CN110209791A - It is a kind of to take turns dialogue intelligent speech interactive system and device more - Google Patents

It is a kind of to take turns dialogue intelligent speech interactive system and device more Download PDF

Info

Publication number
CN110209791A
CN110209791A CN201910505280.4A CN201910505280A CN110209791A CN 110209791 A CN110209791 A CN 110209791A CN 201910505280 A CN201910505280 A CN 201910505280A CN 110209791 A CN110209791 A CN 110209791A
Authority
CN
China
Prior art keywords
semantic
module
text
dialogue
semantic understanding
Prior art date
Legal status (The legal status 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 status listed.)
Granted
Application number
CN201910505280.4A
Other languages
Chinese (zh)
Other versions
CN110209791B (en
Inventor
张韶峰
冯鑫
王世朋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Bairong Yunchuang Technology Co Ltd
Original Assignee
Bairong Yunchuang Technology Co Ltd
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 Bairong Yunchuang Technology Co Ltd filed Critical Bairong Yunchuang Technology Co Ltd
Priority to CN201910505280.4A priority Critical patent/CN110209791B/en
Publication of CN110209791A publication Critical patent/CN110209791A/en
Application granted granted Critical
Publication of CN110209791B publication Critical patent/CN110209791B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G06F16/3329—Natural language query formulation or dialogue systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3343—Query execution using phonetics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3344—Query execution using natural language analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G06F16/353—Clustering; Classification into predefined classes

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Human Computer Interaction (AREA)
  • Health & Medical Sciences (AREA)
  • Acoustics & Sound (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Machine Translation (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

It is a kind of to take turns dialogue intelligent speech interactive system and device more, system includes hybrid semantic understanding module, semantic understanding adaptation module and automatic dialogue management module, voice input is converted into the hybrid semantic understanding module of text input after speech recognition, understand that user is intended to and extracts corresponding state information, automatic dialogue management module is intended to based on user, guide dialog procedure, output dialog text is simultaneously converted to voice output, realize dialogue, semantic understanding adaptation module is used for the Optimization Learning of hybrid semantic understanding module.Multiple modules such as integrating speech sound identification of the present invention, natural language understanding, spatial term, speech synthesis, dialogue management form a whole set of more wheels dialogue intelligent speech interactive system easily extend, can configure, can be applied to any scene.

