CN108153780A - A kind of human-computer dialogue device and its interactive method of realization - Google Patents
A kind of human-computer dialogue device and its interactive method of realization Download PDFInfo
- Publication number
- CN108153780A CN108153780A CN201611106354.XA CN201611106354A CN108153780A CN 108153780 A CN108153780 A CN 108153780A CN 201611106354 A CN201611106354 A CN 201611106354A CN 108153780 A CN108153780 A CN 108153780A
- Authority
- CN
- China
- Prior art keywords
- user
- answer
- inquiry
- question
- characterization information
- 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
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
- G06F16/3346—Query execution using probabilistic model
-
- 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
Abstract
This application discloses a kind of human-computer dialogue device and its realize interactive method, including：In the dialog procedure for providing service to the user, the inquiry of user is obtained；Based on the problems in user's inquiry and issue database, problem is proposed to user, and user view is determined according to the answer of user by the way of mostly wheel question and answer.The technical solution provided by the application in uncertain user view, clarifies user view to user's enquirement by active, intelligently realizes human-computer dialogue process, improve the accuracy rate identified to user view.
Description
Technical field
This application involves interactive, espespecially a kind of human-computer dialogue device and its interactive method of realization.
Background technology
Intelligent interactive system using natural language as interaction medium, provides service to the user.Intelligent human-computer dialogue
System is the interactive mode of a kind of completely new people and machine, and interactive process is more natural, more efficient.For example, just there are work in the sixties
The Eliza systems of name come out, and can simply be talked with user, it seems to be handed over true man to feel many users
Stream.After apple Siri, there are many dialogue products to occur, including Google Now, Microsoft little Na (Cortana).It is domestic that there are many chat
Its machine human hair cloth, such as the small ice of Microsoft, is attracted attention extensively.
Existing intelligence interactive system only does single-wheel dialogue mostly, the inquiry for the service object of user, that is, business
It asks, intelligent interactive system is generally handled in two stages：First stage, identify user view, user view refers to
User wishes the plan for reaching certain purpose, such as：Wish to solve the problems, such as certain, it is desirable to complete some task, wish to reach certain
A target etc.；Second stage, AGENT are to simulate the machine that customer service provides customer service, are solved accordingly for user view
It answers or operates.Here, accurately identification user view is the key that determine entire intelligent interactive system service effectiveness.With
Family intention assessment stage, existing intelligence interactive system are usually directly according to the intention of the statement identification user of user.It is whole
A user view identification process is passive, that is to say, that AGENT will not propose any problem to the inquiry of user, only quilt
The intention of user dynamicly is determined based on the inquiry of user.In this way, if the statement of user is unclear, then, intelligent people
The user view of machine system identification also can be inaccurate, and the answer or behavior that AGENT is provided accordingly can not reach the meaning of user
Figure even results in mistake and risk.
From the point of view of the realization of above-mentioned existing intelligent interactive system, it is impossible to which it is accurate really to ensure to identify user view
Rate, and the realization of existing intelligent interactive system largely relies on artificial participation, virtually improves cost, and realize week
Phase is long.
Invention content
In order to solve the above-mentioned technical problem, this application provides a kind of human-computer dialogue device and its interactive sides of realization
Method can ensure the accuracy rate identified to user view, reduce cost of implementation, shorten performance period.
In order to reach the application purpose, the application provides a kind of human-computer dialogue device, including：Acquiring unit, response unit,
And the problem of storage problem database；Wherein,
Acquiring unit, in the dialog procedure for providing service to the user, obtaining the inquiry of user；
Response unit, for being based on user's inquiry and the problems in issue database, by the way of mostly wheel question and answer to
Family proposes problem, and determines user view according to the answer of user.
Optionally, it further includes：
Unit for determining the problem of dialogue based on talking with language material sample, and is stored in described problem data
In library.
Optionally, the unit is specifically used for：
From it is described dialogue language material sample in extraction customer service to user propose the problem of, by text cluster, opposite user carries
The problem of going out be classified and stored in described problem database.
Optionally, the response unit includes：Intention Anticipation module is intended to decision-making module；Wherein,
Intention Anticipation module, it is corresponding with the inquiry in described problem database for being predicted based on the inquiry of the user
Each question answering certainly probability, formed represent user view characterization information；
Be intended to decision-making module, for based on characterization information in a manner of more wheel question and answer, using user to it is described to
The answer for the problem of family proposes updates the probability that the user predicted in the characterization information answers affirmative；And for basis
The characterization information obtained after more wheel question and answer determines the user view.
Optionally, the intention decision-making module, including：Problem decision submodule and intention decision submodule, wherein,
Problem decision submodule, in the often wheel question and answer of more wheel question and answer, prediction to be obtained in the characterization information
The information gain of each problem, using information gain it is maximum the problem of as being proposed to user the problem of, and to user's proposition problem；
Be intended to decision submodule, for user to it is described to user proposes the problem of the answer update characterization information
The middle probability predicted obtained user and answer affirmative；Until determining user view according to updated characterization information.
Optionally, the Intention Anticipation module is specifically used for：
Based on the inquiry of the user, from dialogue language material learning to each problem and for representing to reach user view
Association between one answer or the aims of systems of behavior；It is predicted based on the question and answer distribution situation in association to described problem data
The probability of each question answering corresponding with the inquiry certainly in library, forms the characterization information for representing user view.
Optionally, the mapping pass for being intended to be stored with pre-set characterization information and user view in decision submodule
System；
It is intended to decision submodule to be specifically used for：According to the mapping relations, the corresponding characterization letter currently formed is found out
The user view of breath.
Interactive method is realized present invention also provides a kind of, including：
In the dialog procedure for providing service to the user, the inquiry of user is obtained；
Based on the problems in user's inquiry and issue database, problem is proposed to user by the way of mostly wheel question and answer, and
User view is determined according to the answer of user.
Optionally, this method further includes：The problem of dialogue is determined based on talking with language material sample, and it is stored in described ask
It inscribes in database.
Optionally, include the problem of the determining dialogue：
From it is described dialogue language material sample in extraction customer service to user propose the problem of, by text cluster, opposite user carries
The problem of going out be classified and stored in described problem database.
Optionally, it is described based on the problems in user's inquiry and issue database, to user by the way of mostly wheel question and answer
It is proposed problem includes：
Each question answering corresponding with the inquiry is agreed in inquiry prediction described problem database based on the user
Fixed probability forms the characterization information for representing user view；
Based on characterization information in a manner of more wheel question and answer, using user to it is described to user propose the problem of answer
Update the probability that the user predicted in the characterization information answers affirmative；And believed according to the characterization obtained after more wheel question and answer
Breath determines the user view.
Optionally, it is described using user to it is described to user propose the problem of answer update in the characterization information and predict
Obtained user answers the probability of affirmative, determines that the user view includes according to the characterization information obtained after more wheel question and answer：
In the often wheel question and answer of more wheel question and answer, prediction obtains the information gain of each problem in the characterization information,
Using information gain it is maximum the problem of as next round in more wheel question and answer to user proposes the problem of, and asked to user's proposition
Topic；
With user to it is described to user propose the problem of answer update the user predicted in the characterization information and return
Answer the probability of affirmative；Until determining user view according to updated characterization information.
Optionally, it is corresponding with the inquiry each in the inquiry prediction described problem database based on the user
The probability of question answering certainly forms and represents that the characterization information of user view includes：
Based on the inquiry of the user, from dialogue language material learning to each problem and for representing to reach user view
Association between one answer or the aims of systems of behavior；It is predicted based on the question and answer distribution situation in association to described problem data
The probability of each question answering corresponding with the inquiry certainly, forms the characterization information for representing user view in library.
Optionally, according to the characterization information of the expression user view of formation, determine that the user view includes：
According to pre-set characterization information and the mapping relations of user view, the corresponding characterization currently formed is found out
The user view of information.
The application provides one kind and is used to implement interactive device again, including at least memory and processor, wherein,
Following executable instruction is stored in memory：In the dialog procedure for providing service to the user, obtain user's
Inquiry；Based on user's inquiry and the problems in issue database, problem is proposed to user by the way of mostly wheel question and answer, and according to
The answer of user determines user view.
The scheme that the application provides includes：In the dialog procedure for providing service to the user, the inquiry of user is obtained；It is based on
The problems in user's inquiry and issue database propose problem, and answering according to user by the way of mostly wheel question and answer to user
User view is determined again.The interactive technical solution of realization that the application provides, in the working method of system, uncertain
During user view, user view is clarified to user's enquirement by active, intelligently realizes human-computer dialogue process, improved pair
The accuracy rate of user view identification.
Other features and advantage will illustrate in the following description, also, partly become from specification
It obtains it is clear that being understood by implementing the application.The purpose of the application and other advantages can be by specification, rights
Specifically noted structure is realized and is obtained in claim and attached drawing.
Description of the drawings
Attached drawing is used for providing further understanding technical scheme, and a part for constitution instruction, with this
The embodiment of application for explaining the technical solution of the application, does not form the limitation to technical scheme together.
Fig. 1 is the composition structure diagram of the man-machine Interface of the application；
Fig. 2 (a) is schematic diagram of the application for the distribution first embodiment for the answer putd question to；
Fig. 2 (b) is schematic diagram of the application for the distribution second embodiment for the answer putd question to；
Fig. 2 (c) is schematic diagram of the application for the distribution 3rd embodiment for the answer putd question to；
Fig. 2 (d) is schematic diagram of the application for the distribution fourth embodiment for the answer putd question to；
Fig. 3 is the flow chart that the man-machine Interface of the application realizes interactive method；
Fig. 4 is the flow diagram for the embodiment that the application realizes interactive method.
Specific embodiment
Purpose, technical scheme and advantage to make the application are more clearly understood, below in conjunction with attached drawing to the application
Embodiment be described in detail.It should be noted that in the absence of conflict, in the embodiment and embodiment in the application
Feature mutually can arbitrarily combine.
In a typical configuration of this application, computing device includes one or more processors (CPU), input/output
Interface, network interface and memory.
Memory may include computer-readable medium in volatile memory, random access memory (RAM) and/or
The forms such as Nonvolatile memory, such as read-only memory (ROM) or flash memory (flash RAM).Memory is computer-readable medium
Example.
Computer-readable medium includes permanent and non-permanent, removable and non-removable media can be by any method
Or technology come realize information store.Information can be computer-readable instruction, data structure, the module of program or other data.
The example of the storage medium of computer includes, but are not limited to phase transition internal memory (PRAM), static RAM (SRAM), moves
State random access memory (DRAM), other kinds of random access memory (RAM), read-only memory (ROM), electric erasable
Programmable read only memory (EEPROM), fast flash memory bank or other memory techniques, CD-ROM read-only memory (CD-ROM),
Digital versatile disc (DVD) or other optical storages, magnetic tape cassette, the storage of tape magnetic rigid disk or other magnetic storage apparatus
Or any other non-transmission medium, available for storing the information that can be accessed by a computing device.It defines, calculates according to herein
Machine readable medium does not include the data-signal and carrier wave of non-temporary computer readable media (transitory media), such as modulation.
Step shown in the flowchart of the accompanying drawings can be in the computer system of such as a group of computer-executable instructions
It performs.Also, although logical order is shown in flow charts, it in some cases, can be to be different from herein suitable
Sequence performs shown or described step.
Fig. 1 is the composition structure diagram of the man-machine Interface of the application, as shown in Figure 1, including acquiring unit, question and answer list
The problem of member and storage problem database；Wherein,
Acquiring unit, in the dialog procedure for providing service to the user, obtaining the inquiry of user；
Response unit, for being based on user's inquiry and the problems in issue database, by the way of mostly wheel question and answer to
Family proposes problem, and determines user view according to the answer of user.
Wherein, the mode of mostly wheel question and answer is exactly the process of the question-response by once or more.Specifically, this reality
A wheel question and answer in example are applied, can specifically refer to that human-computer dialogue device proposes problem to user, user is based on making the problem of proposition
It replies.
It is the problem of storage in issue database：Based on talking with language material sample, by manually learning, cluster analysis,
Or the method for the machine learning inquiry for user that learns the problem of proposing.Dialogue language material sample can be as：Really
User and customer service about the dialogue language material for the natural language for reaching user view or interactive dialog history data etc.
Deng.And the problems in the issue database, it can be updated in real time, specifically, can be according to pair collected for a period of time
Data are talked about, by machine learning or other modes come the problem of determining the inquiry proposition for user.In this way, in issue database
The problem of be that can automatically feedback by the question and answer language material accumulated in service process is provided to the user, be formed with real-time update,
The performance of sustained improvement human-computer dialogue device.The problems in issue database includes but not limited to：The matter of right and wrong refers in particular to problem,
Select permeability etc..
Optionally, the man-machine Interface of the application may also include unit, for true based on talking with language material sample
Surely the problem of talking with, for example learn the problem of inquiry for user proposes, and be stored in described problem database.
By unit, can problem in issue database be obtained based on the language material of history.
For the process by manually learning, learn to may include the problem of the inquiry proposition for user：Assuming that user
Inquiry：" how the xx business that may I ask me has more than is needed", by manually learning, the problem of customer service may propose, has：" you are open-minded
Xx business", " not opening is" " you pay correlative charges", " how to open xx business is" etc.,
These problems can be grouped into same category, formed in issue database for inquiry xx business a series of propositions the problem of.It is right
The sequence putd question to accordingly is had in the proposition of these problems, for example, " you open xx business when customer service is putd question to", user
It answers " not having ", then, customer service can then ask that " not opening is", user answer "no", then client can inquire again " why
Opening xx business is", user answers "Yes", and so far, the problem of customer service also just finishes.For another example, when " you is putd question in customer service
Open xx business", user answers " open-minded ", then customer service can then ask that " you pay correlative charges", user
It answers " not having ", so far, the problem of customer service also just finishes.
Optionally, for the process by machine learning, learn to may include the problem of the proposition for the inquiry of user：
From dialogue language material sample extraction customer service to user proposes the problem of, by text cluster, opposite user proposition
Problem be classified and stored in described problem database.Such as：Synonymous question sentence is grouped into same category, per the forming of category
For a problem set of the inquiry of certain business of user in issue database.Wherein, how to extract customer service inquiry user's
Question sentence, how the specific implementation for being handled and being classified by text cluster mode can be realized by existing various ways, tool
It is readily apparent that on the basis of the technical solution that body is realized and those skilled in the art provide in the application, is not used to limit this
The protection domain of application, this point it is emphasized that：Using inquiry of the language material specimen needle to user is talked with, such as to certain business or certain
The inquiry used obtains a series of problems of the corresponding inquiry.
The problems in issue database can be a kind of problem for different business or be directed to same class business
A kind of problem of different aspect, etc., as long as one kind can be known as.
Optionally, response unit can include：Intention Anticipation module is intended to decision-making module；Wherein,
Intention Anticipation module, it is corresponding with the inquiry in described problem database for being predicted based on the inquiry of the user
Each question answering positive or negative probability, formed represent user view characterization information；Wherein, the characterization of user view
Every information of information corresponds to the probability of a problem and the positive or negative for the problem.
Be intended to decision-making module, for based on characterization information in a manner of more wheel question and answer, using user to it is described to
The answer for the problem of family proposes updates the probability that the user predicted in the characterization information answers (or negative) certainly；And
For determining the user view according to the characterization information obtained after more wheel question and answer.It is being intended to inquiry of the prediction module based on user
The probability of the answer affirmative of each problem in issue database is obtained, after obtaining the characterization information of user view, is asked in more wheels
During answering, user is obtained for answer the problem of proposition based on every wheel question and answer, can update characterization obtained above in real time
The probability of the answer affirmative (or negative) of correspondence problem in information, so as to determine user's based on updated characterization information
It is intended to.
Optionally, it is intended that decision-making module may include：Problem decision submodule and intention decision submodule, wherein,
Problem decision submodule, in the often wheel question and answer of more wheel question and answer, prediction to be obtained in the characterization information
The information gain of each problem, using information gain it is maximum the problem of as being proposed to user the problem of, and to user's proposition problem；
Be intended to decision submodule, for user to it is described to user proposes the problem of the answer update characterization information
The middle probability predicted obtained user and answer (or negative) certainly；Until determining that user anticipates according to updated characterization information
Figure.In the specific implementation, the number of more wheel question and answer can also be limited, i.e., when enquirement reaches preset enquirement number, even if can not be true
User view is made, can also stop being putd question to, and can immediate user view be determined based on characterization information, alternatively, stopping
It only the secondary inquiry and tells user's intention that can not determine user, user is asked to propose inquiry, etc. again.
It is intended to decision-making module in the dialog procedure for providing service to the user, it specifically can be according to user's inquiry and letter in reply
Breath proposes the problem of corresponding (Q1, Q2 ..., Qn) from issue database, and predicts being affirmatively answered or negating for each problem
Probability, to form the characterization information of vectorial (p1, p2 ..., pn) as user view.
If the inquiry for some user, knowing how answer of the user about problem each in issue database
(such as：For the matter of right and wrong, it is known that its answer is certainly or negates), then, just know clearly that this user needs to solve
Certainly what problem or to reach any target.That is, a series of problems in the problem of by extracting database and
Its corresponding answer constitutes a characterization information of user view.
When the inquiry for user proposes a problem, it is assumed that user provides an affirmative acknowledgement (ACK), according to this question and answer, meeting
It updates corresponding question and answer to be predicted as affirming, as shown in Fig. 2 (c)；Later, it is reintroduced according to the strategy of the enquirement in issue database
One problem, it is assumed that user provides a negative answer, according to this question and answer, can update corresponding question and answer and is predicted as and negate,
As shown in Fig. 2 (d).In this way, reach pre- until user view clearly such as reaches pre-set probability threshold value or puts question to enough
The enquirement frequency threshold value first set terminates to put question to.And ultimately form point for the i.e. user view of characterization information for representing user view
Cloth characterization vector (p1, p2 ..., pn).
It, can for be directed to Client-initiated inquiry without the information of any user view more specifically, during beginning of conversation
The all problems of energy, the distribution for priori that there are one answers, as shown in Fig. 2 (a).In user it is set forth for the problem of proposition
After problem or target (providing answer), the answer of each problem can be predicted by question and answer prediction model, forms user view
A distributed expression, as shown in Fig. 2 (b).It is expressed using the distribution of user view, may learn from dialogue language material
Pass between each problem (Q1, Q2 ..., Qn) and aims of systems T (can be an answer or the behavior for reaching user view)
Connection pt=F (p1, p2 ..., pn).Based on question and answer prediction distribution (p1, p2 ..., pn) can predict each problem Qi (i=1,
2 ... n) after affirmative acknowledgement (ACK) (i.e. pi=1) or negative answer (i.e. pi=0) is obtained, the information for target that may bring
Gain InfoGain (Qi) is such as shown in formula (1), i.e., after problem Qi is determined, to the prediction of target T from do not knowing to determining.
InfoGain (Qi)=Entropy (pt)-pi × Entropy (pt | pi=1)-(1-pi) × Entropy (pt | pi
=0)) (1)
When puing question to every time, the problem of taking predictive information gain maximum, puts question to user.When user asks according to one of proposition
After topic provides an answer, according to this question and answer, corresponding question and answer prediction distribution (p1, p2 ..., pn) is updated；After update
Distribution, predict that the information gain of each problem recalculates InfoGain (Qi) again, it is maximum to choose information gain again
The problem of continue to put question to user；Until user view is clear enough, such as the pt values predicted>Pre-set probability threshold value is such as
0.9, accurately answer or behavior can be made, terminates to put question to, alternatively, when the enquirement that the inquiry for user is initiated reaches pre-
The enquirement frequency threshold value first set, it is believed that user view is clear enough, terminates to put question to.
Wherein, the question and answer prediction model can be specifically to expect sample according to dialogue, be trained by machine learning
It obtains, based on the question and answer prediction model, using the inquiry of user as input, you can to obtain each problem in issue database
Answer the probability of positive or negative.
Optionally, it is intended that prediction module is specifically used for：
Based on the inquiry of the user, from dialogue language material learning to each problem and for representing to reach user view
Association between one answer or the aims of systems of behavior；It is predicted based on the question and answer distribution situation in association to described problem data
The probability of each question answering affirmative (or negative) corresponding with the inquiry, forms the characterization letter for representing user view in library
Breath.
Optionally, it is intended that the mapping relations for being previously stored with characterization information and user view are stored in decision submodule,
Being intended to decision-making module only needs according to the mapping relations, to find out the corresponding user view of characterization information currently formed.
The distributed characterization vector of user view, expresses all in dialog procedure and relevant information of user view, root
According to this corresponding user view of characterization vector, the solution finally selected for the user view is pushed to user.
The problem of above-mentioned database, the database that the set of the problem of being based on a kind of business is formed, alternatively, also may be used
The problem of the problem of to be the business based on multiple and different classifications, gathers the database formed, and every class business corresponds to, with industry
Classification of being engaged in is corresponding.If the in this way, database that the set of the problem of business based on multiple and different classifications is formed, then obtain
During the inquiry of one user, can class of service first be determined based on the inquiry of the user, and it is corresponding all to find out the class of service
Problem, and according to the characterization information of the answer composition expression user view to these problems.
Fig. 2 is the flow chart that the man-machine Interface of the application realizes interactive method, as shown in Fig. 2, including：
Step 200：In the dialog procedure for providing service to the user, the inquiry of user is obtained.
Step 201：Based on the problems in user's inquiry and issue database, proposed by the way of mostly wheel question and answer to user
Problem, and user view is determined according to the answer of user.
The problems in being inquired based on user in this step and issue database are carried by the way of mostly wheel question and answer to user
Go wrong including：
Each question answering corresponding with the inquiry is agreed in inquiry prediction described problem database based on the user
Fixed or negative probability, forms the characterization information for representing user view；
Based on characterization information in a manner of more wheel question and answer, using user to it is described to user propose the problem of answer
Update the probability that the user predicted in the characterization information answers (or negative) certainly；And according to being obtained after more wheel question and answer
Characterization information determine the user view.
Wherein,
Inquiry based on the user is predicted to each question answering corresponding with the inquiry in described problem database
Certainly the probability of (or negative) forms and represents that the characterization information of user view includes：
Based on the inquiry of the user, from dialogue language material learning to each problem and for representing to reach user view
Association between one answer or the aims of systems of behavior；It is predicted based on the question and answer distribution situation in association to described problem data
The probability of each question answering affirmative (or negative) corresponding with the inquiry, forms the characterization letter for representing user view in library
Breath.
If the inquiry for some user, knowing how answer of the user about problem each in issue database
(such as：For the matter of right and wrong, it is known that its answer is certainly or negates), then, just know clearly that this user needs to solve
Certainly what problem or to reach any target.That is, a series of problems in the problem of by extracting database and
Its corresponding answer constitutes a characterization information of user view.
When the inquiry for user proposes a problem, it is assumed that user provides an affirmative acknowledgement (ACK), according to this question and answer, meeting
It updates corresponding question and answer to be predicted as affirming, as shown in Fig. 2 (c)；Later, it is reintroduced according to the strategy of the enquirement in issue database
One problem, it is assumed that user provides a negative answer, according to this question and answer, can update corresponding question and answer and is predicted as and negate,
As shown in Fig. 2 (d).In this way, according to the strategy of enquirement until user view is clear enough or enquirement reaches pre-set and carries
It asks frequency threshold value, terminates to put question to.And ultimately form represent user view the characterization information i.e. distribution of user view characterize to
Amount (p1, p2 ..., pn).
More specifically,
Wherein,
Using user to it is described to user propose the problem of answer update the user predicted in the characterization information
The probability of (or negative) certainly is answered, determines that the user view includes according to the characterization information obtained after more wheel question and answer：
In the often wheel question and answer of more wheel question and answer, prediction obtains the information gain of each problem in the characterization information,
Using information gain it is maximum the problem of as next round in more wheel question and answer to user proposes the problem of, and asked to user's proposition
Topic；
With user to it is described to user propose the problem of answer update the user predicted in the characterization information and return
Answer the probability of (or negative) certainly；Until determining user view according to updated characterization information.
It, can for be directed to Client-initiated inquiry without the information of any user view more specifically, during beginning of conversation
The all problems of energy, the distribution for priori that there are one answers, as shown in Fig. 2 (a).In user it is set forth for the problem of proposition
After problem or target (providing answer), the answer of each problem can be predicted by question and answer prediction model, forms user view
A distributed expression, as shown in Fig. 2 (b).It is expressed using the distribution of user view, may learn from dialogue language material
Pass between each problem (Q1, Q2 ..., Qn) and aims of systems T (can be an answer or the behavior for reaching user view)
Connection pt=F (p1, p2 ..., pn).Based on question and answer prediction distribution (p1, p2 ..., pn) can predict each problem Qi (i=1,
2 ... n) after affirmative acknowledgement (ACK) (i.e. pi=1) or negative answer (i.e. pi=0) is obtained, the information for target that may bring
Gain InfoGain (Qi) is such as shown in formula (1), i.e., after problem Qi is determined, to the prediction of target T from do not knowing to determining.
When puing question to every time, the problem of taking predictive information gain maximum, puts question to user.When user asks according to one of proposition
After topic provides an answer, according to this question and answer, corresponding question and answer prediction distribution (p1, p2 ..., pn) is updated；After update
Distribution, predict that the information gain of each problem recalculates InfoGain (Qi) again, it is maximum to choose information gain again
The problem of continue to put question to user；Until user view is clear enough, such as the pt values predicted>Pre-set probability threshold value is such as
0.9, accurately answer or behavior can be made, terminates to put question to, alternatively, when the enquirement that the inquiry for user is initiated reaches pre-
The enquirement frequency threshold value first set, it is believed that user view is clear enough, terminates to put question to.
More specifically,
Wherein, according to the characterization information of the expression user view of formation, determine that the user view includes：
According to pre-set characterization information and the mapping relations of user view, the corresponding characterization currently formed is found out
The user view of information.The distributed characterization vector of user view, it is relevant to express all in dialog procedure and user view
Information, according to this corresponding user view of characterization vector, the solution finally selected for the user view is pushed to user.
The interactive technical solution of realization that the application provides in working method, in uncertain user view, is led to
It crosses and actively clarifies user view to user's enquirement, intelligently realize human-computer dialogue process, improve and user view is identified
Accuracy rate.
The application method further includes：
The problem of dialogue is determined based on talking with language material sample, for example learn the proposition of the inquiry for user
The problem of, and be stored in described problem database.
Wherein, the problems in issue database includes but not limited to：Outside the matter of right and wrong, the problem of also asking other types, such as
It refers in particular to ask, selection is asked.
Optionally, for the process by manually learning, learn to include the problem of the proposition for the inquiry of user：It is false
If user inquires：" how the xx business that may I ask me has more than is needed", by manually learning, the problem of customer service may propose, has：
" you open xx business", " not opening is" " you pay correlative charges", " how to open xx business is
" etc., these problems can be grouped into same category, form a series of propositions for inquiry xx business in issue database
The problem of.The sequence putd question to accordingly is had for the proposition of these problems, for example, " you open xx business when customer service is putd question to
", user answers " not having ", then, customer service can then ask that " not opening is", user answers "no", then client
It can inquire again that " how to open xx business is", user answers "Yes", and so far, the problem of customer service also just finishes.For another example, work as visitor
" you open xx business for clothes enquirement", user answers " open-minded ", then customer service can then ask that " you pay correlative charges
", user answers " not having ", and so far, the problem of customer service also just finishes.
Optionally, for the process by machine learning, learn to include the problem of the proposition for the inquiry of user：
From dialogue language material sample extraction customer service to user proposes the problem of, by text cluster, opposite user proposition
Problem be classified and stored in described problem database.Such as：Synonymous question sentence is grouped into same category, per the forming of category
For a problem set of the inquiry of certain business of user in issue database.Wherein, how to extract customer service inquiry user's
Question sentence, how the specific implementation for being handled and being classified by text cluster mode can be realized by existing various ways, tool
It is readily apparent that on the basis of the technical solution that body is realized and those skilled in the art provide in the application, is not used to limit this
The protection domain of application, this point it is emphasized that：Using inquiry of the language material specimen needle to user is talked with, such as to certain business or certain
The inquiry used obtains a series of problems of the corresponding inquiry.
Further, in the evolution mode of human-computer dialogue device, the session log during human-computer dialogue is put into pair
Language material sample, to correct or the asked questions of accretion learning, the sustained improvement performance of human-computer dialogue device, so as to reduce
Human-computer dialogue device update cost, improves renewal speed.
Interactive method, which is described in detail, to be realized to the application with reference to a specific embodiment.Fig. 4 is this
The flow diagram of the embodiment of interactive method is realized in application, as shown in figure 4, including following sections：
First, single-wheel problem identification is carried out to the inquiry of user, the input of single-wheel interaction problems identification model is user
The description text of a word of inquiry, such as " my account is stolen ", output are that the business corresponding to the description is classified as " such as
What solution limit ".From the point of view of citing, it is defeated to be described as " my Alipay account logs in not up " with inquiry, that is, initial problem of user
For entering, it is 1000 traffic issues that class object is obtained after classification task is performed, then, it is believed that single problem identification of discussing is into
Work(, export the corresponding result solved the problems, such as；If classification results cannot be obtained, then it is assumed that single problem identification of discussing is lost
It loses, more wheel interaction flows can be entered, i.e., further putd question to user and obtain answer to assist in the meaning that user sends out inquiry
Figure；
Then, to need it is further proposed that the problem of predict, and according to user to the answer of forecasting problem come do into
One step identifies.The target of problem prediction is to select a problem, if this problem obtains affirmative reply to being categorized into some
Business helps maximum.Assuming that the problems in issue database number is N, class of service sum is K, it is assumed that Pi=the i-th (i=1~
N) a problem is the probability of affirmative acknowledgement (ACK), Tj=P (business is classified as j | P1P2…PN) it is the conditional probability (j that type of service is j
=1~K), according to formula (1), the information gain when problem i becomes affirmative acknowledgement (ACK) is：InfoGain (i)=Entropy
(T)–Pi×Entropy(T|Pi=1)-(1-Pi)×Entropy(T|Pi=0) the problem of), picking out is exactly i=
argmax(InfoGain(i)).Wherein it is possible to it is modeled using a multilayer neural network FFNN from the answer of N number of problem
The mapping of the distribution of type of service is distributed to, the training data of this model is also derived from sample data, such as each energization
It talks about as a training data, similar with establishing issue database before, can extract that customer service asks from telephonograph text asks
Topic and the answer of user, are converted to (P1P2…PN) vector, in addition the type of service taken on the telephone mark forms data-label
Training language material.
The interactive process of more wheel flows is exactly model prediction and proposes a problem, and user provides an answer；According to this
A question and answer update corresponding question and answer distribution (P1P2…PN).According to updated input, recalculate each question answering and change
Information gain afterwards chooses the problem of information gain is maximum, continues to put question to user, is taken turns until user view clearly i.e. enough more
Problem identification success.The LSTM networks of more wheel interactions can provide the class object higher than threshold values.More wheel problem identifications can lead to
More wheel interaction problems identification models of training are crossed to realize, interaction problems identification models of taking turns with whole number of sessions with user more
According to by taking user proposes problem such as " I spend do not open " as an example, AGENT is further putd question to：" you are sellers ", user
It replies：" yes " etc. classification task is performed for input, class object is identical with single-wheel interaction problems identification model and 1000
Traffic issues.
Wherein, in the problem of input of problem prediction model is whole conversation contents of user, and output is pre-defined library
Some problem.When the confidence point of problem identification result is not higher than the threshold value set, user description information amount will be considered that not
Foot, problem prediction model will be picked out from problem base requries the users, and collect user most helpful problem of classifying
Answer differentiated again with more wheel interaction problems identification models.
Single-wheel interaction problems identification model mentioned above, more wheel interaction problems identification models and problem prediction model can be with
It is such as deep neural network (DNN) model trained in advance.
The application also provides one kind and is used to implement interactive device, including at least memory and processor, wherein, it deposits
Following executable instruction is stored in reservoir：In the dialog procedure for providing service to the user, the inquiry of user is obtained；Based on use
The problems in family inquiry and issue database propose problem, and according to the answer of user by the way of mostly wheel question and answer to user
Determine user view.
Although the embodiment disclosed by the application is as above, the content only for ease of understanding the application and use
Embodiment is not limited to the application.Technical staff in any the application fields, is taken off not departing from the application
Under the premise of the spirit and scope of dew, any modification and variation, but the application can be carried out in the form and details of implementation
Scope of patent protection, still should be subject to the scope of the claims as defined in the appended claims.
Claims (15)
1. a kind of human-computer dialogue device, which is characterized in that including：The problem of acquiring unit, response unit and storage problem number
According to library；Wherein,
Acquiring unit, in the dialog procedure for providing service to the user, obtaining the inquiry of user；
Response unit for being based on the problems in user's inquiry and issue database, is carried by the way of mostly wheel question and answer to user
It goes wrong, and user view is determined according to the answer of user.
2. human-computer dialogue device according to claim 1, which is characterized in that further include：
Unit for determining the problem of dialogue based on talking with language material sample, and is stored in described problem database.
3. human-computer dialogue device according to claim 2, which is characterized in that the unit is specifically used for：
From the dialogue language material sample extraction customer service to user proposes the problem of, by text cluster, opposite user proposition
Problem be classified and stored in described problem database.
4. the human-computer dialogue device according to right 1, which is characterized in that the response unit includes：Intention Anticipation module, meaning
Figure decision-making module；Wherein,
Intention Anticipation module, it is corresponding with the inquiry every in described problem database for being predicted based on the inquiry of the user
The probability of a question answering certainly, forms the characterization information for representing user view；
It is intended to decision-making module, in a manner of more wheel question and answer, being carried based on characterization information using user to described to user
The answer for the problem of going out updates the probability that the user predicted in the characterization information answers affirmative；And for according to more wheels
The characterization information obtained after question and answer determines the user view.
5. human-computer dialogue device according to claim 4, which is characterized in that the intention decision-making module, including：Problem is determined
Plan submodule and intention decision submodule, wherein,
Problem decision submodule, in the often wheel question and answer of more wheel question and answer, prediction to obtain each in the characterization information
The information gain of problem, using information gain it is maximum the problem of as being proposed to user the problem of, and to user propose problem；
Be intended to decision submodule, for user to it is described to user proposes the problem of answer update in the characterization information in advance
The user measured answers the probability of affirmative；Until determining user view according to updated characterization information.
6. human-computer dialogue device according to claim 4, which is characterized in that the Intention Anticipation module is specifically used for：
Based on the inquiry of the user, from dialogue language material learning to each problem and for representing one that reaches user view
Association between answer or the aims of systems of behavior；It is predicted based on the question and answer distribution situation in association in described problem database
The probability of each question answering corresponding with the inquiry certainly, forms the characterization information for representing user view.
7. human-computer dialogue device according to claim 5, which is characterized in that be stored in the intention decision submodule pre-
The characterization information and the mapping relations of user view first set；
It is intended to decision submodule to be specifically used for：According to the mapping relations, the corresponding characterization information currently formed is found out
User view.
8. a kind of realize interactive method, which is characterized in that including：
In the dialog procedure for providing service to the user, the inquiry of user is obtained；
Based on user's inquiry and the problems in issue database, problem is proposed to user by the way of mostly wheel question and answer, and according to
The answer of user determines user view.
9. according to the method described in claim 8, it is characterized in that, this method further includes：Based on talking with language material sample really
Surely the problem of talking with, and be stored in described problem database.
10. according to the method described in claim 9, it is characterized in that, the problem of determining dialogue include：
From the dialogue language material sample extraction customer service to user proposes the problem of, by text cluster, opposite user proposition
Problem be classified and stored in described problem database.
11. according to the method described in right 8, which is characterized in that it is described based on the problems in user's inquiry and issue database, it adopts
It is proposed that problem includes to user with the mode of more wheel question and answer：
Each question answering affirmative corresponding with the inquiry in inquiry prediction described problem database based on the user
Probability forms the characterization information for representing user view；
Based on characterization information in a manner of more wheel question and answer, using user to it is described to user proposes the problem of answer update
The user predicted in the characterization information answers the probability of affirmative；It is and true according to the characterization information obtained after more wheel question and answer
The fixed user view.
12. according to the method for claim 11, which is characterized in that it is described using user to it is described to user propose the problem of
Answer update the probability that the user predicted in the characterization information answers affirmative, according to the characterizations obtained after more wheel question and answer
Information determines that the user view includes：
In the often wheel question and answer of more wheel question and answer, prediction obtains the information gain of each problem in the characterization information, will believe
The problem of breath gain is maximum as next round in more wheel question and answer to user proposes the problem of, and to user's proposition problem；
With user to it is described to user propose the problem of answer update the user predicted in the characterization information and answer and agree
Fixed probability；Until determining user view according to updated characterization information.
13. according to the method for claim 11, which is characterized in that the inquiry prediction described problem based on the user
The probability of each question answering corresponding with the inquiry certainly, forms the characterization information packet for representing user view in database
It includes：
Based on the inquiry of the user, from dialogue language material learning to each problem and for representing one that reaches user view
Association between answer or the aims of systems of behavior；It is predicted based on the question and answer distribution situation in association in described problem database
The probability of each question answering corresponding with the inquiry certainly, forms the characterization information for representing user view.
14. according to the method for claim 12, which is characterized in that according to the characterization information of the expression user view of formation,
Determine that the user view includes：
According to pre-set characterization information and the mapping relations of user view, the corresponding characterization information currently formed is found out
User view.
15. one kind is used to implement interactive device, including at least memory and processor, wherein,
Following executable instruction is stored in memory：In the dialog procedure for providing service to the user, the inquiry of user is obtained；
Based on the problems in user's inquiry and issue database, problem is proposed to user, and according to user by the way of mostly wheel question and answer
Answer determine user view.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111296453.XA CN113961690A (en) | 2016-12-05 | 2016-12-05 | Man-machine conversation device and method for realizing man-machine conversation |
| CN201611106354.XA CN108153780B (en) | 2016-12-05 | 2016-12-05 | Man-machine conversation device and method for realizing man-machine conversation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611106354.XA CN108153780B (en) | 2016-12-05 | 2016-12-05 | Man-machine conversation device and method for realizing man-machine conversation |
Related Child Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111296453.XA Division CN113961690A (en) | 2016-12-05 | 2016-12-05 | Man-machine conversation device and method for realizing man-machine conversation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108153780A true CN108153780A (en) | 2018-06-12 |
| CN108153780B CN108153780B (en) | 2021-11-23 |
Family
ID=62470879
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111296453.XA Pending CN113961690A (en) | 2016-12-05 | 2016-12-05 | Man-machine conversation device and method for realizing man-machine conversation |
| CN201611106354.XA Active CN108153780B (en) | 2016-12-05 | 2016-12-05 | Man-machine conversation device and method for realizing man-machine conversation |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111296453.XA Pending CN113961690A (en) | 2016-12-05 | 2016-12-05 | Man-machine conversation device and method for realizing man-machine conversation |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN113961690A (en) |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109116743A (en) * | 2018-08-01 | 2019-01-01 | 珠海格力电器股份有限公司 | A kind of intelligent appliance interactive approach and system |
| CN109241256A (en) * | 2018-08-20 | 2019-01-18 | 百度在线网络技术（北京）有限公司 | Dialog process method, apparatus, computer equipment and readable storage medium storing program for executing |
| CN109299231A (en) * | 2018-09-14 | 2019-02-01 | 苏州思必驰信息科技有限公司 | Dialogue state tracking, system, electronic equipment and storage medium |
| CN109460451A (en) * | 2018-09-29 | 2019-03-12 | 厦门快商通信息技术有限公司 | Actively obtain method, the intelligent customer service method and system of user information |
| CN109460459A (en) * | 2018-10-31 | 2019-03-12 | 神思电子技术股份有限公司 | A kind of conversational system automatic optimization method based on log study |
| CN109525480A (en) * | 2018-09-14 | 2019-03-26 | 广东神马搜索科技有限公司 | Customer problem collection system and method |
| CN109614464A (en) * | 2018-10-31 | 2019-04-12 | 阿里巴巴集团控股有限公司 | Method and device for traffic issues identification |
| CN109710941A (en) * | 2018-12-29 | 2019-05-03 | 上海点融信息科技有限责任公司 | User's intension recognizing method and device based on artificial intelligence |
| CN109918492A (en) * | 2019-03-18 | 2019-06-21 | 百度在线网络技术（北京）有限公司 | System is arranged in a kind of human-computer dialogue setting method and human-computer dialogue |
| CN109977209A (en) * | 2019-03-22 | 2019-07-05 | 深圳狗尾草智能科技有限公司 | More wheel man-machine interaction methods, system, computer and medium |
| CN110019688A (en) * | 2019-01-23 | 2019-07-16 | 艾肯特公司 | The method that robot is trained |
| CN110059174A (en) * | 2019-04-28 | 2019-07-26 | 科大讯飞股份有限公司 | Inquiry guidance method and device |
| CN110110048A (en) * | 2019-05-10 | 2019-08-09 | 科大讯飞股份有限公司 | Inquiry guidance method and device |
| CN110442690A (en) * | 2019-06-26 | 2019-11-12 | 重庆兆光科技股份有限公司 | A kind of query optimization method, system and medium based on probability inference |
| CN110688459A (en) * | 2019-09-29 | 2020-01-14 | 联想(北京)有限公司 | Intelligent interaction method and intelligent equipment |
| 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 |
| CN111241237A (en) * | 2019-12-31 | 2020-06-05 | 中国建设银行股份有限公司 | Intelligent question and answer data processing method and device based on operation and maintenance service |
| CN111552784A (en) * | 2019-02-12 | 2020-08-18 | 厦门邑通软件科技有限公司 | Man-machine conversation method based on ABC communication rule |
| CN111985248A (en) * | 2020-06-30 | 2020-11-24 | 联想(北京)有限公司 | Information interaction method and device |
| CN112232080A (en) * | 2020-10-20 | 2021-01-15 | 大唐融合通信股份有限公司 | Named entity identification method and device and electronic equipment |
| CN112673367A (en) * | 2018-10-31 | 2021-04-16 | 华为技术有限公司 | Electronic device and method for predicting user intention |
| CN112771506A (en) * | 2018-10-30 | 2021-05-07 | Je国际公司 | Conversation system, conversation robot server device, conversation robot ID management device, conversation mediation server device, program, conversation method, and conversation mediation method |
| CN112988991A (en) * | 2021-02-04 | 2021-06-18 | 支付宝(杭州)信息技术有限公司 | Method and system for anti-fraud intervention through man-machine conversation |
| CN113051375A (en) * | 2019-12-27 | 2021-06-29 | 阿里巴巴集团控股有限公司 | Question-answering data processing method and device based on question-answering equipment |
| CN113139045A (en) * | 2021-05-13 | 2021-07-20 | 八维(杭州)科技有限公司 | Selective question-answering method based on task driving type man-machine conversation |
| CN113837638A (en) * | 2021-09-29 | 2021-12-24 | 支付宝(杭州)信息技术有限公司 | Method, device and equipment for determining dialect |
| WO2022105119A1 (en) * | 2020-11-17 | 2022-05-27 | 平安科技（深圳）有限公司 | Training corpus generation method for intention recognition model, and related device thereof |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20020072092A (en) * | 2001-03-08 | 2002-09-14 | 서정연 | Real-time Natural Language Question-Answering System Using Unit Paragraph Indexing Method |
| CN102456018A (en) * | 2010-10-18 | 2012-05-16 | 腾讯科技（深圳）有限公司 | Interactive search method and device |
| CN102722558A (en) * | 2012-05-29 | 2012-10-10 | 百度在线网络技术（北京）有限公司 | User question recommending method and device |
| JP2013143066A (en) * | 2012-01-12 | 2013-07-22 | Kddi Corp | Question and answer program, server and method which use large amount of comment texts |
| CN105468648A (en) * | 2014-09-11 | 2016-04-06 | 北大方正集团有限公司 | Method and apparatus for establishing specialized knowledge question/answer system |
-
2016
- 2016-12-05 CN CN202111296453.XA patent/CN113961690A/en active Pending
- 2016-12-05 CN CN201611106354.XA patent/CN108153780B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20020072092A (en) * | 2001-03-08 | 2002-09-14 | 서정연 | Real-time Natural Language Question-Answering System Using Unit Paragraph Indexing Method |
| CN102456018A (en) * | 2010-10-18 | 2012-05-16 | 腾讯科技（深圳）有限公司 | Interactive search method and device |
| JP2013143066A (en) * | 2012-01-12 | 2013-07-22 | Kddi Corp | Question and answer program, server and method which use large amount of comment texts |
| CN102722558A (en) * | 2012-05-29 | 2012-10-10 | 百度在线网络技术（北京）有限公司 | User question recommending method and device |
| CN105468648A (en) * | 2014-09-11 | 2016-04-06 | 北大方正集团有限公司 | Method and apparatus for establishing specialized knowledge question/answer system |
Cited By (37)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109116743A (en) * | 2018-08-01 | 2019-01-01 | 珠海格力电器股份有限公司 | A kind of intelligent appliance interactive approach and system |
| CN109241256A (en) * | 2018-08-20 | 2019-01-18 | 百度在线网络技术（北京）有限公司 | Dialog process method, apparatus, computer equipment and readable storage medium storing program for executing |
| CN109525480B (en) * | 2018-09-14 | 2021-09-10 | 阿里巴巴（中国）有限公司 | User question collection system and method |
| CN109299231A (en) * | 2018-09-14 | 2019-02-01 | 苏州思必驰信息科技有限公司 | Dialogue state tracking, system, electronic equipment and storage medium |
| CN109299231B (en) * | 2018-09-14 | 2020-10-30 | 苏州思必驰信息科技有限公司 | Dialog state tracking method, system, electronic device and storage medium |
| CN109525480A (en) * | 2018-09-14 | 2019-03-26 | 广东神马搜索科技有限公司 | Customer problem collection system and method |
| CN109460451A (en) * | 2018-09-29 | 2019-03-12 | 厦门快商通信息技术有限公司 | Actively obtain method, the intelligent customer service method and system of user information |
| CN112771506A (en) * | 2018-10-30 | 2021-05-07 | Je国际公司 | Conversation system, conversation robot server device, conversation robot ID management device, conversation mediation server device, program, conversation method, and conversation mediation method |
| CN109614464B (en) * | 2018-10-31 | 2023-10-27 | 创新先进技术有限公司 | Method and device for identifying business problems |
| US11874876B2 (en) | 2018-10-31 | 2024-01-16 | Huawei Technologies Co., Ltd. | Electronic device and method for predicting an intention of a user |
| CN112673367A (en) * | 2018-10-31 | 2021-04-16 | 华为技术有限公司 | Electronic device and method for predicting user intention |
| CN109460459A (en) * | 2018-10-31 | 2019-03-12 | 神思电子技术股份有限公司 | A kind of conversational system automatic optimization method based on log study |
| CN109614464A (en) * | 2018-10-31 | 2019-04-12 | 阿里巴巴集团控股有限公司 | Method and device for traffic issues identification |
| CN109710941A (en) * | 2018-12-29 | 2019-05-03 | 上海点融信息科技有限责任公司 | User's intension recognizing method and device based on artificial intelligence |
| CN110019688A (en) * | 2019-01-23 | 2019-07-16 | 艾肯特公司 | The method that robot is trained |
| CN111552784A (en) * | 2019-02-12 | 2020-08-18 | 厦门邑通软件科技有限公司 | Man-machine conversation method based on ABC communication rule |
| CN109918492A (en) * | 2019-03-18 | 2019-06-21 | 百度在线网络技术（北京）有限公司 | System is arranged in a kind of human-computer dialogue setting method and human-computer dialogue |
| CN109977209A (en) * | 2019-03-22 | 2019-07-05 | 深圳狗尾草智能科技有限公司 | More wheel man-machine interaction methods, system, computer and medium |
| CN110059174A (en) * | 2019-04-28 | 2019-07-26 | 科大讯飞股份有限公司 | Inquiry guidance method and device |
| CN110059174B (en) * | 2019-04-28 | 2023-05-30 | 科大讯飞股份有限公司 | Query guiding method and device |
| CN110110048A (en) * | 2019-05-10 | 2019-08-09 | 科大讯飞股份有限公司 | Inquiry guidance method and device |
| CN110442690B (en) * | 2019-06-26 | 2021-08-17 | 重庆兆光科技股份有限公司 | Query optimization method, system and medium based on probabilistic reasoning |
| CN110442690A (en) * | 2019-06-26 | 2019-11-12 | 重庆兆光科技股份有限公司 | A kind of query optimization method, system and medium based on probability inference |
| CN110688459A (en) * | 2019-09-29 | 2020-01-14 | 联想(北京)有限公司 | Intelligent interaction method and intelligent equipment |
| 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 |
| CN113051375A (en) * | 2019-12-27 | 2021-06-29 | 阿里巴巴集团控股有限公司 | Question-answering data processing method and device based on question-answering equipment |
| CN111241237A (en) * | 2019-12-31 | 2020-06-05 | 中国建设银行股份有限公司 | Intelligent question and answer data processing method and device based on operation and maintenance service |
| CN111241237B (en) * | 2019-12-31 | 2023-05-23 | 中国建设银行股份有限公司 | Intelligent question-answer data processing method and device based on operation and maintenance service |
| CN111985248A (en) * | 2020-06-30 | 2020-11-24 | 联想(北京)有限公司 | Information interaction method and device |
| CN112232080A (en) * | 2020-10-20 | 2021-01-15 | 大唐融合通信股份有限公司 | Named entity identification method and device and electronic equipment |
| WO2022105119A1 (en) * | 2020-11-17 | 2022-05-27 | 平安科技（深圳）有限公司 | Training corpus generation method for intention recognition model, and related device thereof |
| CN112988991B (en) * | 2021-02-04 | 2023-04-18 | 支付宝(杭州)信息技术有限公司 | Method and system for performing anti-fraud intervention through man-machine conversation |
| CN112988991A (en) * | 2021-02-04 | 2021-06-18 | 支付宝(杭州)信息技术有限公司 | Method and system for anti-fraud intervention through man-machine conversation |
| CN113139045B (en) * | 2021-05-13 | 2023-05-05 | 八维(杭州)科技有限公司 | Selective question-answering method based on task-driven man-machine dialogue |
| CN113139045A (en) * | 2021-05-13 | 2021-07-20 | 八维(杭州)科技有限公司 | Selective question-answering method based on task driving type man-machine conversation |
| CN113837638A (en) * | 2021-09-29 | 2021-12-24 | 支付宝(杭州)信息技术有限公司 | Method, device and equipment for determining dialect |
| CN113837638B (en) * | 2021-09-29 | 2024-04-26 | 支付宝(杭州)信息技术有限公司 | Method, device and equipment for determining speaking skill |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108153780B (en) | 2021-11-23 |
| CN113961690A (en) | 2022-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108153780A (en) | A kind of human-computer dialogue device and its interactive method of realization | |
| CN107329967B (en) | Question answering system and method based on deep learning | |
| CN109493166B (en) | Construction method for task type dialogue system aiming at e-commerce shopping guide scene | |
| CN111046132B (en) | Customer service question-answering processing method and system for searching multiple rounds of conversations | |
| CN109284363A (en) | A kind of answering method, device, electronic equipment and storage medium | |
| Shah et al. | Interactive reinforcement learning for task-oriented dialogue management | |
| CN107589826A (en) | The man-machine interaction method and system of knowledge based collection of illustrative plates | |
| CN109446306A (en) | A kind of intelligent answer method of more wheels dialogue of task based access control driving | |
| CN109829036A (en) | A kind of dialogue management method and relevant apparatus | |
| CN107071193B (en) | Method and device for accessing interactive response system to user | |
| CN106951468A (en) | Talk with generation method and device | |
| CN106845624A (en) | The multi-modal exchange method relevant with the application program of intelligent robot and system | |
| CN113268610B (en) | Intent jump method, device, equipment and storage medium based on knowledge graph | |
| CN109145168A (en) | A kind of expert service robot cloud platform | |
| CN110532363B (en) | Task-oriented automatic dialogue method based on decision tree | |
| US20190220762A1 (en) | Probabilistic modeling system and method | |
| CN109993543A (en) | A kind of complaint handling method and system | |
| CN106164896A (en) | For finding multidimensional recursive learning process and the system of complicated two analogues or many analogues relation | |
| CN110162606A (en) | For solving the session proxy learning model services selection of client-side service request | |
| CN112925888A (en) | Method and device for training question-answer response and small sample text matching model | |
| CN108491480A (en) | Rumour detection method and equipment | |
| Ferreira et al. | Expert-based reward shaping and exploration scheme for boosting policy learning of dialogue management | |
| Misu et al. | Modeling spoken decision making dialogue and optimization of its dialogue strategy | |
| CN110533054B (en) | Multi-mode self-adaptive machine learning method and device | |
| CN108363738B (en) | Recommendation method for industrial equipment data analysis algorithm |
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 | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1256832 Country of ref document: HK |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |