the chat attribute is not limited to the above information from the user chat, but also can be obtained according to the user's long-term historical chat record. It should be understood that the learning ordering model in this embodiment is a well-known technology, and details are not described herein.
  • the rich knowledge chat module 3200 can include a generating sub-module 3210, an extracting sub-module 3220, a rewriting sub-module 3230, and a re-sorting sub-module 3240.
  • the generating sub-module 3210 may generate a search word according to the input information, and perform a search according to the search word to generate a plurality of search results, and then the extracting sub-module 3220 performs sentence extraction on the plurality of search results to obtain similarity with the search word.
  • the rewrite sub-module 3230 can rewrite the sentences in the set of candidate sentences to generate a candidate reply.
  • the reordering sub-module 3240 can reorder the sentences in the candidate sentence set according to the chat attribute of the user. For example, if the input information is "I hope to have the opportunity to travel to Mount Fuji", the input information can be parsed and the corresponding search term "Mount Fuji, Tourism" can be generated, and then multiple search results can be obtained according to the search term, and the search words can be extracted and searched. A sentence with high similarity. Among them, some sentences may include obvious texts such as “journalists”, so these sentences need to be rewritten to make them more fluid, more like the sentences of natural language chat, and the final candidate for the response is “Mt. Fuji due to Due to the weather, there is only a certain period of summer in the year to climb the mountain. Compared with the traditional reply, "I also want to go to Mount Fuji, let's go together.”, with certain knowledge and timeliness, can enable users to Get useful knowledge during the chat.
  • the portrait-based chat module 3300 may include a first acquisition sub-module 3310, a determination sub-module 3320, a transmission sub-module 3330, and a first update sub-module 3340.
  • the human-machine chat system can set its own genus. Sex, state, interest, etc., that is, the system portrait model. It is also possible to set the user's attributes, status, interests, etc., that is, the user portrait model. Of course, in the case of facing different users, the system portrait model used may be the same, or a system image model corresponding to each user may be set. Both the system portrait model and the user portrait model are based on the image knowledge map.
  • the portrait knowledge map is a hierarchical knowledge system. For example, the "family member" node may include two child nodes "brothers" and "parents", and the "parent" child nodes include two child nodes "father” and "mother”.
  • Each node corresponds to a plurality of input information template clusters, such as "Who is your father”, “Who is your father”, and "What is your father's name” belongs to the same input information template cluster.
  • Each input information template cluster corresponds to one or more candidate responses.
  • the input information template cluster and the candidate reply may include variables, such as interest, hobbies, and preferences corresponding to the same attribute "INTEREST", and the attribute values of "INTEREST” may include climbing, music, reading, sports, and the like.
  • the first obtaining sub-module 3310 can obtain the chat context of the user, and the determining sub-module 3320 determines whether the collection condition is satisfied according to the chat context. If it is determined that the collection condition is met, the transmitting sub-module 3330 can send a question to the user. After that, the first update sub-module 3340 can receive the user's answer information according to the question, and update the user portrait model according to the answer information. For example, when talking to a user about a movie-related topic, you can send the question "What movie do you like?" or the user asks the human-machine chat system "What do you like to eat?", the human-machine chat system can ask the user "You like to eat What? After the user answers, the user portrait model can be updated based on the user's answer information, which is more in line with the user's personalized needs.
  • the portrait-based chat module 3300 may further include a second acquisition sub-module 3350, an extraction sub-module 3360, and a second update sub-module 3370.
  • the second obtaining sub-module 3350 can acquire the chat content of the user, and the extracting sub-module 3360 extracts the user portrait data according to the chat content, and then the second update sub-module 3370 updates the user portrait model according to the extracted user portrait data.
  • the user said in the chat process, "I like to climb the mountain and catch fishing when I am fine.”
  • I can extract the user portrait data "hobby climbing, hobby fishing" to update the user portrait model.
  • an appropriate answer can be extracted based on the user's portrait data, and the appropriate answer information can be returned to the user.
  • Crowdsourcing is a method of outsourcing specific tasks to non-specific users on the Internet.
  • human-machine chat the problem that the machine is difficult to answer can be distributed to the performer to manually respond in real time to meet the user's needs. Actual demand.
  • the crowdsourcing-based chat module 3400 can determine whether the input information is suitable for crowdsourcing completion, for example, if the user's mood is low and needs comfort, etc., it is suitable for crowdsourcing completion. For example, if the user's input information contains personal information such as personal identification information, password, and telephone, it is not suitable for crowdsourcing.
  • the crowdsourcing based chat module 3400 can distribute the input information to the corresponding performer.
  • the above information can also be sent to the performer together, and the performer can reply according to the above information and input information.
  • the crowd-based chat module can then receive the reply information of the performer and make a quality judgment on the reply information. If the quality requirement is met, the reply message is used as a candidate reply. For example, if the reply message contains vulgar, reactionary, or pornographic content, the quality is not enough. Or if the executor replies for more than the predetermined duration, the replies of the executor will not be used, and the reply information can be saved to the chat corpus.
  • the artificial intelligence based human-machine chat device may further include an error correction module 6000.
  • the error correction module 6000 is configured to perform error correction and/or rewriting of the input information after receiving the input information input by the user, for correcting the typos in the input information, rewriting the irregular colloquial expression, and the like.
  • the artificial intelligence based human-machine chat device may further include an analysis module 7000.
  • the analyzing module 7000 is configured to perform domain analysis on the input information to obtain an area corresponding to the input information after receiving the input information input by the user, and then the distribution module 2000 may distribute the input information to have the same or similar fields according to the domain corresponding to the input information. Chat service module.
  • the artificial intelligence-based human-machine chat device may further include a first acquiring module 8000, a first determining module 9000, and a completion module 10000.
  • the first obtaining module 8000 is configured to acquire the above information that is chatted with the user after receiving the input information input by the user, and obtain the current topic information of the user according to the above information. Then, the first determining module 9000 can determine, according to the above information, whether the dependency relationship between the input information and the above information is greater than a preset relationship threshold. When the dependency relationship is greater than the preset relationship threshold, the completion module 10000 can complete the input information according to the above information, thereby ensuring the smoothness of the human-machine chat. Specifically, complementing the input information may include referring to the digestion.
  • Completing the input information may also include omitting the completion.
  • the above information "Andy Lau's wife is called Zhu Liqian.”
  • the input information is "I don't know.”
  • the input information can be completed as "I don't know Zhu Liqian.”
  • the artificial intelligence based human-machine chat device may further include a second determining module 11000, a second obtaining module 12000, a first generating module 13000, and a second generating module 14000.
  • the second determining module 11000 is configured to determine whether the input information belongs to chat information without actual content, such as “hehe”, “hoho”, and the like. If it is determined that the chat information belongs to the chat information without the actual content, the second obtaining module 12000 may acquire the current topic, that is, calculate the current topic according to the historical chat record based on the Topic Model. After acquiring the current topic, the first generation module 13000 may generate a guidance topic based on the current topic based on the topic chat map. Among them, the topic chat map is a directed graph with a topic as a node. For example, as shown in FIG.
  • the second generation module 14000 can then generate a candidate reply based on the guided topic. Specifically, a template for a candidate reply may be generated based on a Natural Language Generation, a guide topic is filled into the template to generate a candidate reply, and a sentence including a guide topic may also be selected. As a candidate reply, the user is actively engaged in conversation topic guidance.
  • the artificial intelligence-based human-machine chat device of the embodiment of the present invention receives the input information input by the user, distributes the input information to the chat service module, and then receives the candidate reply returned by the plurality of chat service modules, and selects the candidate response based on the confidence level.
  • the reply is sorted, and the chat information is generated according to the sorting result, and the chat information is provided to the user, and the user can perform multiple rounds of chat, which is real and natural, and has the initiative to return a reply corresponding to the user style for different users, and is more personalized, Intelligent.
  • the present invention also provides a storage medium for storing an application for performing an artificial intelligence based human-machine chat method according to any of the embodiments of the present invention.
  • the present invention also provides an apparatus comprising: one or more processors; a memory; one or more modules, one or more modules stored in the memory when being one or more processors Perform the following operations when performing:
  • the input information is distributed to the chat service module.
  • first and second are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated.
  • features defining “first” or “second” may include at least one of the features, either explicitly or implicitly.
  • the meaning of "a plurality” is at least two, such as two, three, etc., unless specifically defined otherwise.
  • the terms “installation”, “connected”, “connected”, “fixed” and the like shall be understood broadly, and may be either a fixed connection or a detachable connection, unless explicitly stated and defined otherwise. , or integrated; can be mechanical or electrical connection; can be directly connected, or indirectly connected through an intermediate medium, can be the internal communication of two elements or the interaction of two elements, unless otherwise specified Limited.
  • the specific meanings of the above terms in the present invention can be understood on a case-by-case basis.
  • the first feature may be “on” or “under” the second feature, unless otherwise explicitly stated and defined.
  • the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium.
  • the first feature "above”, “above” and “above” the second feature may be that the first feature is directly above or above the second feature, or merely that the first feature level is higher than the second feature.
  • the first feature “below”, “below” and “below” the second feature may be that the first feature is directly below or obliquely below the second feature, or merely that the first feature level is less than the second feature.

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种基于人工智能的人机聊天方法和装置,其中,方法包括以下步骤:接收用户输入的输入信息(S1);将输入信息分发至聊天服务模块(S2);接收多个聊天服务模块返回的候选回复(S3);基于置信度对待选回复进行排序,并根据排序结果生成聊天信息,并向用户提供聊天信息(S4)。基于人工智能的人机聊天方法和装置,通过接收用户输入的输入信息,并将输入信息分发至聊天服务模块,然后接收多个聊天服务模块返回的候选回复,以及基于置信度对待选回复进行排序,并根据排序结果生成聊天信息,并向用户提供聊天信息,能够与用户进行多轮聊天,真实自然,具有主动性,还能针对不同用户返回符合用户风格的回复,更加个性化、智能化。

Description

基于人工智能的人机聊天方法和装置
相关申请的交叉引用
本申请要求百度在线网络技术(北京)有限公司于2015年9月7日提交的、发明名称为“基于人工智能的人机聊天方法和装置”的、中国专利申请号“201510564173.0”的优先权。
技术领域
本发明涉及人工智能技术领域,尤其涉及一种基于人工智能的人机聊天方法和装置。
背景技术
人工智能(Artificial Intelligence)是计算机科学的一个分支,英文缩写为AI,是研究、开发用于模拟、延伸和扩展人的智能的理论、方法、技术及应用***的一门新的技术科学。
人机聊天是指人与计算机之间进行交互聊天的过程。人机聊天可应用于娱乐及情感陪伴、智能服务拟人化等方面。例如:通过人机聊天***可以随时随地进行对话,缓解人们的生活压力,对于儿童还可帮助儿童提高语言能力。
目前,传统的人机聊天***主要基于大规模自动挖掘的聊天句对,每个聊天句对中含有聊天上句P(post)和针对P的下句R(Response)。对于用户输入的聊天句子Q(query),首先计算出与Q相似度最高的多个聊天上句{P1,P2,…,Pn},再对聊天上句对应的聊天下句{R1,R2,…,Rn}进行排序,然后选择出最优的聊天下句R返回给用户。
但是,传统的人机聊天***存在以下缺点:缺乏多轮聊天能力,即用户提出聊天上句,机器回复聊天下句,缺乏主动性,不够真实自然;对于时效性要求较高的对话例如新闻,则无法精准地进行回答;另外,聊天内容缺乏个性化,不同用户提出的相同或相似问题,只能回复同样的答案,缺乏个性化及智能化。
发明内容
本发明旨在至少在一定程度上解决相关技术中的技术问题之一。为此,本发明的一个目的在于提出一种基于人工智能的人机聊天方法,能够与用户进行多轮聊天,真实自然,具有主动性,还能针对不同用户返回符合用户风格的回复,更加个性化、智能化。
本发明的第二个目的在于提出一种基于人工智能的人机聊天装置。
为了实现上述目的,本发明第一方面实施例提出了一种基于人工智能的人机聊天方法, 包括:接收用户输入的输入信息;将所述输入信息分发至聊天服务模块;接收所述多个聊天服务模块返回的候选回复,其中,所述候选回复具有对应的置信度;基于所述置信度对所述待选回复进行排序,并根据排序结果生成聊天信息,并向所述用户提供所述聊天信息。
本发明实施例的基于人工智能的人机聊天方法,通过接收用户输入的输入信息,并将输入信息分发至聊天服务模块,然后接收多个聊天服务模块返回的候选回复,以及基于置信度对待选回复进行排序,并根据排序结果生成聊天信息,并向用户提供聊天信息,能够与用户进行多轮聊天,真实自然,具有主动性,还能针对不同用户返回符合用户风格的回复,更加个性化、智能化。
本发明第二方面实施例提出了一种基于人工智能的人机聊天装置,包括:第一接收模块,用于接收用户输入的输入信息;分发模块,用于将所述输入信息分发至聊天服务模块;第二接收模块,用于接收所述多个聊天服务模块返回的候选回复,其中,所述候选回复具有对应的置信度;提供模块,用于基于所述置信度对所述待选回复进行排序,并根据排序结果生成聊天信息,并向所述用户提供所述聊天信息。
本发明实施例的基于人工智能的人机聊天装置,通过接收用户输入的输入信息,并将输入信息分发至聊天服务模块,然后接收多个聊天服务模块返回的候选回复,以及基于置信度对待选回复进行排序,并根据排序结果生成聊天信息,并向用户提供聊天信息,能够与用户进行多轮聊天,真实自然,具有主动性,还能针对不同用户返回符合用户风格的回复,更加个性化、智能化。
本发明实施例第三方面公开了一种存储介质,用于存储应用程序,所述应用程序用于执行本发明第一方面实施例所述的基于人工智能的人机聊天方法。
本发明实施例第四方面公开了一种设备,包括:一个或者多个处理器;存储器;一个或者多个模块,所述一个或者多个模块存储在所述存储器中,当被所述一个或者多个处理器执行时进行如下操作:
接收用户输入的输入信息;
将所述输入信息分发至聊天服务模块;
接收所述多个聊天服务模块返回的候选回复,其中,所述候选回复具有对应的置信度;
基于所述置信度对所述待选回复进行排序,并根据排序结果生成聊天信息,并向所述用户提供所述聊天信息。
附图说明
本发明所述的和/或附加的方面和优点从下面结合附图对实施例的描述中将变得明显和容易理解,其中:
图1是根据本发明一个实施例的基于人工智能的人机聊天方法的流程图。
图2是根据本发明一个实施例的话题聊天图谱的效果示意图。
图3是根据本发明一个实施例的基于人工智能的人机聊天装置的结构示意图一。
图4是根据本发明一个实施例的基于人工智能的人机聊天装置的结构示意图二。
图5是根据本发明一个实施例的基于搜索的聊天模块的结构示意图。
图6是根据本发明一个实施例的富知识聊天模块的结构示意图。
图7是根据本发明一个实施例的基于画像的聊天模块的结构示意图一。
图8是根据本发明一个实施例的基于画像的聊天模块的结构示意图二。
图9是根据本发明一个实施例的基于人工智能的人机聊天装置的结构示意图三。
图10是根据本发明一个实施例的基于人工智能的人机聊天装置的结构示意图四。
图11是根据本发明一个实施例的基于人工智能的人机聊天装置的结构示意图五。
图12是根据本发明一个实施例的基于人工智能的人机聊天装置的结构示意图六。
具体实施方式
下面详细描述本发明的实施例,所述实施例的示例在附图中示出,其中自始至终相同或类似的标号表示相同或类似的元件或具有相同或类似功能的元件。下面通过参考附图描述的实施例是示例性的,旨在用于解释本发明,而不能理解为对本发明的限制。
下面参考附图描述本发明实施例的基于人工智能的人机聊天方法和装置。
图1是根据本发明一个实施例的基于人工智能的人机聊天方法的流程图。
如图1所示,基于人工智能的人机聊天方法可包括:
S1、接收用户输入的输入信息。
其中,输入信息可以是语音信息,也可以是文本信息。
在接收用户输入的输入信息之后,可对输入信息进行纠错和/或改写,用于纠正输入信息中的错别字,改写不规则的口语化表达等。
另外,还可获取与用户聊天的上文信息,然后根据上文信息判断输入信息与上文信息的依赖关系是否大于预设关系阈值。如果大于预设关系阈值,则可根据上文信息对输入信息进行补全,从而保证人机聊天的流畅度。具体地,对输入信息进行补全可包括指代消解。举例来说,输入信息为“他结婚了么?”,则可根据上文信息“刘德华”将输入信息中的“他”替代为“刘德华”。对输入信息进行补全还可包括省略补全。举例来说,上文信息“刘德华老婆叫朱丽倩。”,输入信息为“我不认识。”,则可将输入信息补全为“我不认识朱丽倩。”。
此外,还可根据上文信息获取用户当前的话题信息,以便后续聊天服务模块对聊天话 题进行引导。
S2、将输入信息分发至聊天服务模块。
具体地,可对输入信息进行领域分析以获取输入信息对应的领域。然后,可根据输入信息对应的领域将输入信息分发至具有相同或相近似领域的聊天服务模块。
其中,聊天服务模块可包括基于搜索的聊天模块、富知识聊天模块、基于画像的聊天模块和基于众包的聊天模块中的一种或多种。
具体地,基于搜索的聊天模块可对输入信息进行切词以生成多个聊天短句,然后可根据多个聊天短句查询聊天语料库从而生成多个聊天语料上句和多个聊天语料上句对应的多个聊天语料下句。其中,聊天语料库为预先建立,语聊来源可包括但不限于贴吧等论坛数据中的“发帖-回帖”、微博中的“博文-回复”、问答社区中的“问题-答案”等。
在此之后,可对多个聊天语料上句进行过滤。具体地,可计算输入信息与多个聊天语料上句之间的相似度。如果相似度小于第一预设相似度阈值,则可将对应的聊天语料上句过滤;如果相似度大于或等于第一预设相似度阈值,则可将对应的聊天语料上句保留。
在对聊天语料上句进行过滤之后,可对过滤之后的聊天语料上句对应的聊天语料下句进行分类。具体地,计算输入信息与多个聊天语料下句之间的相似度,并根据相似度基于GBDT(梯度升压决策树,Gradient Boost Decision Tree)、SVM(支持向量机,Support Vector Machine)等机器学习模型对多个聊天语料下句进行分类。其中,输入信息与多个聊天语料下句之间的相似度可以是输入信息与聊天语料下句之间字面的相似度,也可以是输入信息与聊天语料下句基于深度神经网络训练得到的相似度,也可以是输入信息与聊天语料下句基于机器翻译模型训练得到的相似度。应当理解的是,本实施例中输入信息与多个聊天语料下句之间的相似度以及GBDT、SVM等机器学习模型为公知技术,此处不赘述。
然后基于搜索的聊天模块可对分类之后的聊天语料下句进行重排序,并根据排序结果生成候选回复。具体地,可根据用户聊天的上文信息获取用户的聊天属性,再根据聊天属性对分类之后的聊天语料下句基于学习排序模型(Learning-To-Rank)进行重排序。其中,聊天属性可包括聊天的场合如时间地点等、聊天的趣味性、聊天的风格等。当然,聊天属性不仅限于从用户聊天的上文信息中获取,也可以根据用户长期的历史聊天记录获取。应当理解的是,本实施例中学习排序模型为公知技术,此处不赘述。
富知识聊天模块可根据输入信息生成搜索词,并根据搜索词进行搜索以生成多个搜索结果,然后对多个搜索结果进行句子抽取,以获取与搜索词的相似度大于第二预设相似度阈值的句子的候选句子集合。在此之后,可对候选句子集合中的句子进行改写以生成候选回复。此外,还可根据用户的聊天属性对候选句子集合中的句子进行重排序。举例来说,输入信息为“希望有机会能到富士山旅游”,可对输入信息进行解析并生成对应的搜索词 “富士山、旅游”,然后根据搜索词获得多个搜索结果,并抽取与搜索词相似度高的句子。其中,有的句子可能包括如“记者了解到”等明显节选自网页文本,因此需要对这些句子进行改写,使其更加流畅,更像自然语言聊天的句子,最终生成的候选回复为“富士山由于天气原因,一年中只有规定的夏季的一段时间可以登山”,相对于传统的回复“我也想去富士山,一起吧。”,具有一定的知识性,且具有一定时效性,可使用户能在聊天过程中获取有用的知识。
为了更好地实现拟人化,以及为用户提供个性化服务,人机聊天***可设定自身的属性、状态、兴趣等,即***画像模型。还可设定用户的属性、状态、兴趣等,即用户画像模型。当然,在面对不同的用户时,使用的***画像模型可以是同一个,也可以针对每个用户均可设置与之对应的***画像模型。***画像模型和用户画像模型均基于画像知识图谱。画像知识图谱是一个层次化的知识体系。举例来说,“家庭成员”节点可包括“兄弟姐妹”和“父母”两个子节点,“父母”子节点包括“父亲”和“母亲”两个子节点。每个节点均对应有多个输入信息模板簇,例如“你父亲是谁”、“谁是你父亲”、“你的父亲叫什么”属于同一个输入信息模板簇。每个输入信息模板簇对应一个或多个候选回复。输入信息模板簇和候选回复可包含变量,例如兴趣、爱好、嗜好对应同一属性“INTEREST”,而“INTEREST”的属性值可包括爬山、音乐、读书、运动等。
具体地,基于画像的聊天模块可获取用户的聊天语境,并根据聊天语境判断是否满足收集条件。如果判断满足收集条件,则可向用户发送问题。在此之后,可接收用户根据问题的回答信息,并根据回答信息对用户画像模型进行更新。例如:在与用户聊电影相关的话题时,可向用户发送问题“你喜欢什么电影?”或者用户问人机聊天***“你喜欢吃什么?”,人机聊天***可反问用户“你喜欢吃什么?”,在用户回答后,可基于用户的回答信息对用户画像模型进行更新,更加符合用户个性化的需求。
此外,基于画像的聊天模块还可获取用户的聊天内容,并根据聊天内容提取用户画像数据,然后根据提取的用户画像数据对用户画像模型进行更新。例如:用户在聊天过程中说道“我没事的时候喜欢爬爬山、钓钓鱼。”,可提取用户画像数据“爱好爬山、爱好钓鱼”,从而对用户画像模型进行更新。同时,可基于用户画像数据抽取合适的答案,向用户返回合适的回答信息。
众包(crowdsourcing)是一种将特定任务外包给互联网中非特定用户的方法,对于人机聊天中,机器难以回答的问题,可分发给执行者在线地实时地进行人工回复,从而满足用户的实际需求。
具体地,基于众包的聊天模块可判断输入信息是否适合众包完成,例如用户情绪低落需要安慰等,则适合众包完成。例如用户的输入信息中包含有个人身份信息、密码、电话 等隐私信息,则不适合众包完成。
如果判断适合众包完成,则可将输入信息分发至对应的执行者。当然,同时也可将上文信息一同发送给执行者,执行者可根据上文信息和输入信息进行回复。然后基于众包的聊天模块可接收执行者的回复信息,并对回复信息进行质量判断。如果满足质量要求,则将回复信息作为候选回复。例如:回复信息中如果包含低俗、反动、色情内容,则质量不过关。或者执行者回复的时间超过了预定时长,则该执行者的回复信息将不被采用,同时可将该回复信息保存至聊天语料库中。
在此之外,还可判断输入信息是否属于无实际内容的聊天信息,如“呵呵”、“hoho”等。如果判断是属于无实际内容的聊天信息,则可获取当前话题,即基于话题模型(Topic Model)根据历史聊天记录计算出当前话题。在获取当前话题之后,可基于话题聊天图谱根据当前话题生成引导话题。其中,话题聊天图谱是一个以话题为节点的有向图。例如,如图2所示,节点“休闲”可指向节点“看电影”和节点“听歌”,则说明可从话题“休闲”引导至话题“看电影”或者话题“听歌”。话题“看电影”和话题“听歌”均具有一定的引导概率,可根据引导概率实现话题的引导,从而保证引导话题的多样性。
然后,可根据引导话题生成候选回复。具体地,可基于自然语言生成模型(Natural Language Generation),生成候选回复的模板,将引导话题填充至该模板中生成候选回复;也可以选取包含引导话题的句子作为候选回复,从而实现对用户进行主动地聊天话题引导。
S3、接收多个聊天服务模块返回的候选回复。
其中,候选回复具有对应的置信度。
S4、基于置信度对待选回复进行排序,并根据排序结果生成聊天信息,并向用户提供聊天信息。
具体地,可获取用户的输入信息的特征,并基于输入信息的特征和置信度对待选回复进行排序。其中,输入信息的特征可包括分类特征、字面特征、话题特征等。置信度越高,则待选回复质量越好,可按照置信度从高到低的顺序对待选回复进行排序,最终向用户提供符合用户需求的聊天信息。
另外,还可通过增强学习模型(Reinforcement Learning)根据用户的反馈信息进行更新,从而能够为用户提供更满意的聊天信息。例如:在回复用户的聊天信息中添加评论按钮如“赞”或“踩”以收集用户的反馈信息;或者基于情感分析技术,对用户在聊天中的输入信息进行分析,从而获得用户的评价,例如:“你真智能”等;或者通过记录与用户聊天的交互次数,判断用户的满意度。
本发明实施例的基于人工智能的人机聊天方法,通过接收用户输入的输入信息,并将输入信息分发至聊天服务模块,然后接收多个聊天服务模块返回的候选回复,以及基于置 信度对待选回复进行排序,并根据排序结果生成聊天信息,并向用户提供聊天信息,能够与用户进行多轮聊天,真实自然,具有主动性,还能针对不同用户返回符合用户风格的回复,更加个性化、智能化。
为实现上述目的,本发明还提出一种基于人工智能的人机聊天装置。
图3是根据本发明一个实施例的基于人工智能的人机聊天装置的结构示意图一。
如图3所示,该基于人工智能的人机聊天装置可包括:第一接收模块1000、分发模块2000、聊天服务模块3000、第二接收模块4000和提供模块5000。
第一接收模块1000用于接收用户输入的输入信息。
其中,输入信息可以是语音信息,也可以是文本信息。
分发模块2000用于将输入信息分发至聊天服务模块3000。
第二接收模块4000用于接收多个聊天服务模块3000返回的候选回复。其中,候选回复具有对应的置信度。
提供模块5000用于基于置信度对待选回复进行排序,并根据排序结果生成聊天信息,并向用户提供聊天信息。
具体地,提供模块5000可获取用户的输入信息的特征,并基于输入信息的特征和置信度对待选回复进行排序。其中,输入信息的特征可包括分类特征、字面特征、话题特征等。置信度越高,则待选回复质量越好,可按照置信度从高到低的顺序对待选回复进行排序,最终向用户提供符合用户需求的聊天信息。
如图4所示,聊天服务模块3000可包括基于搜索的聊天模块3100、富知识聊天模块3200、基于画像的聊天模块3300和基于众包的聊天模块3400。
其中,如图5所示,基于搜索的聊天模块3100可包括切词子模块3110、生成子模块3120、过滤子模块3130、分类子模块3140和排序子模块3150。其中,过滤子模块3130可包括计算单元3131、过滤单元3132、保留单元3133,分类子模块3140可包括计算单元3141、分类单元3142,排序子模块3150可包括获取单元3151和排序单元3152。
具体地,切词子模块3110可对输入信息进行切词以生成多个聊天短句,然后生成子模块3120可根据多个聊天短句查询聊天语料库从而生成多个聊天语料上句和多个聊天语料上句对应的多个聊天语料下句。其中,聊天语料库为预先建立,语聊来源可包括但不限于贴吧等论坛数据中的“发帖-回帖”、微博中的“博文-回复”、问答社区中的“问题-答案”等。
在此之后,过滤子模块3130可对多个聊天语料上句进行过滤。具体地,计算单元3131可计算输入信息与多个聊天语料上句之间的相似度。如果相似度小于第一预设相似度阈值,则过滤单元3132可将对应的聊天语料上句过滤;如果相似度大于或等于第一预设相似度阈 值,则保留单元3133可将对应的聊天语料上句保留。
在对聊天语料上句进行过滤之后,分类子模块3140可对过滤之后的聊天语料上句对应的聊天语料下句进行分类。具体地,计算单元3141可计算输入信息与多个聊天语料下句之间的相似度,分类单元3142根据相似度基于GBDT(梯度升压决策树,Gradient Boost Decision Tree)、SVM(支持向量机,Support Vector Machine)等机器学习模型对多个聊天语料下句进行分类。其中,输入信息与多个聊天语料下句之间的相似度可以是输入信息与聊天语料下句之间字面的相似度,也可以是输入信息与聊天语料下句基于深度神经网络训练得到的相似度,也可以是输入信息与聊天语料下句基于机器翻译模型训练得到的相似度。应当理解的是,本实施例中输入信息与多个聊天语料下句之间的相似度以及GBDT、SVM等机器学习模型为公知技术,此处不赘述。
然后排序子模块3150可对分类之后的聊天语料下句进行重排序,并根据排序结果生成候选回复。具体地,获取单元3151可根据用户聊天的上文信息获取用户的聊天属性,排序单元3152根据聊天属性对分类之后的聊天语料下句基于学习排序模型(Learning-To-Rank)进行重排序。其中,聊天属性可包括聊天的场合如时间地点等、聊天的趣味性、聊天的风格等。当然,聊天属性不仅限于从用户聊天的上文信息中获取,也可以根据用户长期的历史聊天记录获取。应当理解的是,本实施例中学习排序模型为公知技术,此处不赘述。
如图6所示,富知识聊天模块3200可包括生成子模块3210、抽取子模块3220、改写子模块3230和重排序子模块3240。具体地,生成子模块3210可根据输入信息生成搜索词,并根据搜索词进行搜索以生成多个搜索结果,然后抽取子模块3220对多个搜索结果进行句子抽取,以获取与搜索词的相似度大于第二预设相似度阈值的句子的候选句子集合。在此之后,改写子模块3230可对候选句子集合中的句子进行改写以生成候选回复。此外,重排序子模块3240可根据用户的聊天属性对候选句子集合中的句子进行重排序。举例来说,输入信息为“希望有机会能到富士山旅游”,可对输入信息进行解析并生成对应的搜索词“富士山、旅游”,然后根据搜索词获得多个搜索结果,并抽取与搜索词相似度高的句子。其中,有的句子可能包括如“记者了解到”等明显节选自网页文本,因此需要对这些句子进行改写,使其更加流畅,更像自然语言聊天的句子,最终生成的候选回复为“富士山由于天气原因,一年中只有规定的夏季的一段时间可以登山”,相对于传统的回复“我也想去富士山,一起吧。”,具有一定的知识性,且具有一定时效性,可使用户能在聊天过程中获取有用的知识。
如图7所示,基于画像的聊天模块3300可包括第一获取子模块3310、判断子模块3320、发送子模块3330、第一更新子模块3340。
为了更好地实现拟人化,以及为用户提供个性化服务,人机聊天***可设定自身的属 性、状态、兴趣等,即***画像模型。还可设定用户的属性、状态、兴趣等,即用户画像模型。当然,在面对不同的用户时,使用的***画像模型可以是同一个,也可以针对每个用户均可设置与之对应的***画像模型。***画像模型和用户画像模型均基于画像知识图谱。画像知识图谱是一个层次化的知识体系。举例来说,“家庭成员”节点可包括“兄弟姐妹”和“父母”两个子节点,“父母”子节点包括“父亲”和“母亲”两个子节点。每个节点均对应有多个输入信息模板簇,例如“你父亲是谁”、“谁是你父亲”、“你的父亲叫什么”属于同一个输入信息模板簇。每个输入信息模板簇对应一个或多个候选回复。输入信息模板簇和候选回复可包含变量,例如兴趣、爱好、嗜好对应同一属性“INTEREST”,而“INTEREST”的属性值可包括爬山、音乐、读书、运动等。
具体地,第一获取子模块3310可获取用户的聊天语境,判断子模块3320根据聊天语境判断是否满足收集条件。如果判断满足收集条件,则发送子模块3330可向用户发送问题。在此之后,第一更新子模块3340可接收用户根据问题的回答信息,并根据回答信息对用户画像模型进行更新。例如:在与用户聊电影相关的话题时,可向用户发送问题“你喜欢什么电影?”或者用户问人机聊天***“你喜欢吃什么?”,人机聊天***可反问用户“你喜欢吃什么?”,在用户回答后,可基于用户的回答信息对用户画像模型进行更新,更加符合用户个性化的需求。
此外,如图8所示,基于画像的聊天模块3300还可包括第二获取子模块3350、提取子模块3360、第二更新子模块3370。
具体地,第二获取子模块3350可获取用户的聊天内容,提取子模块3360根据聊天内容提取用户画像数据,然后第二更新子模块3370根据提取的用户画像数据对用户画像模型进行更新。例如:用户在聊天过程中说道“我没事的时候喜欢爬爬山、钓钓鱼。”,可提取用户画像数据“爱好爬山、爱好钓鱼”,从而对用户画像模型进行更新。同时,可基于用户画像数据抽取合适的答案,向用户返回合适的回答信息。
众包(crowdsourcing)是一种将特定任务外包给互联网中非特定用户的方法,对于人机聊天中,机器难以回答的问题,可分发给执行者在线地实时地进行人工回复,从而满足用户的实际需求。
具体地,基于众包的聊天模块3400可判断输入信息是否适合众包完成,例如用户情绪低落需要安慰等,则适合众包完成。例如用户的输入信息中包含有个人身份信息、密码、电话等隐私信息,则不适合众包完成。
如果判断适合众包完成,则基于众包的聊天模块3400可将输入信息分发至对应的执行者。当然,同时也可将上文信息一同发送给执行者,执行者可根据上文信息和输入信息进行回复。然后基于众包的聊天模块可接收执行者的回复信息,并对回复信息进行质量判断。 如果满足质量要求,则将回复信息作为候选回复。例如:回复信息中如果包含低俗、反动、色情内容,则质量不过关。或者执行者回复的时间超过了预定时长,则该执行者的回复信息将不被采用,同时可将该回复信息保存至聊天语料库中。
另外,如图9所示,该基于人工智能的人机聊天装置还可包括纠错模块6000。
纠错模块6000用于在接收用户输入的输入信息之后,对输入信息进行纠错和/或改写,用于纠正输入信息中的错别字,改写不规则的口语化表达等。
另外,如图10所示,该基于人工智能的人机聊天装置还可包括分析模块7000。
分析模块7000用于在接收用户输入的输入信息之后,对输入信息进行领域分析以获取输入信息对应的领域,然后分发模块2000可根据输入信息对应的领域将输入信息分发至具有相同或相近似领域的聊天服务模块。
另外,如图11所示,该基于人工智能的人机聊天装置还可包括第一获取模块8000、第一判断模块9000、补全模块10000。
第一获取模块8000用于在接收用户输入的输入信息之后,获取与用户聊天的上文信息,并根据上文信息获取用户当前的话题信息。然后,第一判断模块9000可根据上文信息判断输入信息与上文信息的依赖关系是否大于预设关系阈值。在依赖关系大于预设关系阈值时,补全模块10000可根据上文信息对输入信息进行补全,从而保证人机聊天的流畅度。具体地,对输入信息进行补全可包括指代消解。举例来说,输入信息为“他结婚了么?”,则可根据上文信息“刘德华”将输入信息中的“他”替代为“刘德华”。对输入信息进行补全还可包括省略补全。举例来说,上文信息“刘德华老婆叫朱丽倩。”,输入信息为“我不认识。”,则可将输入信息补全为“我不认识朱丽倩。”。
此外,如图12所示,该基于人工智能的人机聊天装置还可包括第二判断模块11000、第二获取模块12000、第一生成模块13000和第二生成模块14000。
第二判断模块11000用于判断输入信息是否属于无实际内容的聊天信息,如“呵呵”、“hoho”等。如果判断是属于无实际内容的聊天信息,则第二获取模块12000可获取当前话题,即基于话题模型(Topic Model)根据历史聊天记录计算出当前话题。在获取当前话题之后,第一生成模块13000可基于话题聊天图谱根据当前话题生成引导话题。其中,话题聊天图谱是一个以话题为节点的有向图。例如,如图2所示,节点“休闲”可指向节点“看电影”和节点“听歌”,则说明可从话题“休闲”引导至话题“看电影”或者话题“听歌”。话题“看电影”和话题“听歌”均具有一定的引导概率,可根据引导概率实现话题的引导,从而保证引导话题的多样性。然后,第二生成模块14000可根据引导话题生成候选回复。具体地,可基于自然语言生成模型(Natural Language Generation),生成候选回复的模板,将引导话题填充至该模板中生成候选回复;也可以选取包含引导话题的句子 作为候选回复,从而实现对用户进行主动地聊天话题引导。
本发明实施例的基于人工智能的人机聊天装置,通过接收用户输入的输入信息,并将输入信息分发至聊天服务模块,然后接收多个聊天服务模块返回的候选回复,以及基于置信度对待选回复进行排序,并根据排序结果生成聊天信息,并向用户提供聊天信息,能够与用户进行多轮聊天,真实自然,具有主动性,还能针对不同用户返回符合用户风格的回复,更加个性化、智能化。
为了实现上述实施例,本发明还提出了一种存储介质,用于存储应用程序,该应用程序用于执行本发明任一个实施例所述的基于人工智能的人机聊天方法。
为了实现上述实施例,本发明还提出了一种设备,包括:一个或者多个处理器;存储器;一个或者多个模块,一个或者多个模块存储在存储器中,当被一个或者多个处理器执行时进行如下操作:
S1’、接收用户输入的输入信息。
S2’、将输入信息分发至聊天服务模块。
S3’、接收多个聊天服务模块返回的候选回复。
S4’、基于置信度对待选回复进行排序,并根据排序结果生成聊天信息,并向用户提供聊天信息。
在本发明的描述中,需要理解的是,术语“中心”、“纵向”、“横向”、“长度”、“宽度”、“厚度”、“上”、“下”、“前”、“后”、“左”、“右”、“竖直”、“水平”、“顶”、“底”“内”、“外”、“顺时针”、“逆时针”、“轴向”、“径向”、“周向”等指示的方位或位置关系为基于附图所示的方位或位置关系,仅是为了便于描述本发明和简化描述,而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作,因此不能理解为对本发明的限制。
此外,术语“第一”、“第二”仅用于描述目的,而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此,限定有“第一”、“第二”的特征可以明示或者隐含地包括至少一个该特征。在本发明的描述中,“多个”的含义是至少两个,例如两个,三个等,除非另有明确具体的限定。
在本发明中,除非另有明确的规定和限定,术语“安装”、“相连”、“连接”、“固定”等术语应做广义理解,例如,可以是固定连接,也可以是可拆卸连接,或成一体;可以是机械连接,也可以是电连接;可以是直接相连,也可以通过中间媒介间接相连,可以是两个元件内部的连通或两个元件的相互作用关系,除非另有明确的限定。对于本领域的普通技术人员而言,可以根据具体情况理解上述术语在本发明中的具体含义。
在本发明中,除非另有明确的规定和限定,第一特征在第二特征“上”或“下”可以 是第一和第二特征直接接触,或第一和第二特征通过中间媒介间接接触。而且,第一特征在第二特征“之上”、“上方”和“上面”可是第一特征在第二特征正上方或斜上方,或仅仅表示第一特征水平高度高于第二特征。第一特征在第二特征“之下”、“下方”和“下面”可以是第一特征在第二特征正下方或斜下方,或仅仅表示第一特征水平高度小于第二特征。
在本说明书的描述中,参考术语“一个实施例”、“一些实施例”、“示例”、“具体示例”、或“一些示例”等的描述意指结合该实施例或示例描述的具体特征、结构、材料或者特点包含于本发明的至少一个实施例或示例中。在本说明书中,对上述术语的示意性表述不必须针对的是相同的实施例或示例。而且,描述的具体特征、结构、材料或者特点可以在任一个或多个实施例或示例中以合适的方式结合。此外,在不相互矛盾的情况下,本领域的技术人员可以将本说明书中描述的不同实施例或示例以及不同实施例或示例的特征进行结合和组合。
尽管上面已经示出和描述了本发明的实施例,可以理解的是,上述实施例是示例性的,不能理解为对本发明的限制,本领域的普通技术人员在本发明的范围内可以对上述实施例进行变化、修改、替换和变型。

Claims (40)

  1. 一种基于人工智能的人机聊天方法,其特征在于,包括以下步骤:
    接收用户输入的输入信息;
    将所述输入信息分发至聊天服务模块;
    接收所述多个聊天服务模块返回的候选回复,其中,所述候选回复具有对应的置信度;
    基于所述置信度对所述待选回复进行排序,并根据排序结果生成聊天信息,并向所述用户提供所述聊天信息。
  2. 如权利要求1所述的方法,其特征在于,在所述接收用户输入的输入信息之后,还包括:
    对所述输入信息进行纠错和/或改写。
  3. 如权利要求1所述的方法,其特征在于,在所述接收用户输入的输入信息之后,还包括:
    对所述输入信息进行领域分析以获取所述输入信息对应的领域,其中,根据所述输入信息对应的领域将所述输入信息分发至具有相同或相近似领域的聊天服务模块。
  4. 如权利要求1所述的方法,其特征在于,在所述接收用户输入的输入信息之后,还包括:
    获取与所述用户聊天的上文信息;
    根据所述上文信息判断所述输入信息与所述上文信息的依赖关系是否大于预设关系阈值;以及
    如果大于所述预设关系阈值,则根据所述上文信息对所述输入信息进行补全。
  5. 如权利要求4所述的方法,其特征在于,还包括:
    根据所述上文信息获取所述用户当前的话题信息。
  6. 如权利要求1所述的方法,其特征在于,所述聊天服务模块包括基于搜索的聊天模块、富知识聊天模块、基于画像的聊天模块和基于众包的聊天模块中的一种或多种。
  7. 如权利要求6所述的方法,其特征在于,还包括:
    所述基于搜索的聊天模块对所述输入信息进行切词以生成多个聊天短句;
    所述基于搜索的聊天模块根据所述多个聊天短句查询聊天语料库以生成多个聊天语料上句,以及所述多个聊天语料上句对应的多个聊天语料下句;
    所述基于搜索的聊天模块对所述多个聊天语料上句进行过滤;
    所述基于搜索的聊天模块对过滤之后的聊天语料上句对应的聊天语料下句进行分类;以及
    所述基于搜索的聊天模块对分类之后的所述聊天语料下句进行重排序,并根据排序结果生成所述候选回复。
  8. 如权利要求7所述的方法,其特征在于,所述基于搜索的聊天模块对所述多个聊天语料上句进行过滤具体包括:
    计算所述输入信息与所述多个聊天语料上句之间的相似度;
    如果所述相似度小于第一预设相似度阈值,则将对应的聊天语料上句过滤;以及
    如果所述相似度大于或等于所述第一预设相似度阈值,则将对应的聊天语料上句保留。
  9. 如权利要求7所述的方法,其特征在于,所述对过滤之后的聊天语料上句对应的聊天语料下句进行分类具体包括:
    计算所述输入信息与所述多个聊天语料下句之间的相似度;以及
    根据所述相似度对所述多个聊天语料下句进行分类。
  10. 如权利要求9所述的方法,其特征在于,所述输入信息与所述多个聊天语料下句之间的相似度包括:
    所述输入信息与所述聊天语料下句之间字面的相似度;
    或者,所述输入信息与所述聊天语料下句基于深度神经网络训练得到的相似度;
    或者,所述输入信息与所述聊天语料下句基于机器翻译模型训练得到的相似度。
  11. 如权利要求7所述的方法,其特征在于,所述对分类之后的所述聊天语料下句进行重排序具体包括:
    根据所述用户聊天的上文信息获取所述用户的聊天属性;
    根据所述聊天属性对所述分类之后的所述聊天语料下句进行重排序。
  12. 如权利要求6所述的方法,其特征在于,还包括:
    所述富知识聊天模块根据所述输入信息生成搜索词,并根据所述搜索词进行搜索以生成多个搜索结果;
    所述富知识聊天模块对所述多个搜索结果进行句子抽取,以获取候选句子集合;
    所述富知识聊天模块对所述候选句子集合中的句子进行改写以生成所述候选回复。
  13. 如权利要求12所述的方法,其特征在于,其中,所述候选句子集合中的句子与所述搜索词的相似度大于第二预设相似度阈值。
  14. 如权利要求12所述的方法,其特征在于,还包括:
    根据所述用户的聊天属性对所述候选句子集合中的句子进行重排序。
  15. 如权利要求6所述的方法,其特征在于,还包括:
    所述基于画像的聊天模块获取所述用户的聊天语境;
    所述基于画像的聊天模块根据所述聊天语境判断是否满足收集条件;
    如果判断满足所述收集条件,则向所述用户发送问题;
    接收所述用户根据所述问题的回答信息,并根据所述回答信息对用户画像模型进行更新。
  16. 如权利要求6所述的方法,其特征在于,还包括:
    所述基于画像的聊天模块获取所述用户的聊天内容;
    所述基于画像的聊天模块根据所述聊天内容提取用户画像数据;
    所述基于画像的聊天模块根据提取的所述用户画像数据对所述用户画像模型进行更新。
  17. 如权利要求6所述的方法,其特征在于,还包括:
    所述基于众包的聊天模块判断所述输入信息是否适合众包完成;
    如果判断适合众包完成,则将所述输入信息分发至对应的执行者;
    接收所述执行者的回复信息,并对所述回复信息进行质量判断;
    如果满足质量要求,则将所述回复信息作为所述候选回复。
  18. 如权利要求1所述的方法,其特征在于,还包括:
    判断所述输入信息是否属于无实际内容的聊天信息;
    如果判断是属于无实际内容的聊天信息,则获取当前话题;
    根据所述当前话题生成引导话题;以及
    根据所述引导话题生成所述候选回复。
  19. 如权利要求1所述的方法,其特征在于,所述基于所述置信度对所述待选回复进行排序具体包括:
    获取所述用户的所述输入信息的特征;以及
    基于所述输入信息的特征和所述置信度对所述待选回复进行排序。
  20. 一种基于人工智能的人机聊天装置,其特征在于,包括:
    第一接收模块,用于接收用户输入的输入信息;
    分发模块,用于将所述输入信息分发至聊天服务模块;
    第二接收模块,用于接收所述多个聊天服务模块返回的候选回复,其中,所述候选回复具有对应的置信度;
    提供模块,用于基于所述置信度对所述待选回复进行排序,并根据排序结果生成聊天信息,并向所述用户提供所述聊天信息。
  21. 如权利要求20所述的装置,其特征在于,所述装置还包括:
    纠错模块,用于在所述接收用户输入的输入信息之后,对所述输入信息进行纠错和/或改写。
  22. 如权利要求20所述的装置,其特征在于,所述装置还包括:
    分析模块,用于在所述接收用户输入的输入信息之后,对所述输入信息进行领域分析以获取所述输入信息对应的领域;
    所述分发模块,用于根据所述输入信息对应的领域将所述输入信息分发至具有相同或相近似领域的聊天服务模块。
  23. 如权利要求20所述的装置,其特征在于,所述装置还包括:
    第一获取模块,用于在所述接收用户输入的输入信息之后,获取与所述用户聊天的上文信息;
    第一判断模块,用于根据所述上文信息判断所述输入信息与所述上文信息的依赖关系是否大于预设关系阈值;以及
    补全模块,用于在所述依赖关系大于所述预设关系阈值时,根据所述上文信息对所述输入信息进行补全。
  24. 如权利要求23所述的装置,其特征在于,所述第一获取模块还用于:
    根据所述上文信息获取所述用户当前的话题信息。
  25. 如权利要求20所述的装置,其特征在于,所述聊天服务模块包括基于搜索的聊天模块、富知识聊天模块、基于画像的聊天模块和基于众包的聊天模块中的一种或多种。
  26. 如权利要求25所述的方法,其特征在于,所述基于搜索的聊天模块,具体包括:
    切词子模块,用于对所述输入信息进行切词以生成多个聊天短句;
    生成子模块,用于根据所述多个聊天短句查询聊天语料库以生成多个聊天语料上句,以及所述多个聊天语料上句对应的多个聊天语料下句;
    过滤子模块,用于对所述多个聊天语料上句进行过滤;
    分类子模块,用于对过滤之后的聊天语料上句对应的聊天语料下句进行分类;以及
    排序子模块,用于对分类之后的所述聊天语料下句进行重排序,并根据排序结果生成所述候选回复。
  27. 如权利要求26所述的装置,其特征在于,所述过滤子模块,具体包括:
    计算单元,用于计算所述输入信息与所述多个聊天语料上句之间的相似度;
    过滤单元,用于如果所述相似度小于第一预设相似度阈值,则将对应的聊天语料上句过滤;以及
    保留单元,用于如果所述相似度大于或等于所述第一预设相似度阈值,则将对应的聊天语料上句保留。
  28. 如权利要求26所述的装置,其特征在于,所述分类子模块具体包括:
    计算单元,用于计算所述输入信息与所述多个聊天语料下句之间的相似度;以及
    分类单元,用于根据所述相似度对所述多个聊天语料下句进行分类。
  29. 如权利要求28所述的装置,其特征在于,所述输入信息与所述多个聊天语料下句之间的相似度包括:
    所述输入信息与所述聊天语料下句之间字面的相似度;
    或者,所述输入信息与所述聊天语料下句基于深度神经网络训练得到的相似度;
    或者,所述输入信息与所述聊天语料下句基于机器翻译模型训练得到的相似度。
  30. 如权利要求26所述的装置,其特征在于,所述排序子模块具体包括:
    获取单元,用于根据所述用户聊天的上文信息获取所述用户的聊天属性;
    排序单元,用于根据所述聊天属性对所述分类之后的所述聊天语料下句进行重排序。
  31. 如权利要求25所述的装置,其特征在于,所述富知识聊天模块,具体包括:
    生成子模块,用于根据所述输入信息生成搜索词,并根据所述搜索词进行搜索以生成多个搜索结果;
    抽取子模块,用于对所述多个搜索结果进行句子抽取,以候选句子集合;
    改写子模块,用于对所述候选句子集合中的句子进行改写以生成所述候选回复。
  32. 如权利要求31所述的装置,其特征在于,其中,所述候选句子集合中的句子与所述搜索词的相似度大于第二预设相似度阈值。
  33. 如权利要求31所述的装置,其特征在于,还包括:
    重排序子模块,用于根据所述用户的聊天属性对所述候选句子集合中的句子进行重排序。
  34. 如权利要求25所述的装置,其特征在于,所述基于画像的聊天模块具体包括:
    第一获取子模块,用于获取所述用户的聊天语境;
    判断子模块,用于根据所述聊天语境判断是否满足收集条件;
    发送子模块,用于当满足所述收集条件时,向所述用户发送问题;
    第一更新子模块,用于接收所述用户根据所述问题的回答信息,并根据所述回答信息对用户画像模型进行更新。
  35. 如权利要求25所述的装置,其特征在于,所述基于画像的聊天模块具体包括:
    第二获取子模块,用于获取所述用户的聊天内容;
    提取子模块,用于根据所述聊天内容提取用户画像数据;
    第二更新子模块,用于根据提取的所述用户画像数据对所述用户画像模型进行更新。
  36. 如权利要求25所述的装置,其特征在于,所述基于众包的聊天模块具体用于:
    判断所述输入信息是否适合众包完成,如果判断适合众包完成,则将所述输入信息分发至对应的执行者,以及接收所述执行者的回复信息,并对所述回复信息进行质量判断, 如果满足质量要求,则将所述回复信息作为所述候选回复。
  37. 如权利要求20所述的装置,其特征在于,还包括:
    第二判断模块,用于判断所述输入信息是否属于无实际内容的聊天信息;
    第二获取模块,用于如果判断是属于无实际内容的聊天信息,则获取当前话题;
    第一生成模块,用于根据所述当前话题生成引导话题;以及
    第二生成模块,用于根据所述引导话题生成所述候选回复。
  38. 如权利要求20所述的装置,其特征在于,所述提供模块具体用于:
    获取所述用户的所述输入信息的特征,并基于所述输入信息的特征和所述置信度对所述待选回复进行排序。
  39. 一种存储介质,其特征在于,用于存储应用程序,所述应用程序用于执行权利要求1至19中任一项所述的基于人工智能的人机聊天方法。
  40. 一种设备,其特征在于,包括:
    一个或者多个处理器;
    存储器;
    一个或者多个模块,所述一个或者多个模块存储在所述存储器中,当被所述一个或者多个处理器执行时进行如下操作:
    接收用户输入的输入信息;
    将所述输入信息分发至聊天服务模块;
    接收所述多个聊天服务模块返回的候选回复,其中,所述候选回复具有对应的置信度;
    基于所述置信度对所述待选回复进行排序,并根据排序结果生成聊天信息,并向所述用户提供所述聊天信息。
PCT/CN2015/096342 2015-09-07 2015-12-03 基于人工智能的人机聊天方法和装置 WO2017041370A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201510564173.0A CN105183848A (zh) 2015-09-07 2015-09-07 基于人工智能的人机聊天方法和装置
CN201510564173.0 2015-09-07

Publications (1)

Publication Number Publication Date
WO2017041370A1 true WO2017041370A1 (zh) 2017-03-16

Family

ID=54905929

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/096342 WO2017041370A1 (zh) 2015-09-07 2015-12-03 基于人工智能的人机聊天方法和装置

Country Status (2)

Country Link
CN (1) CN105183848A (zh)
WO (1) WO2017041370A1 (zh)

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109190114A (zh) * 2018-08-13 2019-01-11 北京百度网讯科技有限公司 用于生成回复信息的方法和装置
CN109242706A (zh) * 2018-08-20 2019-01-18 中国平安人寿保险股份有限公司 辅助坐席人员沟通的方法、装置、计算机设备和存储介质
CN110209897A (zh) * 2018-02-12 2019-09-06 腾讯科技(深圳)有限公司 智能对话方法、装置、存储介质及设备
CN111078893A (zh) * 2019-12-11 2020-04-28 竹间智能科技(上海)有限公司 一种大规模高效获取识别对话意图用语料的方法
CN111291170A (zh) * 2020-01-20 2020-06-16 腾讯科技(深圳)有限公司 一种基于智能客服的会话推荐方法及相关装置
CN111353091A (zh) * 2018-12-24 2020-06-30 北京三星通信技术研究有限公司 信息处理方法、装置、电子设备及可读存储介质
US10826864B2 (en) 2018-07-27 2020-11-03 At&T Intellectual Property I, L.P. Artificially intelligent messaging
CN112967721A (zh) * 2021-02-03 2021-06-15 上海明略人工智能(集团)有限公司 一种基于语音识别技术的销售线索信息识别方法与***
CN113127613A (zh) * 2020-01-10 2021-07-16 北京搜狗科技发展有限公司 聊天信息处理方法及装置
US11397852B2 (en) * 2018-12-17 2022-07-26 Beijing Baidu Netcom Science and Technology Co., Ltd News interaction method, apparatus, device and computer storage medium
CN116628153A (zh) * 2023-05-10 2023-08-22 上海任意门科技有限公司 一种人工智能设备对话的控制方法、装置、设备及介质
CN117216229A (zh) * 2023-11-08 2023-12-12 支付宝(杭州)信息技术有限公司 一种生成客服答案的方法及装置

Families Citing this family (51)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112417181A (zh) * 2016-02-26 2021-02-26 原相科技股份有限公司 主动响应的交互式服务平台及其运作方法
CN105808695A (zh) * 2016-03-03 2016-07-27 陈包容 一种获取聊天回复内容的方法及装置
CN105787560B (zh) * 2016-03-18 2018-04-03 北京光年无限科技有限公司 基于循环神经网络的对话数据交互处理方法及装置
CN107305768B (zh) * 2016-04-20 2020-06-12 上海交通大学 语音交互中的易错字校准方法
CN106489159A (zh) * 2016-06-29 2017-03-08 深圳狗尾草智能科技有限公司 一种基于深度神经网络的用户画像表示学习***及方法
WO2018000271A1 (zh) * 2016-06-29 2018-01-04 深圳狗尾草智能科技有限公司 一种基于用户画像的意图场景识别方法及***
CN107589826B (zh) * 2016-07-07 2019-11-05 苏州狗尾草智能科技有限公司 基于知识图谱的人机交互方法及***
CN106202417A (zh) * 2016-07-12 2016-12-07 北京光年无限科技有限公司 一种用于智能机器人的人机交互方法及***
CN106227740A (zh) * 2016-07-12 2016-12-14 北京光年无限科技有限公司 一种面向对话***的数据处理方法及装置
CN106202476B (zh) * 2016-07-14 2017-06-06 广州安望信息科技有限公司 一种基于知识图谱的人机对话的方法及装置
CN106293348B (zh) * 2016-08-16 2020-02-18 东软集团股份有限公司 应用于新媒体的信息交互方法和装置
CN108021572B (zh) * 2016-11-02 2020-08-07 腾讯科技(深圳)有限公司 回复信息推荐方法和装置
US10481861B2 (en) * 2016-08-30 2019-11-19 Google Llc Using user input to adapt search results provided for presentation to the user
US9947319B1 (en) 2016-09-27 2018-04-17 Google Llc Forming chatbot output based on user state
CN106448670B (zh) * 2016-10-21 2019-11-19 竹间智能科技(上海)有限公司 基于深度学习和强化学习的自动回复对话***
CN108021547B (zh) * 2016-11-04 2021-05-04 株式会社理光 一种自然语言的生成方法、自然语言生成装置及电子设备
CN106779817A (zh) * 2016-11-29 2017-05-31 竹间智能科技(上海)有限公司 基于多维度信息的意图识别方法及***
CN106708983A (zh) * 2016-12-09 2017-05-24 竹间智能科技(上海)有限公司 基于对话交互信息构建用户画像的***及方法
CN106776557B (zh) * 2016-12-13 2020-09-08 竹间智能科技(上海)有限公司 情感机器人的情感状态记忆识别方法及装置
CN106649762A (zh) * 2016-12-27 2017-05-10 竹间智能科技(上海)有限公司 基于追问问题与反馈信息的意图识别方法、***
CN106844646B (zh) * 2017-01-22 2021-03-19 竹间智能科技(上海)有限公司 基于情绪预测的人工智能回话方法及装置
CN108460033A (zh) * 2017-02-17 2018-08-28 微软技术许可有限责任公司 在自动聊天中提供响应
US11341174B2 (en) 2017-03-24 2022-05-24 Microsoft Technology Licensing, Llc Voice-based knowledge sharing application for chatbots
CN108664336A (zh) * 2017-04-01 2018-10-16 北京搜狗科技发展有限公司 推荐方法和装置、用于推荐的装置
CN108766421B (zh) * 2017-04-20 2020-09-15 杭州萤石网络有限公司 语音交互的方法及装置
US11227342B2 (en) 2017-05-26 2022-01-18 Microsoft Technology Licensing, Llc Recommending friends in automated chatting
TWI646529B (zh) * 2017-07-25 2019-01-01 雲拓科技有限公司 主動聊天裝置
CN110020007A (zh) * 2017-09-15 2019-07-16 上海挖数互联网科技有限公司 机器人交互控制方法及装置、存储介质、服务器
CN107665362B (zh) * 2017-09-26 2019-10-29 联想(北京)有限公司 用于实现机器人聊天的训练方法、预测答案的方法及装置
CN107844481B (zh) * 2017-11-21 2019-09-13 新疆科大讯飞信息科技有限责任公司 识别文本检错方法及装置
CN108170749B (zh) * 2017-12-21 2021-06-11 北京百度网讯科技有限公司 基于人工智能的对话方法、装置及计算机可读介质
CN110019722B (zh) * 2017-12-21 2023-11-24 株式会社理光 对话模型的回复排序方法、装置及计算机可读存储介质
CN109994108B (zh) * 2017-12-29 2023-08-29 微软技术许可有限责任公司 用于聊天机器人和人之间的会话交谈的全双工通信技术
CN109995642A (zh) * 2017-12-29 2019-07-09 Tcl集团股份有限公司 一种自动生成快捷回复的方法及装置、即时通讯***
CN108415932B (zh) 2018-01-23 2023-12-22 思必驰科技股份有限公司 人机对话方法及电子设备
CN109086329B (zh) * 2018-06-29 2021-01-05 出门问问信息科技有限公司 基于话题关键词引导的进行多轮对话方法及装置
CN109074502A (zh) * 2018-07-26 2018-12-21 深圳前海达闼云端智能科技有限公司 训练人工智能模型的方法、装置、存储介质及机器人
JP6600398B1 (ja) * 2018-09-15 2019-10-30 株式会社ブロードリーフ Ai創作物の検証装置
CN109376282A (zh) * 2018-09-26 2019-02-22 北京子歌人工智能科技有限公司 一种基于人工智能的人机智能聊天的方法和装置
CN109522927A (zh) * 2018-10-09 2019-03-26 北京奔影网络科技有限公司 用于用户消息的情感分析方法及装置
CN111381685B (zh) * 2018-12-29 2024-03-22 北京搜狗科技发展有限公司 一种句联想方法和装置
CN109670030B (zh) * 2018-12-30 2022-06-28 联想(北京)有限公司 问答交互方法及装置
CN109947895A (zh) * 2019-02-20 2019-06-28 徐江涛 一种对话聊天模式的记账***
CN110674256B (zh) * 2019-09-25 2023-05-12 携程计算机技术(上海)有限公司 Ota酒店的点评与回复的相关度的检测方法及***
CN111161708A (zh) * 2019-12-31 2020-05-15 中国银行股份有限公司 语音信息处理方法及装置
CN111241259B (zh) * 2020-01-08 2023-06-20 百度在线网络技术(北京)有限公司 交互信息推荐方法和装置
CN111930908A (zh) * 2020-08-10 2020-11-13 腾讯科技(深圳)有限公司 基于人工智能的答案识别方法及装置、介质、电子设备
CN112337094B (zh) * 2020-10-27 2022-11-25 四川长虹电器股份有限公司 语音游戏交互方法
CN112597287B (zh) * 2020-12-15 2024-03-15 深圳市优必选科技股份有限公司 一种语句处理方法、语句处理装置及智能设备
CN112883176B (zh) * 2021-02-10 2022-08-09 华南师范大学 基于大数据面向专业职业的人工智能心理健康聊天机器人
CN116541504B (zh) * 2023-06-27 2024-02-06 北京聆心智能科技有限公司 对话生成方法、装置、介质及计算设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050004788A1 (en) * 2003-07-03 2005-01-06 Lee Hang Shun Raymond Multi-level confidence measures for task modeling and its application to task-oriented multi-modal dialog management
CN101221583A (zh) * 2008-01-29 2008-07-16 北京百问百答网络技术有限公司 一种问题推荐方法及***
CN104598445A (zh) * 2013-11-01 2015-05-06 腾讯科技(深圳)有限公司 自动问答***和方法
CN104714954A (zh) * 2013-12-13 2015-06-17 中国电信股份有限公司 基于上下文理解的信息搜索方法和***
CN105068661A (zh) * 2015-09-07 2015-11-18 百度在线网络技术(北京)有限公司 基于人工智能的人机交互方法和***

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050004788A1 (en) * 2003-07-03 2005-01-06 Lee Hang Shun Raymond Multi-level confidence measures for task modeling and its application to task-oriented multi-modal dialog management
CN101221583A (zh) * 2008-01-29 2008-07-16 北京百问百答网络技术有限公司 一种问题推荐方法及***
CN104598445A (zh) * 2013-11-01 2015-05-06 腾讯科技(深圳)有限公司 自动问答***和方法
CN104714954A (zh) * 2013-12-13 2015-06-17 中国电信股份有限公司 基于上下文理解的信息搜索方法和***
CN105068661A (zh) * 2015-09-07 2015-11-18 百度在线网络技术(北京)有限公司 基于人工智能的人机交互方法和***

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110209897A (zh) * 2018-02-12 2019-09-06 腾讯科技(深圳)有限公司 智能对话方法、装置、存储介质及设备
CN110209897B (zh) * 2018-02-12 2023-09-01 腾讯科技(深圳)有限公司 智能对话方法、装置、存储介质及设备
US10826864B2 (en) 2018-07-27 2020-11-03 At&T Intellectual Property I, L.P. Artificially intelligent messaging
CN109190114B (zh) * 2018-08-13 2022-06-07 北京百度网讯科技有限公司 用于生成回复信息的方法和装置
CN109190114A (zh) * 2018-08-13 2019-01-11 北京百度网讯科技有限公司 用于生成回复信息的方法和装置
CN109242706A (zh) * 2018-08-20 2019-01-18 中国平安人寿保险股份有限公司 辅助坐席人员沟通的方法、装置、计算机设备和存储介质
US11397852B2 (en) * 2018-12-17 2022-07-26 Beijing Baidu Netcom Science and Technology Co., Ltd News interaction method, apparatus, device and computer storage medium
CN111353091A (zh) * 2018-12-24 2020-06-30 北京三星通信技术研究有限公司 信息处理方法、装置、电子设备及可读存储介质
CN111078893A (zh) * 2019-12-11 2020-04-28 竹间智能科技(上海)有限公司 一种大规模高效获取识别对话意图用语料的方法
CN113127613A (zh) * 2020-01-10 2021-07-16 北京搜狗科技发展有限公司 聊天信息处理方法及装置
CN113127613B (zh) * 2020-01-10 2024-01-09 北京搜狗科技发展有限公司 聊天信息处理方法及装置
CN111291170A (zh) * 2020-01-20 2020-06-16 腾讯科技(深圳)有限公司 一种基于智能客服的会话推荐方法及相关装置
CN111291170B (zh) * 2020-01-20 2023-09-19 腾讯科技(深圳)有限公司 一种基于智能客服的会话推荐方法及相关装置
CN112967721A (zh) * 2021-02-03 2021-06-15 上海明略人工智能(集团)有限公司 一种基于语音识别技术的销售线索信息识别方法与***
CN112967721B (zh) * 2021-02-03 2024-05-31 上海明略人工智能(集团)有限公司 一种基于语音识别技术的销售线索信息识别方法与***
CN116628153A (zh) * 2023-05-10 2023-08-22 上海任意门科技有限公司 一种人工智能设备对话的控制方法、装置、设备及介质
CN116628153B (zh) * 2023-05-10 2024-03-15 上海任意门科技有限公司 一种人工智能设备对话的控制方法、装置、设备及介质
CN117216229A (zh) * 2023-11-08 2023-12-12 支付宝(杭州)信息技术有限公司 一种生成客服答案的方法及装置

Also Published As

Publication number Publication date
CN105183848A (zh) 2015-12-23

Similar Documents

Publication Publication Date Title
WO2017041370A1 (zh) 基于人工智能的人机聊天方法和装置
CN105068661B (zh) 基于人工智能的人机交互方法和***
Ghazvininejad et al. A knowledge-grounded neural conversation model
CN107943998B (zh) 一种基于知识图谱的人机对话控制***及方法
CN106227815B (zh) 一种多模态线索的个性化应用程序功能推荐方法及其***
CN104933113B (zh) 一种基于语义理解的表情输入方法和装置
CN112074857A (zh) 组合机器学习和社交数据以生成个性化推荐
CN105808590B (zh) 搜索引擎实现方法、搜索方法以及装置
US20160012106A1 (en) User interface for summarizing the relevance of a document to a query
CN112313643A (zh) 基于检索/创建的数字媒体文件预测潜在相关的主题
EP2511869A2 (en) Method and system for providing user-customized content
CN110263180A (zh) 意图知识图谱生成方法、意图识别方法及装置
CN106940726B (zh) 一种基于知识网络的创意自动生成方法与终端
CN113934941B (zh) 一种基于多维度信息的用户推荐***及方法
CN111460145A (zh) 一种学习资源推荐方法、设备及存储介质
CN115470344A (zh) 一种基于文本聚类的视频弹幕与评论主题融合的方法
WO2017034818A1 (en) Method and system for matching people with choices
US20170098168A1 (en) Method and system for providing synthetic answers to a personal question
Hoyt et al. PodcastRE Analytics: Using RSS to Study the Cultures and Norms of Podcasting.
CN107977883A (zh) 旅行包的推荐方法、装置及计算机设备
Parizi et al. EmoNews: an Emotional News Recommender System.
CN115550304B (zh) 确定一组用户的活动实例集合的方法、设备和存储介质
Elahi Empirical evaluation of active learning strategies in collaborative filtering
Bellogín et al. Discovering relevant preferences in a personalised recommender system using machine learning techniques
CN114896391A (zh) 基于任务提示的小样本句型分类方法、***、设备及介质

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: 15903473

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: 15903473

Country of ref document: EP

Kind code of ref document: A1