Description

It is a kind of to take turns dialogue intelligent speech interactive system and device more
Technical field
The invention belongs to field of computer technology, are related to natural language processing and artificial intelligence field, for a kind of more wheels pair Talk about intelligent speech interactive system.
Background technique
How to allow computer understanding human language is all the popular research side of artificial intelligence, natural language processing all the time To, and modern artificial intelligence field one of the key problem to be solved.Speech recognition, image recognition technology application increasingly at It is ripe instantly, although research temperature of the depth learning technology in semantic understanding field is very high, the people of real spoken dialog Work intellectual product is still very few.Common speech robot people on the market, is mostly voice assistant humanoid robot, is often based upon keyword Language matching, can Understanding content not only very simple, but also be difficult to accomplish continuous more wheel interactions, usually give an irrelevant answer
Several intelligent sound interaction schemes of the prior art are described below.
(1) prior art one: speech recognition ASR+ text matches
Speech recognition adds text matches to be the mode for realizing that intelligent speech interactive system is most traditional, and this mode is easy due to it Traditional call center is commonly applied in the advantages such as realization, low to data degree of dependence.Text matches usually use, and accurate Match, such as judge that character string is equal or fuzzy matching, such as in the way of regular expression progress wildcard, from speech recognition " keyword " is extracted in text, and instruction distribution is carried out to " keyword ", to achieve the purpose that speech understanding.
However, this kind of scheme has the following problems: 1) accuracy rate of speech recognition and dialogue field tight association, and train The speech recognition system of specific area has that cost is excessively high again, 2) spoken dialog is different from text conversation, in a word Usually will appear multiple intentions or even inconsistent intention, 3) when interactive content becomes complexity, the matching artificially write Grammer can be increased rapidly to the degree for being difficult to safeguard, and the conflict between matching grammer usually occur, therefore this kind of interactive voice The semantics recognition accuracy rate of system is very limited.
(2) prior art two: speech recognition ASR+ intention assessment+semanteme slot
It is also a kind of common intelligent sound interactive mode that intention assessment, which adds semantic slot to extract, acquires dialog text number in advance According to, text data is labeled and is classified, such as:
By " how is Beijing weather tomorrow? " it is labeled as " tomorrow/Beijing TIME/LOC weather/INT how/B ", and On a large amount of labeled data, train classification models carry out intention assessment, and common intention assessment model has SVM model or CNN, RNN Even depth learning model；And according to the extraction for being intended to the semantic slot of progress, common semantic slot extraction has syntactic analysis, name entity It extracts and sequence to the modes such as series model.Shown in example as above, it is intended that be " Cha Tianqi " that extractible semanteme slot includes the " time (TIME) ", " place (LOC) " two factors.The mode that this kind of intention assessment adds semantic slot to extract, adds somewhat to language The accuracy rate of justice identification and more wheel dialogue abilities of robot.However, this mode 1) rely on a large amount of artificial labeled data, mark The quality of note data determines the quality finally interacted with quantity, and artificial labeled data is often time-consuming very long, is difficult in the short time Interior completion；2) simple 1-2 wheel language interaction can only be realized, interactive content is also limited by the content of semantic slot.Therefore, this Scheme can be limited to the speed ability of its expansion, often can be only applied to the well-defined business field of very mature and slot value Scape is difficult to adapt to the scene demand of content fast development variation.
(3) prior art three: speech recognition ASR+ sequence to Sequence Learning Seq2Seq+ Text To Speech TTS
Speech recognition, series model add the mode of speech synthesis, that is, so-called end to end model, are produced by internet Raw a large amount of text data is directly predicted in output text using the Seq2Seq model in deep learning by input text Hold, by speech synthesis, converts text to voice and exported.
But above-mentioned end to end model is used, output text is directly predicted by input text, it appears that be to improve robot Intelligent interaction ability, but this interactive mode only considers text currently entered due to the limitation of state-of-the art, Without considering semanteme above, so that robot loses the ability of more wheel dialogues, can only be used as chatting.Moreover, by It is provided completely by model in the content of its interaction, so that interaction loses controllability.The most common situation is that model is only certain The content for having remembered training data in degree is not to produce self intelligently to go to judge conversation content, so that dialogue interaction Logical miss is given an irrelevant answer.Therefore, this scheme is more used to entertain, and is difficult to apply to production practices.
Summary of the invention
The problem to be solved in the present invention is: existing intelligent sound identification technology identification method is single, is not for grammer Very rigorous spoken recognition effect is bad, cannot identify complicated voice content, or a large amount of people are needed in identification learning Work mark, adaptive ability is insufficient, requires manual intervention and completes data update, and it is the more wheel dialogues of realization that segmentation scheme, which seems, but The logical relation between each wheel dialogue, the only simple stacking of single-wheel dialogue are not identified in practical identification process.Needle of the present invention To the deficiencies in the prior art, propose that one kind takes turns the Intelligent voice dialog system of dialogue more.
The technical solution of the present invention is as follows: a kind of more wheel dialogue intelligent speech interactive systems, including hybrid semantic understanding mould Block, semantic understanding adaptation module and automatic dialogue management module, it is mixed that voice input is converted into text input after speech recognition Box-like semantic understanding module understands that user is intended to and extracts corresponding state information, and automatic dialogue management module is intended to based on user, Dialog procedure is guided, dialog text is exported and is converted to voice output, realizes dialogue, semantic understanding adaptation module is for mixing The Optimization Learning of formula semantic understanding module,
Hybrid semantic understanding module, by the way of Model Fusion, in conjunction with text matches, semantic similarity matching, letter The semantic understandings schemes such as retrieval, more intent classifier models are ceased, comprehensive distinguishing dialog semantics, wherein text matches belong to preposition calculation Method first pre-processes sentence, obtains pretreated dialog text, then by semantic similarity matching, information retrieval and Duo Yi The mode of figure disaggregated model fusion, exports final result jointly；
Semantic understanding adaptation module optimizes existing hybrid semantic reason by way of transfer learning, re -training Solve model, including Bi-LSTM language model, similarity mode model and more intent classifier models；
Automatic dialogue management module for realizing human-computer interactive control and expands configuration, based on shape locating for current session The information that state, current user's intention of identification are interacted with dialog history, the interactive instruction of comprehensive distinguishing output；And by more Intention assessment realizes that multiple interaction and state are converted by a human-computer dialogue.
Further, in hybrid semantic understanding module, text matches are for establishing semantic understanding rule, semantic similarity Matching establishes semantic matches model using the Bi-LSTM neural network language model based on attention mechanism, by the dialogue of input Text combination semantic understanding rule carries out vectorization expression, and using trained twin network by the way of, to semantic matches model into Row accurate adjustment fine-tune finally indicates the regression model of training convolutional neural networks based on vectorization, to predict between two texts Semantic similarity；Information retrieval passes through the corresponding standard corpus database of dialog text and the intention classification of standard corpus, base In semantic similarity retrieved from standard corpus library with the highest corpus of dialog text semantic similarity, be intended to class Not as the intent classifier of dialog text, the intention of identification input text is realized；More intent classifier models coupling standard corpus numbers According to the business datum with application, generate the labeled data for being intended to classification, one Bi- based on attention mechanism of training more LSTM network carries out more classification predictions, provides the intention classification of standard corpus as more intent classifier models.
Further, semantic understanding adaptation module is used to optimize model in hybrid semantic understanding module, including with Lower optimization:
1) newly-increased corpus data is imported in the training data of Bi-LSTM semantic matches model, is trained, updates it Network weight；
2) cleaning filtering is carried out to the newly-increased labeled data of more intent classifier models, filters out optimal a part mark number According to, be mixed into standard corpus library, and Intention Anticipation again is carried out to whole labeled data, according to the text feature of corpus and its Performance on labeled data collection establishes the order models and corresponding index monitoring mechanism for having supervision, monitors each intention assessment Accuracy rate and the corpus of recall rate and tagged corpus change；
3) newly-increased labeled data is imported into the more intent classifier models of Bi-LSTM, updates its network weight, and monitor it and testing Accuracy rate and recall rate on card collection；
4) model in the online updated hybrid semantic understanding module of automatically dispose and standard corpus library.
Further, automatic dialogue management module is being executed using finite state machine in such a way that intensified learning combines When each round dialogue interaction, current session state in which, current user are intended to the information interacted with history, in conjunction with preparatory The dialogue rules of interaction of setting and the interactive strategy obtained by study, carry out comprehensive descision, and output dialogue robot should execute Interactive operation instruction.
The present invention also proposes that one kind takes turns dialogue intelligent sound interactive device more, and described device is the calculating with storage medium Machine device, computer program is mounted in the storage medium, and the computer program talks with intelligence for realizing above-mentioned more wheels It can voice interactive system.
Integrating speech sound identification of the present invention, natural language understanding, spatial term, speech synthesis, dialogue management etc. are multiple Module forms a whole set of more wheels dialogue intelligent speech interactive system easily extend, can configure, can be applied to any scene.This In terms of natural language understanding, innovative has used hybrid semantic understanding model for invention, excavates the spoken rule of dialogue in depth Rule on the basis of summarizing summary semantic classes, in conjunction with traditional natural language processing techniques and deep neural network algorithm, is managed in real time Solution is spoken semantic, realizes man-machine smooth exchange, promotes interactive experience.Meanwhile in order to reduce model optimization to the full extent to people The dependence of work mark, customizes natural language understanding adaptation module, automatically updates model parameter, make model that self may be implemented more Newly with optimization.
Detailed description of the invention
Fig. 1 is the embodiment of the present invention system architecture figure.
Fig. 2 is the identification process schematic diagram of the embodiment of the present invention.
The position Fig. 3 embodiment of the present invention takes turns conversation process figure more.
Specific embodiment
The present invention provides one kind to take turns dialogue intelligent speech interactive system, including hybrid semantic understanding module, semanteme more Understand adaptation module and automatic dialogue management module, voice input is converted into the hybrid semanteme of text input after speech recognition Understanding Module understands that user is intended to and extracts corresponding state information, and automatic dialogue management module is intended to based on user, guidance dialogue Process exports dialog text and is converted to voice output, realizes dialogue, semantic understanding adaptation module is for hybrid semantic reason The Optimization Learning of module is solved,
Hybrid semantic understanding module, by the way of Model Fusion, in conjunction with text matches, semantic similarity matching, letter The semantic understandings schemes such as retrieval, more intent classifier models are ceased, comprehensive distinguishing dialog semantics, wherein text matches belong to preposition calculation Method first pre-processes sentence, obtains pretreated dialog text, then by semantic similarity matching, information retrieval and Duo Yi The mode of figure disaggregated model fusion, exports final result jointly；
Semantic understanding adaptation module optimizes existing hybrid semantic reason by way of transfer learning, re -training Solve model, including Bi-LSTM language model, similarity mode model and more intent classifier models；
Automatic dialogue management module for realizing human-computer interactive control and expands configuration, based on shape locating for current session The information that state, current user's intention of identification are interacted with dialog history, the interactive instruction of comprehensive distinguishing output；And by more Intention assessment realizes that multiple interaction and state are converted by a human-computer dialogue.
As shown in Figure 1, for the system structure diagram of one specific implementation of the present invention, wherein ASR module is speech recognition Text is asked in module, the speech transcription for acquiring user terminal；NLU module is semantic understanding module, for understanding that user is intended to With extraction corresponding information, i.e., hybrid semantic understanding module；Adaptation module is semantic understanding adaptation module, assists NLU The self-renewing of module, DM module are automatic dialogue management module, are intended to based on user, guide dialog procedure；NLG module is text This generation module, is intended to based on user and knowledge base is extracted, and carries out the text generation of voice output；TTS module is speech synthesis Module, the information that will be exported are converted into corresponding voice.
The realization of modules of the invention is specifically described below.
Hybrid semantic understanding module
Hybrid semantic understanding module, by the way of Model Fusion, mixing text matches, semantic similarity matching, letter The semantic understandings schemes such as retrieval, intent classifier are ceased, comprehensive distinguishing dialog semantics considerably increase the accuracy and spirit of semantic understanding Activity.Text matches are for establishing semantic understanding rule, and semantic similarity matching is using the Bi-LSTM mind based on attention mechanism Through netspeak model foundation semantic matches model, the dialog text combination semantic understanding rule of input is subjected to vectorization table Show, and by the way of the twin network of training, accurate adjustment fine-tune is carried out to semantic matches model, is finally based on vectorization table Show the regression model of training convolutional neural networks, to predict the semantic similarity between two texts；Information retrieval passes through dialogue text The intention classification of this corresponding standard corpus database and standard corpus, is retrieved from standard corpus library based on semantic similarity With the highest corpus of dialog text semantic similarity, it is intended to intent classifier of the classification as dialog text, realizes and know Not Shu Ru text intention；The business datum of more intent classifier models coupling standard corpus data and application generates more meanings The labeled data of figure classification, training one based on the Bi-LSTM network of attention mechanism as more intent classifier models, carry out more Classification prediction, provides the intention classification of standard corpus.
1) text matches
Text matches, in conjunction with traditional natural Language Processing algorithm, such as keyword extraction, syntactic analysis and name Entity recognition Technology, fine granularity is carried out to the common vocabulary of business scenario and is summarized as Feature Words.On the basis of vocabulary is sorted out, establish semantic Understand that rule, a syntax rule are as follows:
[today] [can with] [refund]
Wherein, include three fine-grained Feature Words: [today], [can with], [refund], these three Feature Words include Corresponding common saying, for example, [today] may include, " today, this afternoon, for a moment, at once, wait a mement " etc., can also Sorted out with carrying out more fine-grained Feature Words according to business scenario.
The extraction of Feature Words can be by keyword extraction, syntactic analysis and name entity recognition techniques automatically from text It extracts, word can also be characterized by human configuration some vocabulary, be extracted by string processing algorithm, very flexibly. Text matches rule quickly, can be configured flexibly, not need a large amount of artificial marks, thus can fast implement a semantic reason Model is solved, is built from scratch, to solve the problems, such as that new business scene is cold-started.
2) based on the semantic similarity Matching Model of attention mechanism
Chinese spoken dialog has many characteristics, such as that syntactic structure is loose, expression way is changeable, centre word is expressed the meaning obviously.Pay attention to Power mechanism is one kind in model training, and more weights are put on the vocabulary for determining text semantic, and is neglected to a certain extent The neural network mechanism of slightly common such as spoken word, conjunction.Therefore, for most effective excavation short text semantic information, The present invention has trained Bi-LSTM neural network language model (the Attention based Bi- based on attention mechanism LSTM Neural Language Model), semantic matches model is constructed, input text is subjected to vectorization expression.Meanwhile being Overcome artificial labeled data is at high cost, labeled data set is small difficulty, the present invention is using the twin network (Siamese of training Network mode) carries out fine-tune to Bi-LSTM neural network semantic matches model here, allows to maximum The adaptation service scene of degree generates the term vector expression for more adapting to current business scene.
Meanwhile on similarity labeled data, similarity labeled data each is two texts separated with comma, and The corresponding similarity manually marked, similarity is divided into 1-5,5 grades, wherein 1 indicates least similar, 5 indicate completely the same. Indicate that the present invention has trained the regression model of a convolutional neural networks (CNN), is used in conjunction with the vectorization of semantic matches model Predict the semantic similarity between two texts.When the Matching Score of two texts is (movable more than the threshold value being previously set State adjustment) when, it is believed that the content of two text expressions is consistent, that is, is determined as that text is synonymous, is based on semantic similarity, identification Input the intention of text.
3) information retrieval
The intention classification of the corresponding standard corpus database of dialog text and standard corpus is established, dialog text and mark are calculated Semantic similarity in quasi- corpus retrieves and the highest language of the dialog text semantic similarity from standard corpus library Material, the intent classifier as dialog text.What it is in interactive system is in use process, by carrying out to obtained business datum Mark supplements standard corpus library, the coverage rate and accuracy rate of interactive system semantic understanding ability can be improved, substantially increase language The optimization efficiency of sound interactive system, so that whole system achievees the effect that more with more intelligent.
4) more intent classifier models
Since Chinese spoken expressive faculty is quite abundant, same intention may there are many expression ways of multiplicity, however, We can not carry out unconfined supplement for standard corpus again.Therefore, combined standard corpus data and business of the present invention Data, generate the labeled data of more intention classifications, and Bi-LSTM network of the training one based on attention mechanism is intended to divide more as Class model carries out more classification predictions, provides the intention classification of standard corpus, provide support for information retrieval.Due to neural network The ability of feature is automatically extracted, largely, reduces dependence of the semantic understanding module to standard corpus library, and further Improve the predictablity rate of model.
Semantic understanding adaptation module
Semantic understanding adaptation module, it is intended to cost of labor is minimized, by way of transfer learning, re -training, intelligence The model in existing hybrid semantic understanding module, including Bi-LSTM semantic matches model, similarity mode model can be optimized With more intent classifier models etc., to improve the semantic understanding ability of existing business scenario.
Carry out intention mark for input text data, 1) corpus data that will be newly-increased imports Bi-LSTM semantic matches mould It in the training data of type, is trained, updates its network weight；2) cleaning filtering is carried out to newly-increased labeled data, filtered out most Excellent a part of labeled data, is mixed into standard corpus library, and carries out Intention Anticipation again to whole labeled data.According to corpus Text feature and its performance on labeled data collection, establishing has the order models and corresponding index monitoring equipment of supervision System monitors the accuracy rate and recall rate and the corpus variation of tagged corpus etc. of each intention assessment；3) newly-increased labeled data is led Enter the more intent classifier models of Bi-LSTM, update its network weight, and monitors its accuracy rate and recall rate on verifying collection；4) The online updated model of automatically dispose and tagged corpus.Semantic understanding adaptation module will update term vector, text language Adopted similarity model, standard corpus library, intent classifier etc. all realize automation, to minimize manual intervention.
Automatic dialogue management module
Automatic dialogue management module is the important module for realizing human-computer interaction controllably with quick expansion configuration, and is people The smooth natural important leverage of machine interaction.Automatic dialogue management module, based on current session state in which, current user's meaning Scheme the information interacted with history, the interactive instruction of comprehensive distinguishing output；And pass through more intention assessments, automatic dialogue management module It can realize that multiple interaction and state are converted, to greatly increase the fluency of human-computer interaction by a human-computer dialogue.
Automatic dialogue management module is combined using finite state machine (Finite-State Machine) with intensified learning Mode current session state in which, current user are intended to the letter interacted with history when executing each round interaction Breath is passed to dialogue management module together.Dialogue management module will combine preset rules of interaction and pass through what study obtained Interactive strategy, the interactive operation that should execute of output robot, such as execute next round interaction, KnowledgeBase-query, whether breaking machine Whether device people voice broadcast executes and defaults intention etc..
Meanwhile in automatic dialogue management module, it can be handed over by the interaction flow of the customized more wheel dialogues of user, i.e. setting Mutually rule quickly expands configuration to realize.
Illustrate implementation of the invention below by specific implementation scene.
Illustrate implementation of the invention by taking the communication of financial scenario as an example, is different from other traditional services industries, financial field The occasion of scape is just extremely tight to accuracy, compliance, and scene is flexible and changeable, and the objective group that every client faces is different, for interaction Demand it is also not identical, therefore to the semantic understanding accuracy, interaction flow controllability, scene of intelligent interaction robot expand Requirement on flexibility is high.And existing several technical solutions, it is difficult to meet above-mentioned several demands simultaneously.Due to financial scenario Communication feature, the present invention are to provide the intelligent speech interactive system of a complete more wheel dialogues, get through speech recognition, nature The modules such as language understanding, spatial term and speech synthesis are realized to the strict control of conversation process, can quickly be opened up Exhibition configuration, flexibly cope with distinct interaction scene, and can self-adaptation, automatic lifting natural language understanding ability, minimize Manual intervention.
With credit card debt collection, the example as a true application scenarios of the invention is illustrated, and the present invention is at this The pure language angle of scene (only considering different texts, the frequency occurred without considering text) accuracy rate of intention assessment is more than 85%, reach industry-leading level.Pass through first as the voice interactive system of the dialogue of wheel more than one in each round interaction ASR module, the text for being by the voice input transcription of user, and content of text is passed to semantic understanding module；Semantic understanding mould Block carries out processing and feature extraction to input text, and predicts its intention.By the intention of prediction, it is passed to dialogue management module, is sentenced The interactive operation that other robot should execute in next step；According to the interactive instruction that dialogue management module exports, inquiry knowledge is executed The operation such as library (optional), text generation, speech production；Voice transfer is broadcasted to client by network, such as Fig. 2 institute Show.
More wheel conversation process of credit card collection scene are as shown in figure 3, the node wherein selected with bifurcated is main stream Cheng Jiedian, grayed-out nodes are the on-hook node that process terminates, and white nodes are corresponding intention.Pass through automatic dialogue management mould Entire human-computer interaction process is together in series, to realize noninductive human-computer interaction by block according to intention type.
The present invention has developed hybrid semantic understanding module, semantic understanding adaptation module and automatic dialogue management module etc. Three important modules, realize the strict controls of more wheel dialogues, the self-promotion of natural language understanding ability, and can spirit Configuration living, meets different interaction demands.

Claims (5)

1. the more wheel dialogue intelligent speech interactive systems of one kind, it is characterized in that certainly including hybrid semantic understanding module, semantic understanding Module and automatic dialogue management module are adapted to, voice input is converted into the hybrid semantic understanding mould of text input after speech recognition Block understands that user is intended to and extracts corresponding state information, and automatic dialogue management module is intended to based on user, guides dialog procedure, Output dialog text is simultaneously converted to voice output, realizes dialogue, and semantic understanding adaptation module is used for hybrid semantic understanding mould The Optimization Learning of block；
Hybrid semantic understanding module, by the way of Model Fusion, in conjunction with text matches, semantic similarity matching, information inspection The semantic understandings schemes such as rope, more intent classifier models, comprehensive distinguishing dialog semantics, wherein text matches belong to preposition algorithm, first Sentence is pre-processed, obtains pretreated dialog text, then by semantic similarity matching, information retrieval and more intent classifiers The mode of Model Fusion, exports final result jointly；
Semantic understanding adaptation module optimizes existing hybrid semantic understanding mould by way of transfer learning, re -training Type, including Bi-LSTM language model, similarity mode model and more intent classifier models；
Automatic dialogue management module for realizing human-computer interactive control and is expanded configuration, based on current session state in which, is known Other current user is intended to the information interacted with dialog history, the interactive instruction of comprehensive distinguishing output；And by being intended to more Identification realizes that multiple interaction and state are converted by a human-computer dialogue.
2. a kind of more wheel dialogue intelligent speech interactive systems according to claim 1, it is characterized in that hybrid semantic understanding In module, text matches use the Bi-LSTM based on attention mechanism for establishing semantic understanding rule, semantic similarity matching Neural network language model establishes semantic matches model, and the dialog text combination semantic understanding rule of input is carried out vectorization table Show, and by the way of the twin network of training, accurate adjustment fine-tune is carried out to semantic matches model, is finally based on vectorization table Show the regression model of training convolutional neural networks, to predict the semantic similarity between two texts；Information retrieval passes through dialogue text The intention classification of this corresponding standard corpus database and standard corpus, is retrieved from standard corpus library based on semantic similarity With the highest corpus of dialog text semantic similarity, it is intended to intent classifier of the classification as dialog text, realizes and know Not Shu Ru text intention；The business datum of more intent classifier models coupling standard corpus data and application generates more meanings The labeled data of figure classification, training one based on the Bi-LSTM network of attention mechanism as more intent classifier models, carry out more Classification prediction, provides the intention classification of standard corpus.
3. a kind of more wheel dialogue intelligent speech interactive systems according to claim 2, it is characterized in that semantic understanding is adaptive Module is used to optimize the model in hybrid semantic understanding module, including following optimization:
1) newly-increased corpus data is imported in the training data of Bi-LSTM semantic matches model, is trained, updates its network Weight；
2) cleaning filtering is carried out to the newly-increased labeled data of more intent classifier models, filters out optimal a part of labeled data, Be mixed into standard corpus library, and Intention Anticipation again carried out to whole labeled data, according to the text feature of corpus and its marking The performance on data set is infused, the order models and corresponding index monitoring mechanism for having supervision is established, monitors each intention assessment The corpus of accuracy rate and recall rate and tagged corpus changes；
3) newly-increased labeled data is imported into the more intent classifier models of Bi-LSTM, updates its network weight, and monitored it and collect in verifying On accuracy rate and recall rate；
4) model in the online updated hybrid semantic understanding module of automatically dispose and standard corpus library.
4. a kind of more wheel dialogue intelligent speech interactive systems according to claim 1, it is characterized in that automatic dialogue management mould Block using finite state machine in such a way that intensified learning combines, execute each round dialogue interaction when, will be locating for current session State, current user be intended to information interact with history, learn in conjunction with preset dialogue rules of interaction and passing through The interactive strategy arrived carries out comprehensive descision, the interactive operation instruction that output dialogue robot should execute.
5. a kind of more wheel dialogue intelligent sound interactive devices, it is characterized in that described device is the computer dress with storage medium It sets, computer program is mounted in the storage medium, the computer program is for realizing described in claim any one of 1-4 More wheels talk with intelligent speech interactive system.
CN201910505280.4A 2019-06-12 2019-06-12 Multi-round dialogue intelligent voice interaction system and device Active CN110209791B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910505280.4A CN110209791B (en) 2019-06-12 2019-06-12 Multi-round dialogue intelligent voice interaction system and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910505280.4A CN110209791B (en) 2019-06-12 2019-06-12 Multi-round dialogue intelligent voice interaction system and device

Publications (2)

Publication Number Publication Date
CN110209791A true CN110209791A (en) 2019-09-06
CN110209791B CN110209791B (en) 2021-03-26

Family

ID=67792172

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910505280.4A Active CN110209791B (en) 2019-06-12 2019-06-12 Multi-round dialogue intelligent voice interaction system and device

Country Status (1)

Country Link
CN (1) CN110209791B (en)

Cited By (75)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110633475A (en) * 2019-09-27 2019-12-31 安徽咪鼠科技有限公司 Natural language understanding method, device and system based on computer scene and storage medium
CN110705311A (en) * 2019-09-27 2020-01-17 安徽咪鼠科技有限公司 Semantic understanding accuracy improving method, device and system applied to intelligent voice mouse and storage medium
CN110737762A (en) * 2019-10-09 2020-01-31 尹曦 old people personal information assistant system based on voice interaction
CN110765270A (en) * 2019-11-04 2020-02-07 苏州思必驰信息科技有限公司 Training method and system of text classification model for spoken language interaction
CN110781685A (en) * 2019-10-18 2020-02-11 四川长虹电器股份有限公司 Method for automatically marking correctness of semantic analysis result based on user feedback
CN110866099A (en) * 2019-10-30 2020-03-06 南昌众荟智盈信息技术有限公司 Intelligent steward service method and system based on intelligent sound box voice interaction
CN110942774A (en) * 2019-12-12 2020-03-31 北京声智科技有限公司 Man-machine interaction system, and dialogue method, medium and equipment thereof
CN110956142A (en) * 2019-12-03 2020-04-03 中国太平洋保险（集团）股份有限公司 Intelligent interactive training system
CN111078846A (en) * 2019-11-25 2020-04-28 青牛智胜(深圳)科技有限公司 Multi-turn dialog system construction method and system based on business scene
CN111090726A (en) * 2019-12-04 2020-05-01 中国南方电网有限责任公司 NLP-based electric power industry character customer service interaction method
CN111091826A (en) * 2019-12-13 2020-05-01 中博信息技术研究院有限公司 Intelligent voice robot system based on deep learning and finite-state machine
CN111090730A (en) * 2019-12-05 2020-05-01 中科数智(北京)科技有限公司 Intelligent voice scheduling system and method
CN111091011A (en) * 2019-12-20 2020-05-01 科大讯飞股份有限公司 Domain prediction method, domain prediction device and electronic equipment
CN111104502A (en) * 2019-12-24 2020-05-05 携程计算机技术（上海）有限公司 Dialogue management method, system, electronic device and storage medium for outbound system
CN111128175A (en) * 2020-01-19 2020-05-08 大连即时智能科技有限公司 Spoken language dialogue management method and system
CN111128126A (en) * 2019-12-30 2020-05-08 上海浩琨信息科技有限公司 Multi-language intelligent voice conversation method and system
CN111128147A (en) * 2019-11-18 2020-05-08 云知声智能科技股份有限公司 System and method for terminal equipment to automatically access AI multi-turn conversation capability
CN111143525A (en) * 2019-12-17 2020-05-12 广东广信通信服务有限公司 Vehicle information acquisition method and device and intelligent vehicle moving system
CN111158648A (en) * 2019-12-18 2020-05-15 西安电子科技大学 Interactive help system development method based on live-action semantic understanding and platform thereof
CN111241236A (en) * 2019-12-30 2020-06-05 新大陆数字技术股份有限公司 Task-oriented question-answering method, system, electronic equipment and readable storage medium
CN111257971A (en) * 2020-01-17 2020-06-09 河北冀云气象技术服务有限责任公司 Meteorological platform with artificial intelligence service ability and learning ability
CN111312242A (en) * 2020-02-13 2020-06-19 上海凯岸信息科技有限公司 Intelligent voice robot scheme capable of interrupting intention without influencing dialogue management
CN111314451A (en) * 2020-02-07 2020-06-19 普强时代(珠海横琴)信息技术有限公司 Language processing system based on cloud computing application
CN111324708A (en) * 2020-02-07 2020-06-23 普强时代(珠海横琴)信息技术有限公司 Natural language processing system based on human-computer interaction
CN111428483A (en) * 2020-03-31 2020-07-17 华为技术有限公司 Voice interaction method and device and terminal equipment
CN111427996A (en) * 2020-03-02 2020-07-17 云知声智能科技股份有限公司 Method and device for extracting date and time from human-computer interaction text
CN111428017A (en) * 2020-03-24 2020-07-17 科大讯飞股份有限公司 Human-computer interaction optimization method and related device
CN111462752A (en) * 2020-04-01 2020-07-28 北京思特奇信息技术股份有限公司 Client intention identification method based on attention mechanism, feature embedding and BI-L STM
CN111460118A (en) * 2020-03-26 2020-07-28 聚好看科技股份有限公司 Artificial intelligence conflict semantic recognition method and device
CN111475616A (en) * 2020-03-13 2020-07-31 平安科技（深圳）有限公司 Multi-turn dialogue method and device based on dialogue state prediction and computer equipment
CN111508488A (en) * 2020-04-13 2020-08-07 江苏止芯科技有限公司 Intelligent robot dialogue system
CN111538814A (en) * 2020-04-26 2020-08-14 云知声智能科技股份有限公司 Method for supporting custom standardization of protocol in semantic understanding
CN111611378A (en) * 2020-05-15 2020-09-01 金日泽 Behavior training dialogue control method, behavior training dialogue control system, storage medium, program, and terminal
CN111639223A (en) * 2020-05-26 2020-09-08 广东小天才科技有限公司 Audio generation method of virtual object for spoken language practice and electronic equipment
CN111833872A (en) * 2020-07-08 2020-10-27 北京声智科技有限公司 Voice control method, device, equipment, system and medium for elevator
CN111858888A (en) * 2020-07-13 2020-10-30 北京航空航天大学 Multi-round dialogue system of check-in scene
CN111949775A (en) * 2020-07-09 2020-11-17 北京声智科技有限公司 Method, device, equipment and medium for generating guide conversation
CN112017654A (en) * 2020-07-17 2020-12-01 武汉赛思云科技有限公司 Method and system for realizing non-interface office based on human-computer voice interaction
CN112035632A (en) * 2020-08-21 2020-12-04 惠州市德赛西威汽车电子股份有限公司 Preferred distribution method and system suitable for multi-conversation robot collaboration task
CN112115244A (en) * 2020-08-21 2020-12-22 深圳市欢太科技有限公司 Dialogue interaction method and device, storage medium and electronic equipment
CN112199486A (en) * 2020-10-21 2021-01-08 中国电子科技集团公司第十五研究所 Task type multi-turn conversation method and system for office scene
CN112256854A (en) * 2020-11-05 2021-01-22 云南电网有限责任公司 Intelligent AI conversation method and device based on AI natural language understanding
CN112487179A (en) * 2019-09-11 2021-03-12 珠海格力电器股份有限公司 Spoken language semantic understanding method, device and system
CN112527969A (en) * 2020-12-22 2021-03-19 上海浦东发展银行股份有限公司 Incremental intention clustering method, device, equipment and storage medium
CN112550306A (en) * 2019-09-10 2021-03-26 奥迪股份公司 Vehicle driving assistance system, vehicle including the same, and corresponding method and medium
CN112599124A (en) * 2020-11-20 2021-04-02 内蒙古电力(集团)有限责任公司电力调度控制分公司 Voice scheduling method and system for power grid scheduling
CN112686051A (en) * 2020-12-22 2021-04-20 科大讯飞股份有限公司 Semantic recognition model training method, recognition method, electronic device, and storage medium
CN112732887A (en) * 2021-01-22 2021-04-30 南京英诺森软件科技有限公司 Processing device and system for multi-turn conversation
CN112750434A (en) * 2020-12-16 2021-05-04 马上消费金融股份有限公司 Method and device for optimizing voice recognition system and electronic equipment
CN112784027A (en) * 2021-01-21 2021-05-11 军事科学院***工程研究院***总体研究所 Natural language interaction system and method in intelligent networking
CN112818097A (en) * 2021-01-26 2021-05-18 山西三友和智慧信息技术股份有限公司 Off-task training system based on dialog box state tracking model
CN112818107A (en) * 2021-02-24 2021-05-18 中国人民大学 Conversation robot for daily life and chat method thereof
CN112836030A (en) * 2021-01-29 2021-05-25 成都视海芯图微电子有限公司 Intelligent dialogue system and method
CN112905747A (en) * 2021-03-08 2021-06-04 国能大渡河流域水电开发有限公司 Professional system archive question-answering robot system based on semantic analysis technology
CN112992132A (en) * 2019-12-02 2021-06-18 浙江思考者科技有限公司 AI intelligent voice interaction program bridging one-key application applet
CN112989829A (en) * 2021-02-10 2021-06-18 海尔数字科技（上海）有限公司 Named entity identification method, device, equipment and storage medium
CN113204971A (en) * 2021-03-26 2021-08-03 南京邮电大学 Scene self-adaptive Attention multi-intention identification method based on deep learning
WO2021174717A1 (en) * 2020-03-05 2021-09-10 苏宁易购集团股份有限公司 Text intent recognition method and apparatus, computer device and storage medium
CN113409631A (en) * 2021-06-18 2021-09-17 上海锡鼎智能科技有限公司 AI auxiliary teaching robot
CN113515616A (en) * 2021-07-12 2021-10-19 中国电子科技集团公司第二十八研究所 Task driving system based on natural language
CN113689851A (en) * 2021-07-27 2021-11-23 国家电网有限公司 Scheduling professional language understanding system and method
CN113687719A (en) * 2021-08-23 2021-11-23 广东电网有限责任公司 Intelligent interaction method and device suitable for voice information
CN113935309A (en) * 2021-09-13 2022-01-14 惠州市德赛西威汽车电子股份有限公司 Skill optimization processing method and system based on semantic platform
CN114036277A (en) * 2021-11-15 2022-02-11 深圳壹账通智能科技有限公司 Dialogue robot route skipping method and device, electronic equipment and medium
CN114171021A (en) * 2021-11-29 2022-03-11 国网江苏省电力有限公司南京供电分公司 Distribution network virtual scheduling system and method based on artificial intelligence
CN114582314A (en) * 2022-02-28 2022-06-03 江苏楷文电信技术有限公司 ASR-based human-computer audio-video interaction logic model design method
CN114706965A (en) * 2022-03-22 2022-07-05 广州营客信息科技有限公司 AI intelligent customer service system
CN114970559A (en) * 2022-05-18 2022-08-30 马上消费金融股份有限公司 Intelligent response method and device
CN115168593A (en) * 2022-09-05 2022-10-11 深圳爱莫科技有限公司 Intelligent dialogue management system, method and processing equipment capable of self-learning
CN115794065A (en) * 2022-11-01 2023-03-14 中犇科技有限公司 Visual intelligent programming method based on AI voice interaction
CN115936011A (en) * 2022-12-28 2023-04-07 南京易米云通网络科技有限公司 Multi-intention semantic recognition method in intelligent conversation
US11798545B2 (en) 2020-01-08 2023-10-24 Beijing Xiaomi Pinecone Electronics Co., Ltd. Speech interaction method and apparatus, device and storage medium
CN117648408A (en) * 2024-01-30 2024-03-05 北京水滴科技集团有限公司 Intelligent question-answering method and device based on large model, electronic equipment and storage medium
CN117671212A (en) * 2023-12-13 2024-03-08 江苏麦克数字空间营造有限公司 Exhibition hall exhibition system based on meta universe and interaction method thereof
CN117807215A (en) * 2024-03-01 2024-04-02 青岛海尔科技有限公司 Statement multi-intention recognition method, device and equipment based on model

Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106354835A (en) * 2016-08-31 2017-01-25 上海交通大学 Artificial dialogue auxiliary system based on context semantic understanding
CN106407178A (en) * 2016-08-25 2017-02-15 中国科学院计算技术研究所 Session abstract generation method and device
CN107193978A (en) * 2017-05-26 2017-09-22 武汉泰迪智慧科技有限公司 A kind of many wheel automatic chatting dialogue methods and system based on deep learning
CN107346340A (en) * 2017-07-04 2017-11-14 北京奇艺世纪科技有限公司 A kind of user view recognition methods and system
CN107515944A (en) * 2017-08-31 2017-12-26 广东美的制冷设备有限公司 Exchange method, user terminal and storage medium based on artificial intelligence
CN108228764A (en) * 2017-12-27 2018-06-29 神思电子技术股份有限公司 A kind of single-wheel dialogue and the fusion method of more wheel dialogues
US20180232376A1 (en) * 2017-02-16 2018-08-16 Microsoft Technology Licensing, Llc Conversational virtual assistant
CN108415923A (en) * 2017-10-18 2018-08-17 北京邮电大学 The intelligent interactive system of closed domain
CN108717439A (en) * 2018-05-16 2018-10-30 哈尔滨理工大学 A kind of Chinese Text Categorization merged based on attention mechanism and characteristic strengthening
CN108829667A (en) * 2018-05-28 2018-11-16 南京柯基数据科技有限公司 It is a kind of based on memory network more wheels dialogue under intension recognizing method
CN108874782A (en) * 2018-06-29 2018-11-23 北京寻领科技有限公司 A kind of more wheel dialogue management methods of level attention LSTM and knowledge mapping
CN108874972A (en) * 2018-06-08 2018-11-23 青岛里奥机器人技术有限公司 A kind of more wheel emotion dialogue methods based on deep learning
CN109241255A (en) * 2018-08-20 2019-01-18 华中师范大学 A kind of intension recognizing method based on deep learning
CN109522393A (en) * 2018-10-11 2019-03-26 平安科技（深圳）有限公司 Intelligent answer method, apparatus, computer equipment and storage medium

Patent Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106407178A (en) * 2016-08-25 2017-02-15 中国科学院计算技术研究所 Session abstract generation method and device
CN106354835A (en) * 2016-08-31 2017-01-25 上海交通大学 Artificial dialogue auxiliary system based on context semantic understanding
US20180232376A1 (en) * 2017-02-16 2018-08-16 Microsoft Technology Licensing, Llc Conversational virtual assistant
CN107193978A (en) * 2017-05-26 2017-09-22 武汉泰迪智慧科技有限公司 A kind of many wheel automatic chatting dialogue methods and system based on deep learning
CN107346340A (en) * 2017-07-04 2017-11-14 北京奇艺世纪科技有限公司 A kind of user view recognition methods and system
CN107515944A (en) * 2017-08-31 2017-12-26 广东美的制冷设备有限公司 Exchange method, user terminal and storage medium based on artificial intelligence
CN108415923A (en) * 2017-10-18 2018-08-17 北京邮电大学 The intelligent interactive system of closed domain
CN108228764A (en) * 2017-12-27 2018-06-29 神思电子技术股份有限公司 A kind of single-wheel dialogue and the fusion method of more wheel dialogues
CN108717439A (en) * 2018-05-16 2018-10-30 哈尔滨理工大学 A kind of Chinese Text Categorization merged based on attention mechanism and characteristic strengthening
CN108829667A (en) * 2018-05-28 2018-11-16 南京柯基数据科技有限公司 It is a kind of based on memory network more wheels dialogue under intension recognizing method
CN108874972A (en) * 2018-06-08 2018-11-23 青岛里奥机器人技术有限公司 A kind of more wheel emotion dialogue methods based on deep learning
CN108874782A (en) * 2018-06-29 2018-11-23 北京寻领科技有限公司 A kind of more wheel dialogue management methods of level attention LSTM and knowledge mapping
CN109241255A (en) * 2018-08-20 2019-01-18 华中师范大学 A kind of intension recognizing method based on deep learning
CN109522393A (en) * 2018-10-11 2019-03-26 平安科技（深圳）有限公司 Intelligent answer method, apparatus, computer equipment and storage medium

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
刘娇等: "人机对话***中意图识别方法综述", 《计算机工程与应用》 *
史梦飞: "基于中文知识库的问答***研究与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *
洪源: "基于深度学习的智能问答***设计", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *
黄佳伟: "人机对话***中用户意图分类方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *

Cited By (107)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112550306A (en) * 2019-09-10 2021-03-26 奥迪股份公司 Vehicle driving assistance system, vehicle including the same, and corresponding method and medium
CN112487179B (en) * 2019-09-11 2024-05-31 珠海格力电器股份有限公司 Spoken language semantic understanding method, device and system
CN112487179A (en) * 2019-09-11 2021-03-12 珠海格力电器股份有限公司 Spoken language semantic understanding method, device and system
CN110705311A (en) * 2019-09-27 2020-01-17 安徽咪鼠科技有限公司 Semantic understanding accuracy improving method, device and system applied to intelligent voice mouse and storage medium
CN110633475A (en) * 2019-09-27 2019-12-31 安徽咪鼠科技有限公司 Natural language understanding method, device and system based on computer scene and storage medium
CN110705311B (en) * 2019-09-27 2022-11-25 安徽咪鼠科技有限公司 Semantic understanding accuracy improving method, device and system applied to intelligent voice mouse and storage medium
CN110737762A (en) * 2019-10-09 2020-01-31 尹曦 old people personal information assistant system based on voice interaction
CN110781685A (en) * 2019-10-18 2020-02-11 四川长虹电器股份有限公司 Method for automatically marking correctness of semantic analysis result based on user feedback
CN110866099A (en) * 2019-10-30 2020-03-06 南昌众荟智盈信息技术有限公司 Intelligent steward service method and system based on intelligent sound box voice interaction
CN110765270A (en) * 2019-11-04 2020-02-07 苏州思必驰信息科技有限公司 Training method and system of text classification model for spoken language interaction
CN110765270B (en) * 2019-11-04 2022-07-01 思必驰科技股份有限公司 Training method and system of text classification model for spoken language interaction
CN111128147A (en) * 2019-11-18 2020-05-08 云知声智能科技股份有限公司 System and method for terminal equipment to automatically access AI multi-turn conversation capability
CN111078846A (en) * 2019-11-25 2020-04-28 青牛智胜(深圳)科技有限公司 Multi-turn dialog system construction method and system based on business scene
CN112992132A (en) * 2019-12-02 2021-06-18 浙江思考者科技有限公司 AI intelligent voice interaction program bridging one-key application applet
CN110956142A (en) * 2019-12-03 2020-04-03 中国太平洋保险（集团）股份有限公司 Intelligent interactive training system
CN111090726A (en) * 2019-12-04 2020-05-01 中国南方电网有限责任公司 NLP-based electric power industry character customer service interaction method
CN111090730A (en) * 2019-12-05 2020-05-01 中科数智(北京)科技有限公司 Intelligent voice scheduling system and method
CN110942774A (en) * 2019-12-12 2020-03-31 北京声智科技有限公司 Man-machine interaction system, and dialogue method, medium and equipment thereof
CN111091826A (en) * 2019-12-13 2020-05-01 中博信息技术研究院有限公司 Intelligent voice robot system based on deep learning and finite-state machine
CN111091826B (en) * 2019-12-13 2023-09-01 中博信息技术研究院有限公司 Intelligent voice robot system based on deep learning and finite state machine
CN111143525A (en) * 2019-12-17 2020-05-12 广东广信通信服务有限公司 Vehicle information acquisition method and device and intelligent vehicle moving system
CN111158648A (en) * 2019-12-18 2020-05-15 西安电子科技大学 Interactive help system development method based on live-action semantic understanding and platform thereof
CN111158648B (en) * 2019-12-18 2023-04-07 西安电子科技大学 Interactive help system development method based on live-action semantic understanding and platform thereof
CN111091011A (en) * 2019-12-20 2020-05-01 科大讯飞股份有限公司 Domain prediction method, domain prediction device and electronic equipment
CN111104502A (en) * 2019-12-24 2020-05-05 携程计算机技术（上海）有限公司 Dialogue management method, system, electronic device and storage medium for outbound system
CN111128126B (en) * 2019-12-30 2023-04-07 海智讯通(上海)智能科技有限公司 Multi-language intelligent voice conversation method and system
CN111241236B (en) * 2019-12-30 2023-08-22 新大陆数字技术股份有限公司 Task-oriented question-answering method, system, electronic device and readable storage medium
CN111241236A (en) * 2019-12-30 2020-06-05 新大陆数字技术股份有限公司 Task-oriented question-answering method, system, electronic equipment and readable storage medium
CN111128126A (en) * 2019-12-30 2020-05-08 上海浩琨信息科技有限公司 Multi-language intelligent voice conversation method and system
US11798545B2 (en) 2020-01-08 2023-10-24 Beijing Xiaomi Pinecone Electronics Co., Ltd. Speech interaction method and apparatus, device and storage medium
CN111257971A (en) * 2020-01-17 2020-06-09 河北冀云气象技术服务有限责任公司 Meteorological platform with artificial intelligence service ability and learning ability
CN111128175A (en) * 2020-01-19 2020-05-08 大连即时智能科技有限公司 Spoken language dialogue management method and system
CN111324708A (en) * 2020-02-07 2020-06-23 普强时代(珠海横琴)信息技术有限公司 Natural language processing system based on human-computer interaction
CN111314451A (en) * 2020-02-07 2020-06-19 普强时代(珠海横琴)信息技术有限公司 Language processing system based on cloud computing application
CN111312242A (en) * 2020-02-13 2020-06-19 上海凯岸信息科技有限公司 Intelligent voice robot scheme capable of interrupting intention without influencing dialogue management
CN111427996B (en) * 2020-03-02 2023-10-20 云知声智能科技股份有限公司 Method and device for extracting date and time from man-machine interaction text
CN111427996A (en) * 2020-03-02 2020-07-17 云知声智能科技股份有限公司 Method and device for extracting date and time from human-computer interaction text
WO2021174717A1 (en) * 2020-03-05 2021-09-10 苏宁易购集团股份有限公司 Text intent recognition method and apparatus, computer device and storage medium
CN111475616B (en) * 2020-03-13 2023-08-22 平安科技（深圳）有限公司 Multi-round dialogue method and device based on dialogue state prediction and computer equipment
CN111475616A (en) * 2020-03-13 2020-07-31 平安科技（深圳）有限公司 Multi-turn dialogue method and device based on dialogue state prediction and computer equipment
CN111428017A (en) * 2020-03-24 2020-07-17 科大讯飞股份有限公司 Human-computer interaction optimization method and related device
CN111428017B (en) * 2020-03-24 2022-12-02 科大讯飞股份有限公司 Human-computer interaction optimization method and related device
CN111460118B (en) * 2020-03-26 2023-10-20 聚好看科技股份有限公司 Artificial intelligence conflict semantic recognition method and device
CN111460118A (en) * 2020-03-26 2020-07-28 聚好看科技股份有限公司 Artificial intelligence conflict semantic recognition method and device
CN111428483A (en) * 2020-03-31 2020-07-17 华为技术有限公司 Voice interaction method and device and terminal equipment
CN111462752A (en) * 2020-04-01 2020-07-28 北京思特奇信息技术股份有限公司 Client intention identification method based on attention mechanism, feature embedding and BI-L STM
CN111462752B (en) * 2020-04-01 2023-10-13 北京思特奇信息技术股份有限公司 Attention mechanism, feature embedding and BI-LSTM (business-to-business) based customer intention recognition method
CN111508488A (en) * 2020-04-13 2020-08-07 江苏止芯科技有限公司 Intelligent robot dialogue system
CN111538814A (en) * 2020-04-26 2020-08-14 云知声智能科技股份有限公司 Method for supporting custom standardization of protocol in semantic understanding
CN111538814B (en) * 2020-04-26 2024-03-08 云知声智能科技股份有限公司 Method for supporting custom standardization by protocol in semantic understanding
CN111611378A (en) * 2020-05-15 2020-09-01 金日泽 Behavior training dialogue control method, behavior training dialogue control system, storage medium, program, and terminal
CN111639223A (en) * 2020-05-26 2020-09-08 广东小天才科技有限公司 Audio generation method of virtual object for spoken language practice and electronic equipment
CN111639223B (en) * 2020-05-26 2024-04-19 广东小天才科技有限公司 Audio generation method of virtual object for spoken language exercise and electronic equipment
CN111833872A (en) * 2020-07-08 2020-10-27 北京声智科技有限公司 Voice control method, device, equipment, system and medium for elevator
CN111949775A (en) * 2020-07-09 2020-11-17 北京声智科技有限公司 Method, device, equipment and medium for generating guide conversation
CN111949775B (en) * 2020-07-09 2024-06-11 北京声智科技有限公司 Method, device, equipment and medium for generating guide dialogue
CN111858888B (en) * 2020-07-13 2023-05-30 北京航空航天大学 Multi-round dialogue system of check-in scene
CN111858888A (en) * 2020-07-13 2020-10-30 北京航空航天大学 Multi-round dialogue system of check-in scene
CN112017654A (en) * 2020-07-17 2020-12-01 武汉赛思云科技有限公司 Method and system for realizing non-interface office based on human-computer voice interaction
CN112115244B (en) * 2020-08-21 2024-05-03 深圳市欢太科技有限公司 Dialogue interaction method and device, storage medium and electronic equipment
CN112115244A (en) * 2020-08-21 2020-12-22 深圳市欢太科技有限公司 Dialogue interaction method and device, storage medium and electronic equipment
CN112035632A (en) * 2020-08-21 2020-12-04 惠州市德赛西威汽车电子股份有限公司 Preferred distribution method and system suitable for multi-conversation robot collaboration task
CN112199486A (en) * 2020-10-21 2021-01-08 中国电子科技集团公司第十五研究所 Task type multi-turn conversation method and system for office scene
CN112256854A (en) * 2020-11-05 2021-01-22 云南电网有限责任公司 Intelligent AI conversation method and device based on AI natural language understanding
CN112599124A (en) * 2020-11-20 2021-04-02 内蒙古电力(集团)有限责任公司电力调度控制分公司 Voice scheduling method and system for power grid scheduling
CN112750434A (en) * 2020-12-16 2021-05-04 马上消费金融股份有限公司 Method and device for optimizing voice recognition system and electronic equipment
CN112686051A (en) * 2020-12-22 2021-04-20 科大讯飞股份有限公司 Semantic recognition model training method, recognition method, electronic device, and storage medium
CN112686051B (en) * 2020-12-22 2024-05-31 中国科学技术大学 Semantic recognition model training method, recognition method, electronic device and storage medium
CN112527969B (en) * 2020-12-22 2022-11-15 上海浦东发展银行股份有限公司 Incremental intention clustering method, device, equipment and storage medium
CN112527969A (en) * 2020-12-22 2021-03-19 上海浦东发展银行股份有限公司 Incremental intention clustering method, device, equipment and storage medium
CN112784027A (en) * 2021-01-21 2021-05-11 军事科学院***工程研究院***总体研究所 Natural language interaction system and method in intelligent networking
CN112784027B (en) * 2021-01-21 2024-05-14 军事科学院***工程研究院***总体研究所 Natural language interaction system and method in intelligent networking
CN112732887A (en) * 2021-01-22 2021-04-30 南京英诺森软件科技有限公司 Processing device and system for multi-turn conversation
CN112818097A (en) * 2021-01-26 2021-05-18 山西三友和智慧信息技术股份有限公司 Off-task training system based on dialog box state tracking model
CN112836030A (en) * 2021-01-29 2021-05-25 成都视海芯图微电子有限公司 Intelligent dialogue system and method
CN112989829B (en) * 2021-02-10 2024-03-08 卡奥斯数字科技(上海)有限公司 Named entity recognition method, device, equipment and storage medium
CN112989829A (en) * 2021-02-10 2021-06-18 海尔数字科技（上海）有限公司 Named entity identification method, device, equipment and storage medium
CN112818107B (en) * 2021-02-24 2023-10-31 中国人民大学 Conversation robot for daily life and chat method thereof
CN112818107A (en) * 2021-02-24 2021-05-18 中国人民大学 Conversation robot for daily life and chat method thereof
CN112905747A (en) * 2021-03-08 2021-06-04 国能大渡河流域水电开发有限公司 Professional system archive question-answering robot system based on semantic analysis technology
CN113204971B (en) * 2021-03-26 2024-01-26 南京邮电大学 Scene self-adaptive Attention multi-intention recognition method based on deep learning
CN113204971A (en) * 2021-03-26 2021-08-03 南京邮电大学 Scene self-adaptive Attention multi-intention identification method based on deep learning
CN113409631A (en) * 2021-06-18 2021-09-17 上海锡鼎智能科技有限公司 AI auxiliary teaching robot
CN113515616B (en) * 2021-07-12 2024-05-14 中国电子科技集团公司第二十八研究所 Task driving system based on natural language
CN113515616A (en) * 2021-07-12 2021-10-19 中国电子科技集团公司第二十八研究所 Task driving system based on natural language
CN113689851B (en) * 2021-07-27 2024-02-02 国家电网有限公司 Scheduling professional language understanding system and method
CN113689851A (en) * 2021-07-27 2021-11-23 国家电网有限公司 Scheduling professional language understanding system and method
CN113687719A (en) * 2021-08-23 2021-11-23 广东电网有限责任公司 Intelligent interaction method and device suitable for voice information
CN113935309A (en) * 2021-09-13 2022-01-14 惠州市德赛西威汽车电子股份有限公司 Skill optimization processing method and system based on semantic platform
CN114036277A (en) * 2021-11-15 2022-02-11 深圳壹账通智能科技有限公司 Dialogue robot route skipping method and device, electronic equipment and medium
CN114171021A (en) * 2021-11-29 2022-03-11 国网江苏省电力有限公司南京供电分公司 Distribution network virtual scheduling system and method based on artificial intelligence
CN114582314B (en) * 2022-02-28 2023-06-23 江苏楷文电信技术有限公司 Man-machine audio-video interaction logic model design method based on ASR
CN114582314A (en) * 2022-02-28 2022-06-03 江苏楷文电信技术有限公司 ASR-based human-computer audio-video interaction logic model design method
CN114706965A (en) * 2022-03-22 2022-07-05 广州营客信息科技有限公司 AI intelligent customer service system
CN114706965B (en) * 2022-03-22 2022-11-11 广州营客信息科技有限公司 AI intelligent customer service system
CN114970559B (en) * 2022-05-18 2024-02-02 马上消费金融股份有限公司 Intelligent response method and device
CN114970559A (en) * 2022-05-18 2022-08-30 马上消费金融股份有限公司 Intelligent response method and device
CN115168593A (en) * 2022-09-05 2022-10-11 深圳爱莫科技有限公司 Intelligent dialogue management system, method and processing equipment capable of self-learning
CN115794065A (en) * 2022-11-01 2023-03-14 中犇科技有限公司 Visual intelligent programming method based on AI voice interaction
CN115794065B (en) * 2022-11-01 2023-11-03 中犇科技有限公司 Visual intelligent programming method based on AI voice interaction
CN115936011B (en) * 2022-12-28 2023-10-20 南京易米云通网络科技有限公司 Multi-intention semantic recognition method in intelligent dialogue
CN115936011A (en) * 2022-12-28 2023-04-07 南京易米云通网络科技有限公司 Multi-intention semantic recognition method in intelligent conversation
CN117671212A (en) * 2023-12-13 2024-03-08 江苏麦克数字空间营造有限公司 Exhibition hall exhibition system based on meta universe and interaction method thereof
CN117648408B (en) * 2024-01-30 2024-04-30 北京水滴科技集团有限公司 Intelligent question-answering method and device based on large model, electronic equipment and storage medium
CN117648408A (en) * 2024-01-30 2024-03-05 北京水滴科技集团有限公司 Intelligent question-answering method and device based on large model, electronic equipment and storage medium
CN117807215A (en) * 2024-03-01 2024-04-02 青岛海尔科技有限公司 Statement multi-intention recognition method, device and equipment based on model
CN117807215B (en) * 2024-03-01 2024-05-24 青岛海尔科技有限公司 Statement multi-intention recognition method, device and equipment based on model

Also Published As

Publication number Publication date
CN110209791B (en) 2021-03-26

Similar Documents

Publication Publication Date Title
CN110209791A (en) It is a kind of to take turns dialogue intelligent speech interactive system and device more
CN106328166B (en) Human-computer dialogue abnormality detection system and method
CN103458056B (en) Speech intention judging system based on automatic classification technology for automatic outbound system
CN106297773B (en) A kind of neural network acoustic training model method
CN106951468A (en) Talk with generation method and device
CN111177310B (en) Intelligent scene conversation method and device for power service robot
CN109446306A (en) A kind of intelligent answer method of more wheels dialogue of task based access control driving
CN106503805A (en) A kind of bimodal based on machine learning everybody talk with sentiment analysis system and method
CN106357942A (en) Intelligent response method and system based on context dialogue semantic recognition
CN111651609A (en) Multi-turn dialogue method and system integrating knowledge graph and emotion supervision
CN105810200A (en) Man-machine dialogue apparatus and method based on voiceprint identification
CN110059169B (en) Intelligent robot chat context implementation method and system based on corpus labeling
CN111429915A (en) Scheduling system and scheduling method based on voice recognition
CN109460459A (en) A kind of conversational system automatic optimization method based on log study
CN105244042B (en) A kind of speech emotional interactive device and method based on finite-state automata
WO2022022746A1 (en) Intent recognition method and intent recognition system having self learning capability
CN110019688A (en) The method that robot is trained
CN109637526A (en) The adaptive approach of DNN acoustic model based on personal identification feature
CN110321564A (en) A kind of more wheel dialogue intension recognizing methods
CN110442700A (en) Man-machine more wheel dialogue methods and system, smart machine for human-computer interaction
CN111914555B (en) Automatic relation extraction system based on Transformer structure
CN109829729A (en) A kind of intelligence outgoing call system and method
CN106127526A (en) Intelligent robot system and method for work thereof
CN110046232A (en) Natural expression processing method, response method, equipment and the system of natural intelligence
CN106653020A (en) Multi-business control method and system for smart sound and video equipment based on deep learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant