CN105183848A - Human-computer chatting method and device based on artificial intelligence - Google Patents

Human-computer chatting method and device based on artificial intelligence Download PDF

Info

Publication number
CN105183848A
CN105183848A CN201510564173.0A CN201510564173A CN105183848A CN 105183848 A CN105183848 A CN 105183848A CN 201510564173 A CN201510564173 A CN 201510564173A CN 105183848 A CN105183848 A CN 105183848A
Authority
CN
China
Prior art keywords
chat
module
sentence
user
information
Prior art date
Application number
CN201510564173.0A
Other languages
Chinese (zh)
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
Priority to CN201510564173.0A priority Critical patent/CN105183848A/en
Publication of CN105183848A publication Critical patent/CN105183848A/en

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/903—Querying
    • G06F16/90335—Query processing
    • G06F16/90348—Query processing by searching ordered data, e.g. alpha-numerically ordered data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation

Abstract

The invention discloses a human-computer chatting method and device based on artificial intelligence. The method includes the following steps of receiving input information input by a user; distributing the input information to chatting service modules; receiving candidate replying returned by the multiple chatting service modules; ranking the to-be-selected replying based on confidence degrees, generating chatting information according to the ranking result, and providing the chatting information for the user. According to the human-computer chatting method and device based on the artificial intelligence, the input information input by the user is received and distributed to the chatting service modules, then the candidate replying returned by the multiple chatting service modules is received, the to-be-selected replying is ranked based on the confidence degrees, and the chatting information is generated according to the ranking result and provided for the user, so that multiple times of chatting with the user can be carried out, the true effect and the natural effect are achieved, initiative is achieved, replying conforming to user styles can be further returned for different users, and the more personalized effect and the more intelligent effect are achieved.

Description

Based on man-machine chat method and the device of artificial intelligence

Technical field

The present invention relates to field of artificial intelligence, particularly relate to a kind of man-machine chat method based on artificial intelligence and device.

Background technology

Artificial intelligence (ArtificialIntelligence) is a branch of computer science, english abbreviation is AI, is research, develops the theory of intelligence for simulating, extending and expand people, method, one of application system new technological sciences.

Man-machine chat refers to the process of carrying out interactive chat between people and computing machine.Man-machine chat can be applicable to the aspects such as amusement and emotion are accompanied, Intelligent Service personalizes.Such as: can be engaged in the dialogue whenever and wherever possible by man-machine chat system, alleviate the life stress of people, children also can be helped to improve language ability for children.

At present, the chat sentence of traditional man-machine chat system mainly based on extensive automatic mining is right, and each chat sentence centering contains the upper sentence P (post) of chat and the lower sentence R (Response) for P.For the chat sentence Q (query) of user's input, first calculate with Q similarity the most much higher chat go up sentence P1, P2 ... Pn}, then to sentence { R1, R2 under the chat of the upper sentence correspondence of chat,, Rn} sorts, and under then selecting optimum chat, sentence R returns to user.

But traditional man-machine chat system exists following shortcoming: lack and take turns chat capability more, namely user proposes upper sentence of chatting, the lower sentence of chat replied by machine, lacks initiative, inadequate true nature; For the dialogue such as news that ageing requirement is higher, then cannot answer accurately; In addition, lacking individuality of chat content, the same or similar problem that different user proposes, can only reply same answer, lacking individuality and intellectuality.

Summary of the invention

The present invention is intended to solve one of technical matters in correlation technique at least to a certain extent.For this reason, one object of the present invention is to propose a kind of man-machine chat method based on artificial intelligence, can carry out taking turns chat with user more, true nature, there is initiative, can also return for different user the reply meeting user styles, more personalized, intelligent.

Second object of the present invention is to propose a kind of man-machine chat device based on artificial intelligence.

To achieve these goals, first aspect present invention embodiment proposes a kind of man-machine chat method based on artificial intelligence, comprising: the input information receiving user's input; By described input distribution of information to chatting service module; Receive candidate's reply that described multiple chatting service module returns, wherein, described candidate replys has corresponding degree of confidence; Based on described degree of confidence, described reply to be selected is sorted, and generate chat message according to ranking results, and provide described chat message to described user.

The man-machine chat method based on artificial intelligence of the embodiment of the present invention, by receiving the input information of user's input, and distribution of information will be inputted to chatting service module, then candidate's reply that multiple chatting service module returns is received, and based on degree of confidence, reply to be selected is sorted, and generate chat message according to ranking results, and provide chat message to user, can carry out taking turns chat with user more, true nature, there is initiative, can also return for different user the reply meeting user styles, more personalized, intelligent.

Second aspect present invention embodiment proposes a kind of man-machine chat device based on artificial intelligence, comprising: the first receiver module, for receiving the input information of user's input; Distribution module, for by described input distribution of information to chatting service module; Second receiver module, the candidate returned for receiving described multiple chatting service module replys, and wherein, described candidate replys has corresponding degree of confidence; Module is provided, for sorting to described reply to be selected based on described degree of confidence, and generates chat message according to ranking results, and provide described chat message to described user.

The man-machine chat device based on artificial intelligence of the embodiment of the present invention, by receiving the input information of user's input, and distribution of information will be inputted to chatting service module, then candidate's reply that multiple chatting service module returns is received, and based on degree of confidence, reply to be selected is sorted, and generate chat message according to ranking results, and provide chat message to user, can carry out taking turns chat with user more, true nature, there is initiative, can also return for different user the reply meeting user styles, more personalized, intelligent.

Accompanying drawing explanation

Fig. 1 is according to an embodiment of the invention based on the process flow diagram of the man-machine chat method of artificial intelligence.

Fig. 2 is the effect schematic diagram of topic chat collection of illustrative plates according to an embodiment of the invention.

Fig. 3 is according to an embodiment of the invention based on the structural representation one of the man-machine chat device of artificial intelligence.

Fig. 4 is according to an embodiment of the invention based on the structural representation two of the man-machine chat device of artificial intelligence.

Fig. 5 is according to an embodiment of the invention based on the structural representation of the chat module of search.

Fig. 6 is the structural representation of rich according to an embodiment of the invention knowledge chat module.

Fig. 7 is according to an embodiment of the invention based on the structural representation one of the chat module of portrait.

Fig. 8 is according to an embodiment of the invention based on the structural representation two of the chat module of portrait.

Fig. 9 is according to an embodiment of the invention based on the structural representation three of the man-machine chat device of artificial intelligence.

Figure 10 is according to an embodiment of the invention based on the structural representation four of the man-machine chat device of artificial intelligence.

Figure 11 is according to an embodiment of the invention based on the structural representation five of the man-machine chat device of artificial intelligence.

Figure 12 is according to an embodiment of the invention based on the structural representation six of the man-machine chat device of artificial intelligence.

Embodiment

Be described below in detail embodiments of the invention, the example of described embodiment is shown in the drawings, and wherein same or similar label represents same or similar element or has element that is identical or similar functions from start to finish.Be exemplary below by the embodiment be described with reference to the drawings, be intended to for explaining the present invention, and can not limitation of the present invention be interpreted as.

Below with reference to the accompanying drawings man-machine chat method based on artificial intelligence and the device of the embodiment of the present invention are described.

Fig. 1 is according to an embodiment of the invention based on the process flow diagram of the man-machine chat method of artificial intelligence.

As shown in Figure 1, the man-machine chat method based on artificial intelligence can comprise:

The input information of S1, reception user input.

Wherein, input information can be voice messaging, also can be text message.

After the input information receiving user's input, error correction and/or rewriting can being carried out to input information, for correcting the wrongly written or mispronounced characters in input information, rewriting irregular colloquial style expression etc.

In addition, also can obtain the information above of chatting with user, then judge whether the dependence of input information and information is above greater than preset relation threshold value according to information above.If be greater than preset relation threshold value, then can carry out completion according to information above to input information, thus ensure the fluency of man-machine chat.Particularly, completion is carried out to input information and can comprise reference resolution.For example, input information for " he has married? ", then according to information " Liu Dehua " above, " he " in input information can be replaced by " Liu Dehua ".Completion is carried out to input information and also can comprise omission completion.For example, above information " Liu De China wife be Zhu Liqian.", input information is not for " I is familiar with.", then can by input information completion for " I be familiar with Zhu Liqian.”。

In addition, topic information that also can be current according to acquisition of information user above, so that follow-up chatting service module guides chat topic.

S2, distribution of information will be inputted to chatting service module.

Particularly, domain analysis can be carried out to obtain field corresponding to input information to input information.Then, distribution of information will can be inputted to the chatting service module with identical or close field in the field corresponding according to input information.

Wherein, chatting service module can comprise based on search chat module, rich knowledge chat module, based on portrait chat module and based on one or more in the chat module of mass-rent.

Particularly, chat module based on search can cut word to generate multiple chat short sentence to input information, then can sentence according to multiple chat short sentence inquiry chat corpus thus under generating multiple chat language materials of sentence correspondence on sentence and multiple chat language material on multiple chat language material.Wherein, chat corpus is for set up in advance, and language is chatted to originate and can be included but not limited to "-the money order receipt to be signed and returned to the sender of posting " in the forum data such as mhkc, " blog article-reply " in microblogging, " problem-answer " in Ask-Answer Community etc.

After this, can filter sentence on multiple chat language material.Particularly, the similarity between sentence on input information and multiple chat language material can be calculated.If similarity is less than first preset similarity threshold, then sentence on the chat language material of correspondence can be filtered; If similarity is more than or equal to first preset similarity threshold, then sentence on the chat language material of correspondence can be retained.

To after on chat language material, sentence filters, can classify to sentence under the chat language material of sentence correspondence on the chat language material after filtering.Particularly, calculate the similarity between sentence under input information and multiple chat language material, and according to similarity based on GBDT (gradient boost decision tree, GradientBoostDecisionTree), the machine learning model such as SVM (support vector machine, SupportVectorMachine) is classified to sentence under multiple chat language material.Wherein, similarity under input information and multiple chat language material between sentence can be similarity literal between sentence under input information and chat language material, also can be that input information trains with sentence under chat language material the similarity obtained based on deep neural network, also can be that input information trains with sentence under chat language material the similarity obtained based on Machine Translation Model.Should be understood that, the machine learning model such as similarity and GBDT, SVM in the present embodiment under input information and multiple chat language material between sentence is known technology, does not repeat herein.

Then the chat module based on search can reorder to sentence under the chat language material after classification, and generates candidate's reply according to ranking results.Particularly, according to the chat attribute of the user of acquisition of information above of user's chat, then according to chat attribute, sentence under the chat language material after classification can be reordered based on study order models (Learning-To-Rank).Wherein, the occasion that attribute of chatting can comprise chat is as time place etc., the interest of chat, the style etc. of chat.Certainly, chat attribute is not limited only to obtain from the information above of user's chat, also can obtain by the history chat record long-term according to user.Should be understood that, the present embodiment learning order models is known technology, does not repeat herein.

Rich knowledge chat module can generate search word according to input information, and carry out searching for generate multiple Search Results according to search word, then sentence extraction is carried out to multiple Search Results, be greater than with the similarity of search word the candidate sentences set that second presets the sentence of similarity threshold to obtain.After this, can rewrite to generate candidate to the sentence in candidate sentences set to reply.In addition, also can reorder to the sentence in candidate sentences set according to the chat attribute of user.For example, input information is " wishing that chance can arrive Fuji and travel ", can resolve input information and generate corresponding search word " Fuji, tourism ", then obtaining multiple Search Results according to search word, and extracting the sentence high with search word similarity.Wherein, some sentences may comprise as " reporter learned " etc. obviously selects from web page text, therefore need to rewrite these sentences, make it more smooth, more as the sentence of natural language chat, the final candidate generated replys as " Fuji, due to weather reason, only has a period of time in the summer of regulation to climb the mountain in 1 year ", relative to traditional reply, " I also thinks Fuji, together.", there is certain intellectual, and have necessarily ageing, user can be made to obtain useful knowledge in chat process.

Personalize to realize better, and provide personalized service for user, man-machine chat system can set self attribute, state, interest etc., and namely system draws a portrait model.Also can set the attribute, state, interest etc. of user, namely user draws a portrait model.Certainly, when different users, the system portrait model of use can be same, also can arrange system portrait model corresponding with it per family for each use.System portrait model and user draw a portrait model all based on portrait knowledge mapping.Portrait knowledge mapping is the knowledge hierarchy of a stratification.For example, " kinsfolk " node can comprise " siblings " and " father and mother " two child nodes, and " father and mother " child node comprises " father " and " mother " two child nodes.Each node is all to there being multiple input information model bunch, and such as " whom your father is ", " who be you father ", " what your father cries " belong to same input information model bunch.Each input information model bunch correspondence one or more candidate reply.Input information model bunch and candidate reply and can comprise variable, such as interest, hobby, the corresponding same attribute " INTEREST " of hobby, and the property value of " INTEREST " can comprise climb the mountain, music, reading, motion etc.

Particularly, the chat module based on portrait can obtain the chat linguistic context of user, and judges whether to meet collection condition according to chat linguistic context.If judge to meet collection condition, then can send problem to user.After this, the answer information of user according to problem can be received, and according to answer information, model be drawn a portrait to user and upgrade.Such as: when chatting the relevant topic of film to user, can to user send problem " what film you like? " or user ask man-machine chat system " you like what is eaten? " man-machine chat system can ask in reply user " you like what is eaten? " after user answers, model can be drawn a portrait based on the answer information of user to user to upgrade, more meet the demand of user individual.

In addition, the chat module based on portrait also can obtain the chat content of user, and extracts user's representation data according to chat content, then draws a portrait model according to the user's representation data extracted to user and upgrades.Such as: user says in chat process, and " I likes when having nothing to do climbing the mountain, fishing fishing.", user's representation data " hobby is climbed the mountain, love fishing " can be extracted, thus model drawn a portrait to user upgrade.Meanwhile, suitable answer can be extracted based on user's representation data, return suitable answer information to user.

Mass-rent (crowdsourcing) is a kind of method particular task being contracted out to non-user-specific in internet, for in man-machine chat, machine is difficult to the problem answered, and can be distributed to executor and manually reply in real time online, thus meet the actual demand of user.

Particularly, the chat module based on mass-rent can judge whether input information is applicable to mass-rent and completes, and the low needs comfort of such as user emotion etc., be then applicable to mass-rent and complete.Such as, include the privacy informations such as personally identifiable information, password, phone in the input information of user, be then not suitable for mass-rent and complete.

If judge that being applicable to mass-rent completes, then can by input distribution of information to corresponding executor.Certainly, simultaneously also information above together can be sent to executor, executor can according to information and input information are replied above.Then can receive the return information of executor based on the chat module of mass-rent, and Quality estimation is carried out to return information.If meet quality requirements, then return information is alternatively replied.Such as: if comprise vulgar, reaction, Pornograph in return information, then quality does not pass a test.Or the time that executor replys has exceeded scheduled duration, then the return information of this executor will not be used, this return information can be saved in chat corpus simultaneously.

Outside this, also can judge whether input information belongs to the chat message without actual content, as " laughing a great ho-ho ", " hoho " etc.If judge it is belong to the chat message without actual content, then can obtain actualite, namely calculate actualite based on topic model (TopicModel) according to history chat record.After acquisition actualite, collection of illustrative plates can be chatted according to actualite generation guiding topic based on topic.Wherein, the digraph of topic chat collection of illustrative plates to be one with topic be node.Such as, as shown in Figure 2, node " leisure " can point to node and " to see a film " and node " listens song ", then illustrate that can guide to topic from topic " leisure " " sees a film " or topic " listens song ".Topic " is seen a film " and topic " is listened song " and all had certain Guiding probability, can realize the guiding of topic according to Guiding probability, thus ensures the diversity guiding topic.

Then, candidate can be generated according to guiding topic to reply.Particularly, based on spatial term model (NaturalLanguageGeneration), the template that candidate replys can be generated, guiding topic is filled in this template and generate candidate's reply; Also can choose to comprise guides the sentence of topic alternatively to reply, thus realizes carrying out chat topic guiding on one's own initiative to user.

S3, receive the candidate that multiple chatting service module returns and reply.

Wherein, candidate replys and has corresponding degree of confidence.

S4, based on degree of confidence, reply to be selected to be sorted, and generate chat message according to ranking results, and provide chat message to user.

Particularly, the feature of the input information of user can be obtained, and based on the characteristic sum degree of confidence of input information, reply to be selected be sorted.Wherein, the feature inputting information can comprise characteristic of division, literal feature, topic feature etc.Degree of confidence is higher, then reply quality to be selected is better, can sort, provide the chat message meeting user's request eventually to user according to degree of confidence order from high to low to reply to be selected.

In addition, also upgrade according to the feedback information of user by strengthening learning model (ReinforcementLearning), thus more satisfied chat message can be provided for user.Such as: in the chat message of replying user, add review button if " praising " or " stepping on " is to collect the feedback information of user; Or based on sentiment analysis technology, the input information of user in chat is analyzed, thus obtain the evaluation of user, such as: " you are intelligence very " etc.; Or by the interaction times that record is chatted with user, judge the satisfaction of user.

The man-machine chat method based on artificial intelligence of the embodiment of the present invention, by receiving the input information of user's input, and distribution of information will be inputted to chatting service module, then candidate's reply that multiple chatting service module returns is received, and based on degree of confidence, reply to be selected is sorted, and generate chat message according to ranking results, and provide chat message to user, can carry out taking turns chat with user more, true nature, there is initiative, can also return for different user the reply meeting user styles, more personalized, intelligent.

For achieving the above object, the present invention also proposes a kind of man-machine chat device based on artificial intelligence.

Fig. 3 is according to an embodiment of the invention based on the structural representation one of the man-machine chat device of artificial intelligence.

As shown in Figure 3, can should comprise based on the man-machine chat device of artificial intelligence: the first receiver module 1000, distribution module 2000, chatting service module 3000, second receiver module 4000 and module 5000 is provided.

First receiver module 1000 is for receiving the input information of user's input.

Wherein, input information can be voice messaging, also can be text message.

Distribution module 2000 will be for inputting distribution of information to chatting service module 3000.

Candidate's reply that second receiver module 4000 returns for receiving multiple chatting service module 3000.Wherein, candidate replys and has corresponding degree of confidence.

There is provided module 5000 for sorting to reply to be selected based on degree of confidence, and generate chat message according to ranking results, and provide chat message to user.

Particularly, the feature of the input information providing module 5000 can obtain user, and based on the characteristic sum degree of confidence of input information, reply to be selected is sorted.Wherein, the feature inputting information can comprise characteristic of division, literal feature, topic feature etc.Degree of confidence is higher, then reply quality to be selected is better, can sort, provide the chat message meeting user's request eventually to user according to degree of confidence order from high to low to reply to be selected.

As shown in Figure 4, chatting service module 3000 can comprise based on the chat module 3100 of search, rich knowledge chat module 3200, based on the chat module 3300 of portrait and chat module 3400 based on mass-rent.

Wherein, as shown in Figure 5, based on the chat module 3100 of search can comprise cut lexon module 3110, generate submodule 3120, filter submodule 3130, classification submodule 3140 and sorting sub-module 3150.Wherein, filter submodule 3130 and can comprise computing unit 3131, filter element 3132, stick unit 3133, classification submodule 3140 can comprise computing unit 3141, taxon 3142, and sorting sub-module 3150 can comprise acquiring unit 3151 and sequencing unit 3152.

Particularly, cut lexon module 3110 and can cut word to generate multiple chat short sentence to input information, then generating submodule 3120 can sentence according to multiple chat short sentence inquiry chat corpus thus under generating multiple chat language materials of sentence correspondence on sentence and multiple chat language material on multiple chat language material.Wherein, chat corpus is for set up in advance, and language is chatted to originate and can be included but not limited to "-the money order receipt to be signed and returned to the sender of posting " in the forum data such as mhkc, " blog article-reply " in microblogging, " problem-answer " in Ask-Answer Community etc.

After this, filter submodule 3130 to filter sentence on multiple chat language material.Particularly, computing unit 3131 can calculate the similarity on input information and multiple chat language material between sentence.If similarity is less than first preset similarity threshold, then sentence on the chat language material of correspondence can filter by filter element 3132; If similarity is more than or equal to first preset similarity threshold, then sentence on the chat language material of correspondence can retain by stick unit 3133.

To after on chat language material, sentence filters, classification submodule 3140 can be classified to sentence under the chat language material of sentence correspondence on the chat language material after filtering.Particularly, computing unit 3141 can calculate the similarity under input information and multiple chat language material between sentence, taxon 3142 according to similarity based on GBDT (gradient boost decision tree, GradientBoostDecisionTree), the machine learning model such as SVM (support vector machine, SupportVectorMachine) is classified to sentence under multiple chat language material.Wherein, similarity under input information and multiple chat language material between sentence can be similarity literal between sentence under input information and chat language material, also can be that input information trains with sentence under chat language material the similarity obtained based on deep neural network, also can be that input information trains with sentence under chat language material the similarity obtained based on Machine Translation Model.Should be understood that, the machine learning model such as similarity and GBDT, SVM in the present embodiment under input information and multiple chat language material between sentence is known technology, does not repeat herein.

Then sorting sub-module 3150 can reorder to sentence under the chat language material after classification, and generates candidate's reply according to ranking results.Particularly, acquiring unit 3151 can according to the chat attribute of the user of acquisition of information above of user's chat, and sequencing unit 3152 reorders based on study order models (Learning-To-Rank) to sentence under the chat language material after classification according to chat attribute.Wherein, the occasion that attribute of chatting can comprise chat is as time place etc., the interest of chat, the style etc. of chat.Certainly, chat attribute is not limited only to obtain from the information above of user's chat, also can obtain by the history chat record long-term according to user.Should be understood that, the present embodiment learning order models is known technology, does not repeat herein.

As shown in Figure 6, rich knowledge chat module 3200 can comprise generation submodule 3210, extracts submodule 3220, rewrite submodule 3230 and the submodule 3240 that reorders.Particularly, generate submodule 3210 and can generate search word according to input information, and carry out searching for generate multiple Search Results according to search word, then extract submodule 3220 and sentence extraction is carried out to multiple Search Results, be greater than with the similarity of search word the candidate sentences set that second presets the sentence of similarity threshold to obtain.After this, rewrite submodule 3230 to rewrite the sentence in candidate sentences set to generate candidate's reply.In addition, the submodule 3240 that reorders can reorder to the sentence in candidate sentences set according to the chat attribute of user.For example, input information is " wishing that chance can arrive Fuji and travel ", can resolve input information and generate corresponding search word " Fuji, tourism ", then obtaining multiple Search Results according to search word, and extracting the sentence high with search word similarity.Wherein, some sentences may comprise as " reporter learned " etc. obviously selects from web page text, therefore need to rewrite these sentences, make it more smooth, more as the sentence of natural language chat, the final candidate generated replys as " Fuji, due to weather reason, only has a period of time in the summer of regulation to climb the mountain in 1 year ", relative to traditional reply, " I also thinks Fuji, together.", there is certain intellectual, and have necessarily ageing, user can be made to obtain useful knowledge in chat process.

As shown in Figure 7, the chat module 3300 based on portrait can comprise the first acquisition submodule 3310, judges submodule 3320, send submodule 3330, first renewal submodule 3340.

Personalize to realize better, and provide personalized service for user, man-machine chat system can set self attribute, state, interest etc., and namely system draws a portrait model.Also can set the attribute, state, interest etc. of user, namely user draws a portrait model.Certainly, when different users, the system portrait model of use can be same, also can arrange system portrait model corresponding with it per family for each use.System portrait model and user draw a portrait model all based on portrait knowledge mapping.Portrait knowledge mapping is the knowledge hierarchy of a stratification.For example, " kinsfolk " node can comprise " siblings " and " father and mother " two child nodes, and " father and mother " child node comprises " father " and " mother " two child nodes.Each node is all to there being multiple input information model bunch, and such as " whom your father is ", " who be you father ", " what your father cries " belong to same input information model bunch.Each input information model bunch correspondence one or more candidate reply.Input information model bunch and candidate reply and can comprise variable, such as interest, hobby, the corresponding same attribute " INTEREST " of hobby, and the property value of " INTEREST " can comprise climb the mountain, music, reading, motion etc.

Particularly, first obtains the chat linguistic context that submodule 3310 can obtain user, judges that submodule 3320 judges whether to meet collection condition according to chat linguistic context.If judge to meet collection condition, then send submodule 3330 and can send problem to user.After this, first upgrades submodule 3340 can receive the answer information of user according to problem, and draws a portrait model according to answer information to user and upgrade.Such as: when chatting the relevant topic of film to user, can to user send problem " what film you like? " or user ask man-machine chat system " you like what is eaten? " man-machine chat system can ask in reply user " you like what is eaten? " after user answers, model can be drawn a portrait based on the answer information of user to user to upgrade, more meet the demand of user individual.

In addition, as shown in Figure 8, the chat module 3300 based on portrait also can comprise the second acquisition submodule 3350, extract submodule 3360, second renewal submodule 3370.

Particularly, second obtains submodule 3350 can obtain the chat content of user, extracts submodule 3360 and extracts user's representation data according to chat content, and then second upgrades submodule 3370 and draw a portrait model according to the user's representation data extracted to user and upgrade.Such as: user says in chat process, and " I likes when having nothing to do climbing the mountain, fishing fishing.", user's representation data " hobby is climbed the mountain, love fishing " can be extracted, thus model drawn a portrait to user upgrade.Meanwhile, suitable answer can be extracted based on user's representation data, return suitable answer information to user.

Mass-rent (crowdsourcing) is a kind of method particular task being contracted out to non-user-specific in internet, for in man-machine chat, machine is difficult to the problem answered, and can be distributed to executor and manually reply in real time online, thus meet the actual demand of user.

Particularly, the chat module 3400 based on mass-rent can judge whether input information is applicable to mass-rent and completes, and the low needs comfort of such as user emotion etc., be then applicable to mass-rent and complete.Such as, include the privacy informations such as personally identifiable information, password, phone in the input information of user, be then not suitable for mass-rent and complete.

If judge that being applicable to mass-rent completes, then the chat module 3400 based on mass-rent can by input distribution of information to corresponding executor.Certainly, simultaneously also information above together can be sent to executor, executor can according to information and input information are replied above.Then can receive the return information of executor based on the chat module of mass-rent, and Quality estimation is carried out to return information.If meet quality requirements, then return information is alternatively replied.Such as: if comprise vulgar, reaction, Pornograph in return information, then quality does not pass a test.Or the time that executor replys has exceeded scheduled duration, then the return information of this executor will not be used, this return information can be saved in chat corpus simultaneously.

In addition, as shown in Figure 9, also correction module 6000 can should be comprised based on the man-machine chat device of artificial intelligence.

Correction module 6000, for after the input information receiving user's input, being carried out error correction and/or rewriting to input information, for correcting the wrongly written or mispronounced characters in input information, being rewritten irregular colloquial style expression etc.

In addition, as shown in Figure 10, also analysis module 7000 can should be comprised based on the man-machine chat device of artificial intelligence.

Analysis module 7000 is for after the input information receiving user's input, carry out domain analysis to obtain field corresponding to input information to input information, then distribution module 2000 can will input distribution of information to the chatting service module with identical or close field in the field corresponding according to input information.

In addition, as shown in figure 11, the first acquisition module 8000, first judge module 9000, completion module 10000 can also should be comprised based on the man-machine chat device of artificial intelligence.

First acquisition module 8000, for after the input information receiving user's input, obtains the information above of chatting with user, and the topic information current according to acquisition of information user above.Then, according to information above, the first judge module 9000 can judge whether input information is greater than preset relation threshold value with the dependence of information above.When dependence is greater than preset relation threshold value, completion module 10000 can carry out completion according to information above to input information, thus ensures the fluency of man-machine chat.Particularly, completion is carried out to input information and can comprise reference resolution.For example, input information for " he has married? ", then according to information " Liu Dehua " above, " he " in input information can be replaced by " Liu Dehua ".Completion is carried out to input information and also can comprise omission completion.For example, above information " Liu De China wife be Zhu Liqian.", input information is not for " I is familiar with.", then can by input information completion for " I be familiar with Zhu Liqian.”。

In addition, as shown in figure 12, the second judge module 11000, second acquisition module 12000, first generation module 13000 and the second generation module 14000 can also should be comprised based on the man-machine chat device of artificial intelligence.

Second judge module 11000 for judging whether input information belongs to the chat message without actual content, as " laughing a great ho-ho ", " hoho " etc.If judge it is belong to the chat message without actual content, then the second acquisition module 12000 can obtain actualite, namely calculates actualite based on topic model (TopicModel) according to history chat record.After acquisition actualite, the first generation module 13000 can chat collection of illustrative plates according to actualite generation guiding topic based on topic.Wherein, the digraph of topic chat collection of illustrative plates to be one with topic be node.Such as, as shown in Figure 2, node " leisure " can point to node and " to see a film " and node " listens song ", then illustrate that can guide to topic from topic " leisure " " sees a film " or topic " listens song ".Topic " is seen a film " and topic " is listened song " and all had certain Guiding probability, can realize the guiding of topic according to Guiding probability, thus ensures the diversity guiding topic.Then, the second generation module 14000 can generate candidate's reply according to guiding topic.Particularly, based on spatial term model (NaturalLanguageGeneration), the template that candidate replys can be generated, guiding topic is filled in this template and generate candidate's reply; Also can choose to comprise guides the sentence of topic alternatively to reply, thus realizes carrying out chat topic guiding on one's own initiative to user.

The man-machine chat device based on artificial intelligence of the embodiment of the present invention, by receiving the input information of user's input, and distribution of information will be inputted to chatting service module, then candidate's reply that multiple chatting service module returns is received, and based on degree of confidence, reply to be selected is sorted, and generate chat message according to ranking results, and provide chat message to user, can carry out taking turns chat with user more, true nature, there is initiative, can also return for different user the reply meeting user styles, more personalized, intelligent.

In describing the invention, it will be appreciated that, term " " center ", " longitudinal direction ", " transverse direction ", " length ", " width ", " thickness ", " on ", D score, " front ", " afterwards ", " left side ", " right side ", " vertically ", " level ", " top ", " end " " interior ", " outward ", " clockwise ", " counterclockwise ", " axis ", " radial direction ", orientation or the position relationship of the instruction such as " circumference " are based on orientation shown in the drawings or position relationship, only the present invention for convenience of description and simplified characterization, instead of indicate or imply that the device of indication or element must have specific orientation, with specific azimuth configuration and operation, therefore limitation of the present invention can not be interpreted as.

In addition, term " first ", " second " only for describing object, and can not be interpreted as instruction or hint relative importance or imply the quantity indicating indicated technical characteristic.Thus, be limited with " first ", the feature of " second " can express or impliedly comprise at least one this feature.In describing the invention, the implication of " multiple " is at least two, such as two, three etc., unless otherwise expressly limited specifically.

In the present invention, unless otherwise clearly defined and limited, the term such as term " installation ", " being connected ", " connection ", " fixing " should be interpreted broadly, and such as, can be fixedly connected with, also can be removably connect, or integral; Can be mechanical connection, also can be electrical connection; Can be directly be connected, also indirectly can be connected by intermediary, can be the connection of two element internals or the interaction relationship of two elements, unless otherwise clear and definite restriction.For the ordinary skill in the art, above-mentioned term concrete meaning in the present invention can be understood as the case may be.

In the present invention, unless otherwise clearly defined and limited, fisrt feature second feature " on " or D score can be that the first and second features directly contact, or the first and second features are by intermediary indirect contact.And, fisrt feature second feature " on ", " top " and " above " but fisrt feature directly over second feature or oblique upper, or only represent that fisrt feature level height is higher than second feature.Fisrt feature second feature " under ", " below " and " below " can be fisrt feature immediately below second feature or tiltedly below, or only represent that fisrt feature level height is less than second feature.

In the description of this instructions, specific features, structure, material or feature that the description of reference term " embodiment ", " some embodiments ", " example ", " concrete example " or " some examples " etc. means to describe in conjunction with this embodiment or example are contained at least one embodiment of the present invention or example.In this manual, to the schematic representation of above-mentioned term not must for be identical embodiment or example.And the specific features of description, structure, material or feature can combine in one or more embodiment in office or example in an appropriate manner.In addition, when not conflicting, the feature of the different embodiment described in this instructions or example and different embodiment or example can carry out combining and combining by those skilled in the art.

Although illustrate and describe embodiments of the invention above, be understandable that, above-described embodiment is exemplary, can not be interpreted as limitation of the present invention, and those of ordinary skill in the art can change above-described embodiment within the scope of the invention, revises, replace and modification.

Claims (38)

1., based on a man-machine chat method for artificial intelligence, it is characterized in that, comprise the following steps:
Receive the input information of user's input;
By described input distribution of information to chatting service module;
Receive candidate's reply that described multiple chatting service module returns, wherein, described candidate replys has corresponding degree of confidence;
Based on described degree of confidence, described reply to be selected is sorted, and generate chat message according to ranking results, and provide described chat message to described user.
2. the method for claim 1, is characterized in that, after the input information of described reception user input, also comprises:
Error correction and/or rewriting are carried out to described input information.
3. the method for claim 1, is characterized in that, after the input information of described reception user input, also comprises:
Domain analysis is carried out to obtain field corresponding to described input information to described input information, wherein, according to field corresponding to described input information by described input distribution of information to the chatting service module with identical or close field.
4. the method for claim 1, is characterized in that, after the input information of described reception user input, also comprises:
Obtain the information above of chatting with described user;
Judge whether the dependence of described input information and described information is above greater than preset relation threshold value according to described information above; And
If be greater than described preset relation threshold value, then according to described information above, completion is carried out to described input information.
5. method as claimed in claim 4, is characterized in that, also comprise:
According to the described topic information that user described in acquisition of information is current above.
6. the method for claim 1, is characterized in that, described chatting service module comprise based on search chat module, rich knowledge chat module, based on portrait chat module and based on one or more in the chat module of mass-rent.
7. method as claimed in claim 6, is characterized in that, also comprise:
The described chat module based on search cuts word to generate multiple chat short sentence to described input information;
The described chat module based on search according to described multiple chat short sentence inquiry chat corpus to generate sentence on multiple chat language material, and on described multiple chat language material sentence correspondence multiple chat language materials under sentence;
The described chat module based on search is filtered sentence on described multiple chat language material;
The described chat module based on search is classified to sentence under the chat language material of sentence correspondence on the chat language material after filtration; And
The described chat module based on search reorders to sentence under the described chat language material after classification, and generates described candidate reply according to ranking results.
8. method as claimed in claim 7, is characterized in that, the described chat module based on search is filtered sentence on described multiple chat language material and specifically comprised:
Calculate the similarity between sentence on described input information and described multiple chat language material;
If described similarity is less than first preset similarity threshold, then sentence on the chat language material of correspondence is filtered; And
If described similarity is more than or equal to described first preset similarity threshold, then sentence on the chat language material of correspondence is retained.
9. method as claimed in claim 7, is characterized in that, the described classification to sentence under the chat language material of sentence correspondence on the chat language material after filtration specifically comprises:
Calculate the similarity between sentence under described input information and described multiple chat language material; And
According to described similarity, sentence under described multiple chat language material is classified.
10. method as claimed in claim 9, is characterized in that, the similarity under described input information and described multiple chat language material between sentence comprises:
Similarity literal between sentence under described input information and described chat language material;
Or the similarity obtained trained in sentence under described input information and described chat language material based on deep neural network;
Or the similarity obtained trained in sentence under described input information and described chat language material based on Machine Translation Model.
11. methods as claimed in claim 7, is characterized in that, described reordering to sentence under the described chat language material after classification specifically comprises:
According to the chat attribute of the user described in acquisition of information above that described user chats;
According to described chat attribute, sentence under the described chat language material after described classification is reordered.
12. methods as claimed in claim 6, is characterized in that, also comprise:
Described rich knowledge chat module generates search word according to described input information, and carries out searching for generate multiple Search Results according to described search word;
Described rich knowledge chat module carries out sentence extraction to described multiple Search Results, to obtain candidate sentences set;
Described rich knowledge chat module is rewritten to generate described candidate to the sentence in described candidate sentences set and is replied.
13. methods as claimed in claim 12, is characterized in that, wherein, the sentence in described candidate sentences set and the similarity of described search word are greater than second and preset similarity threshold.
14. methods as claimed in claim 12, is characterized in that, also comprise:
Chat attribute according to described user reorders to the sentence in described candidate sentences set.
15. methods as claimed in claim 6, is characterized in that, also comprise:
The described chat module based on portrait obtains the chat linguistic context of described user;
The described chat module based on portrait judges whether to meet collection condition according to described chat linguistic context;
If judge to meet described collection condition, then send problem to described user;
Receive the answer information of described user according to described problem, and according to described answer information, model drawn a portrait to user and upgrade.
16. methods as claimed in claim 6, is characterized in that, also comprise:
The described chat module based on portrait obtains the chat content of described user;
The described chat module based on portrait extracts user's representation data according to described chat content;
The described chat module based on portrait is drawn a portrait model according to the described user's representation data extracted to described user and is upgraded.
17. methods as claimed in claim 6, is characterized in that, also comprise:
The described chat module based on mass-rent judges whether described input information is applicable to mass-rent and completes;
If judge that being applicable to mass-rent completes, then by described input distribution of information to corresponding executor;
Receive the return information of described executor, and Quality estimation is carried out to described return information;
If meet quality requirements, then described return information is replied as described candidate.
18. the method for claim 1, is characterized in that, also comprise:
Judge whether described input information belongs to the chat message without actual content;
If judge it is belong to the chat message without actual content, then obtain actualite;
Generate according to described actualite and guide topic; And
Generate described candidate according to described guiding topic to reply.
19. the method for claim 1, is characterized in that, described sequence to described reply to be selected based on described degree of confidence specifically comprises:
Obtain the feature of the described input information of described user; And
Based on described input information characteristic sum described in degree of confidence described reply to be selected is sorted.
20. 1 kinds, based on the man-machine chat device of artificial intelligence, is characterized in that, comprising:
First receiver module, for receiving the input information of user's input;
Distribution module, for by described input distribution of information to chatting service module;
Second receiver module, the candidate returned for receiving described multiple chatting service module replys, and wherein, described candidate replys has corresponding degree of confidence;
Module is provided, for sorting to described reply to be selected based on described degree of confidence, and generates chat message according to ranking results, and provide described chat message to described user.
21. devices as claimed in claim 20, it is characterized in that, described device also comprises:
Correction module, for after the input information of described reception user input, carries out error correction and/or rewriting to described input information.
22. devices as claimed in claim 20, it is characterized in that, described device also comprises:
Analysis module, for after the input information of described reception user input, carries out domain analysis to obtain field corresponding to described input information to described input information;
Described distribution module, for according to field corresponding to described input information by described input distribution of information to the chatting service module with identical or close field.
23. devices as claimed in claim 20, it is characterized in that, described device also comprises:
First acquisition module, for after the input information of described reception user input, obtains the information above of chatting with described user;
According to described information above, first judge module, for judging whether the dependence of described input information and described information is above greater than preset relation threshold value; And
Completion module, for when described dependence is greater than described preset relation threshold value, carries out completion according to described information above to described input information.
24. devices as claimed in claim 23, is characterized in that, described first acquisition module also for:
According to the described topic information that user described in acquisition of information is current above.
25. devices as claimed in claim 20, is characterized in that, described chatting service module comprise based on search chat module, rich knowledge chat module, based on portrait chat module and based on one or more in the chat module of mass-rent.
26. methods as claimed in claim 25, is characterized in that, the described chat module based on search, specifically comprises:
Cut lexon module, for cutting word to described input information to generate multiple chat short sentence;
Generate submodule, for according to described multiple chat short sentence inquiry chat corpus to generate sentence on multiple chat language material, and on described multiple chat language material sentence correspondence multiple chat language materials under sentence;
Filter submodule, for filtering sentence on described multiple chat language material;
Classification submodule, under the chat language material to sentence correspondence on the chat language material after filtration, sentence is classified; And
Sorting sub-module, for reordering to sentence under the described chat language material after classification, and generates described candidate reply according to ranking results.
27. devices as claimed in claim 26, it is characterized in that, described filtration submodule, specifically comprises:
Computing unit, for calculating the similarity on described input information and described multiple chat language material between sentence;
Filter element, if be less than first for described similarity to preset similarity threshold, then filters sentence on the chat language material of correspondence; And
Stick unit, if be more than or equal to described first for described similarity to preset similarity threshold, then retains sentence on the chat language material of correspondence.
28. devices as claimed in claim 26, it is characterized in that, described classification submodule specifically comprises:
Computing unit, for calculating the similarity under described input information and described multiple chat language material between sentence; And
Taxon, for classifying to sentence under described multiple chat language material according to described similarity.
29. devices as claimed in claim 28, is characterized in that, the similarity under described input information and described multiple chat language material between sentence comprises:
Similarity literal between sentence under described input information and described chat language material;
Or the similarity obtained trained in sentence under described input information and described chat language material based on deep neural network;
Or the similarity obtained trained in sentence under described input information and described chat language material based on Machine Translation Model.
30. devices as claimed in claim 26, it is characterized in that, described sorting sub-module specifically comprises:
Acquiring unit, for the chat attribute of user described in acquisition of information above of chatting according to described user;
Sequencing unit, for reordering to sentence under the described chat language material after described classification according to described chat attribute.
31. devices as claimed in claim 25, is characterized in that, described rich knowledge chat module, specifically comprises:
Generating submodule, for generating search word according to described input information, and carrying out searching for generate multiple Search Results according to described search word;
Extract submodule, for carrying out sentence extraction, with candidate sentences set to described multiple Search Results;
Rewrite submodule, reply for rewriting to generate described candidate to the sentence in described candidate sentences set.
32. devices as claimed in claim 31, is characterized in that, wherein, the sentence in described candidate sentences set and the similarity of described search word are greater than second and preset similarity threshold.
33. devices as claimed in claim 31, is characterized in that, also comprise:
Reorder submodule, reorders to the sentence in described candidate sentences set for the chat attribute according to described user.
34. devices as claimed in claim 25, is characterized in that, the described chat module based on portrait specifically comprises:
First obtains submodule, for obtaining the chat linguistic context of described user;
Judge submodule, for judging whether to meet collection condition according to described chat linguistic context;
Sending submodule, for when meeting described collection condition, sending problem to described user;
First upgrades submodule, for receiving the answer information of described user according to described problem, and draws a portrait model according to described answer information to user and upgrades.
35. devices as claimed in claim 25, is characterized in that, the described chat module based on portrait specifically comprises:
Second obtains submodule, for obtaining the chat content of described user;
Extract submodule, for extracting user's representation data according to described chat content;
Second upgrades submodule, upgrades for drawing a portrait model according to the described user's representation data extracted to described user.
36. devices as claimed in claim 25, is characterized in that, the described chat module based on mass-rent specifically for:
Judge whether described input information is applicable to mass-rent and completes, if judge that being applicable to mass-rent completes, then by described input distribution of information to corresponding executor, and receive the return information of described executor, and Quality estimation is carried out to described return information, if meet quality requirements, then described return information is replied as described candidate.
37. devices as claimed in claim 20, is characterized in that, also comprise:
Second judge module, for judging whether described input information belongs to the chat message without actual content;
Second acquisition module, if be belong to the chat message without actual content for judging, then obtains actualite;
First generation module, guides topic for generating according to described actualite; And
Second generation module, replys for generating described candidate according to described guiding topic.
38. devices as claimed in claim 20, is characterized in that, described in provide module specifically for:
Obtain the feature of the described input information of described user, and based on described input information characteristic sum described in degree of confidence described reply to be selected is sorted.
CN201510564173.0A 2015-09-07 2015-09-07 Human-computer chatting method and device based on artificial intelligence CN105183848A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510564173.0A CN105183848A (en) 2015-09-07 2015-09-07 Human-computer chatting method and device based on artificial intelligence

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201510564173.0A CN105183848A (en) 2015-09-07 2015-09-07 Human-computer chatting method and device based on artificial intelligence
PCT/CN2015/096342 WO2017041370A1 (en) 2015-09-07 2015-12-03 Human-computer chatting method and device based on artificial intelligence

Publications (1)

Publication Number Publication Date
CN105183848A true CN105183848A (en) 2015-12-23

Family

ID=54905929

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510564173.0A CN105183848A (en) 2015-09-07 2015-09-07 Human-computer chatting method and device based on artificial intelligence

Country Status (2)

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

Cited By (27)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105787560A (en) * 2016-03-18 2016-07-20 北京光年无限科技有限公司 Dialogue data interaction processing method and device based on recurrent neural network
CN105808695A (en) * 2016-03-03 2016-07-27 陈包容 Method and device for obtaining chat reply contents
CN106202476A (en) * 2016-07-14 2016-12-07 广州安望信息科技有限公司 A kind of interactive method and device of knowledge based collection of illustrative plates
CN106202417A (en) * 2016-07-12 2016-12-07 北京光年无限科技有限公司 A kind of man-machine interaction method for intelligent robot and system
CN106227740A (en) * 2016-07-12 2016-12-14 北京光年无限科技有限公司 A kind of data processing method towards conversational system and device
CN106293348A (en) * 2016-08-16 2017-01-04 东软集团股份有限公司 It is applied to information interacting method and the device of new media
CN106448670A (en) * 2016-10-21 2017-02-22 竹间智能科技（上海）有限公司 Dialogue automatic reply system based on deep learning and reinforcement learning
CN106649762A (en) * 2016-12-27 2017-05-10 竹间智能科技（上海）有限公司 Intention recognition method and system based on inquiry question and feedback information
CN106708983A (en) * 2016-12-09 2017-05-24 竹间智能科技（上海）有限公司 Dialogue interactive information-based user portrait construction system and method
CN106776557A (en) * 2016-12-13 2017-05-31 竹间智能科技（上海）有限公司 Affective state memory recognition methods and the device of emotional robot
CN106779817A (en) * 2016-11-29 2017-05-31 竹间智能科技（上海）有限公司 Intension recognizing method and system based on various dimensions information
CN106844646A (en) * 2017-01-22 2017-06-13 竹间智能科技（上海）有限公司 Artificial intelligence reply method and device based on emotional prediction
CN107305768A (en) * 2016-04-20 2017-10-31 上海交通大学 Easy wrongly written character calibration method in interactive voice
WO2018000271A1 (en) * 2016-06-29 2018-01-04 深圳狗尾草智能科技有限公司 Intention scene recognition method and system based on user portrait
WO2018000281A1 (en) * 2016-06-29 2018-01-04 深圳狗尾草智能科技有限公司 User portrait representation learning system and method based on deep neural network
WO2018006469A1 (en) * 2016-07-07 2018-01-11 深圳狗尾草智能科技有限公司 Knowledge graph-based human-robot interaction method and system
CN107665362A (en) * 2017-09-26 2018-02-06 联想(北京)有限公司 For realizing the training method of robot chat, predicting the method and device of answer
CN107844481A (en) * 2017-11-21 2018-03-27 新疆科大讯飞信息科技有限责任公司 Identify text error-detecting method and device
CN108021547A (en) * 2016-11-04 2018-05-11 株式会社理光 A kind of generation method of natural language, spatial term device and electronic equipment
CN108021572A (en) * 2016-11-02 2018-05-11 腾讯科技（深圳）有限公司 Return information recommends method and apparatus
CN108170749A (en) * 2017-12-21 2018-06-15 北京百度网讯科技有限公司 Dialogue method, device and computer-readable medium based on artificial intelligence
CN108415932A (en) * 2018-01-23 2018-08-17 苏州思必驰信息科技有限公司 Interactive method and electronic equipment
CN108766421A (en) * 2017-04-20 2018-11-06 杭州萤石网络有限公司 The method and device of interactive voice
WO2018214164A1 (en) * 2017-05-26 2018-11-29 Microsoft Technology Licensing, Llc Recommending friends in automated chatting
CN109086329A (en) * 2018-06-29 2018-12-25 出门问问信息科技有限公司 Dialogue method and device are taken turns in progress based on topic keyword guidance more
CN109995642A (en) * 2017-12-29 2019-07-09 Tcl集团股份有限公司 A kind of method and device automatically generating quickly revert, instant communicating system
CN110020007A (en) * 2017-09-15 2019-07-16 上海挖数互联网科技有限公司 Robot interactive control method and device, storage medium, server

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10826864B2 (en) 2018-07-27 2020-11-03 At&T Intellectual Property I, L.P. Artificially intelligent messaging

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101221583A (en) * 2008-01-29 2008-07-16 北京百问百答网络技术有限公司 Question recommending method and system
CN104598445A (en) * 2013-11-01 2015-05-06 腾讯科技（深圳）有限公司 Automatic question-answering system and method
CN104714954A (en) * 2013-12-13 2015-06-17 中国电信股份有限公司 Information searching method and system based on context understanding

Family Cites Families (2)

* 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
CN105068661B (en) * 2015-09-07 2018-09-07 百度在线网络技术（北京）有限公司 Man-machine interaction method based on artificial intelligence and system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101221583A (en) * 2008-01-29 2008-07-16 北京百问百答网络技术有限公司 Question recommending method and system
CN104598445A (en) * 2013-11-01 2015-05-06 腾讯科技（深圳）有限公司 Automatic question-answering system and method
CN104714954A (en) * 2013-12-13 2015-06-17 中国电信股份有限公司 Information searching method and system based on context understanding

Cited By (36)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105808695A (en) * 2016-03-03 2016-07-27 陈包容 Method and device for obtaining chat reply contents
CN105787560B (en) * 2016-03-18 2018-04-03 北京光年无限科技有限公司 Dialogue data interaction processing method and device based on Recognition with Recurrent Neural Network
CN105787560A (en) * 2016-03-18 2016-07-20 北京光年无限科技有限公司 Dialogue data interaction processing method and device based on recurrent neural network
CN107305768A (en) * 2016-04-20 2017-10-31 上海交通大学 Easy wrongly written character calibration method in interactive voice
CN107305768B (en) * 2016-04-20 2020-06-12 上海交通大学 Error-prone character calibration method in voice interaction
WO2018000281A1 (en) * 2016-06-29 2018-01-04 深圳狗尾草智能科技有限公司 User portrait representation learning system and method based on deep neural network
WO2018000271A1 (en) * 2016-06-29 2018-01-04 深圳狗尾草智能科技有限公司 Intention scene recognition method and system based on user portrait
WO2018006469A1 (en) * 2016-07-07 2018-01-11 深圳狗尾草智能科技有限公司 Knowledge graph-based human-robot interaction method and system
CN106227740A (en) * 2016-07-12 2016-12-14 北京光年无限科技有限公司 A kind of data processing method towards conversational system and device
CN106202417A (en) * 2016-07-12 2016-12-07 北京光年无限科技有限公司 A kind of man-machine interaction method for intelligent robot and system
CN106202476A (en) * 2016-07-14 2016-12-07 广州安望信息科技有限公司 A kind of interactive method and device of knowledge based collection of illustrative plates
CN106293348B (en) * 2016-08-16 2020-02-18 东软集团股份有限公司 Information interaction method and device applied to new media
CN106293348A (en) * 2016-08-16 2017-01-04 东软集团股份有限公司 It is applied to information interacting method and the device of new media
CN106448670A (en) * 2016-10-21 2017-02-22 竹间智能科技（上海）有限公司 Dialogue automatic reply system based on deep learning and reinforcement learning
CN106448670B (en) * 2016-10-21 2019-11-19 竹间智能科技（上海）有限公司 Conversational system is automatically replied based on deep learning and intensified learning
CN108021572A (en) * 2016-11-02 2018-05-11 腾讯科技（深圳）有限公司 Return information recommends method and apparatus
CN108021572B (en) * 2016-11-02 2020-08-07 腾讯科技（深圳）有限公司 Reply information recommendation method and device
CN108021547A (en) * 2016-11-04 2018-05-11 株式会社理光 A kind of generation method of natural language, spatial term device and electronic equipment
CN106779817A (en) * 2016-11-29 2017-05-31 竹间智能科技（上海）有限公司 Intension recognizing method and system based on various dimensions information
CN106708983A (en) * 2016-12-09 2017-05-24 竹间智能科技（上海）有限公司 Dialogue interactive information-based user portrait construction system and method
CN106776557A (en) * 2016-12-13 2017-05-31 竹间智能科技（上海）有限公司 Affective state memory recognition methods and the device of emotional robot
CN106776557B (en) * 2016-12-13 2020-09-08 竹间智能科技（上海）有限公司 Emotional state memory identification method and device of emotional robot
CN106649762A (en) * 2016-12-27 2017-05-10 竹间智能科技（上海）有限公司 Intention recognition method and system based on inquiry question and feedback information
CN106844646A (en) * 2017-01-22 2017-06-13 竹间智能科技（上海）有限公司 Artificial intelligence reply method and device based on emotional prediction
CN108766421B (en) * 2017-04-20 2020-09-15 杭州萤石网络有限公司 Voice interaction method and device
CN108766421A (en) * 2017-04-20 2018-11-06 杭州萤石网络有限公司 The method and device of interactive voice
WO2018214164A1 (en) * 2017-05-26 2018-11-29 Microsoft Technology Licensing, Llc Recommending friends in automated chatting
CN110020007A (en) * 2017-09-15 2019-07-16 上海挖数互联网科技有限公司 Robot interactive control method and device, storage medium, server
CN107665362A (en) * 2017-09-26 2018-02-06 联想(北京)有限公司 For realizing the training method of robot chat, predicting the method and device of answer
CN107844481B (en) * 2017-11-21 2019-09-13 新疆科大讯飞信息科技有限责任公司 Identify text error-detecting method and device
CN107844481A (en) * 2017-11-21 2018-03-27 新疆科大讯飞信息科技有限责任公司 Identify text error-detecting method and device
CN108170749A (en) * 2017-12-21 2018-06-15 北京百度网讯科技有限公司 Dialogue method, device and computer-readable medium based on artificial intelligence
CN109995642A (en) * 2017-12-29 2019-07-09 Tcl集团股份有限公司 A kind of method and device automatically generating quickly revert, instant communicating system
CN108415932A (en) * 2018-01-23 2018-08-17 苏州思必驰信息科技有限公司 Interactive method and electronic equipment
CN109086329A (en) * 2018-06-29 2018-12-25 出门问问信息科技有限公司 Dialogue method and device are taken turns in progress based on topic keyword guidance more
CN109086329B (en) * 2018-06-29 2021-01-05 出门问问信息科技有限公司 Topic keyword guide-based multi-turn conversation method and device

Also Published As

Publication number Publication date
WO2017041370A1 (en) 2017-03-16

Similar Documents

Publication Publication Date Title
US20190311025A1 (en) Methods and systems for modeling complex taxonomies with natural language understanding
US10235355B2 (en) System, method, and computer-readable storage device for providing cloud-based shared vocabulary/typing history for efficient social communication
CN106384593B (en) A kind of conversion of voice messaging, information generating method and device
CN105740228B (en) A kind of internet public feelings analysis method and system
CN104268197B (en) A kind of industry comment data fine granularity sentiment analysis method
KR101895249B1 (en) Method, system and storage medium for realizing intelligent answering of questions
KR102170563B1 (en) Human machine interactive method and apparatus based on artificial intelligence
US9043285B2 (en) Phrase-based data classification system
CN104794212B (en) Context sensibility classification method and categorizing system based on user comment text
CN106095833B (en) Human-computer dialogue content processing method
CN105808200B (en) Intelligent automation assistant
CN105895087B (en) Voice recognition method and device
Jeong et al. Semi-supervised speech act recognition in emails and forums
CN105426514B (en) Personalized mobile application APP recommended method
US10002185B2 (en) Context-aware cognitive processing
DE112013004082T5 (en) Search system of the emotion entity for the microblog
WO2018149326A1 (en) Natural language question answering method and apparatus, and server
US20150074112A1 (en) Multimedia Question Answering System and Method
CN107229684A (en) Statement classification method, system, electronic equipment, refrigerator and storage medium
CN105512228A (en) Bidirectional question-answer data processing method and system based on intelligent robot
CN105095182B (en) A kind of return information recommendation method and device
CN106484664A (en) Similarity calculating method between a kind of short text
CN105704400A (en) Learning system based on multi-platform terminal and cloud service and operation method thereof
CN105426436A (en) Artificial intelligent robot based information provision method and apparatus
CN104933113B (en) A kind of expression input method and device based on semantic understanding

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20151223