CN109739969A - Answer generation method and intelligent conversational system - Google Patents
Answer generation method and intelligent conversational system Download PDFInfo
- Publication number
- CN109739969A CN109739969A CN201811653405.XA CN201811653405A CN109739969A CN 109739969 A CN109739969 A CN 109739969A CN 201811653405 A CN201811653405 A CN 201811653405A CN 109739969 A CN109739969 A CN 109739969A
- Authority
- CN
- China
- Prior art keywords
- data
- answer
- read statement
- information
- reply
- 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.)
- Pending
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Present disclose provides a kind of answer generation methods, are applied in intelligent conversational system, and the intelligence conversational system can be responded and be provided feedback information to the input information received, this method comprises: obtaining read statement data;Identification characterizes the focus information of the read statement data;Obtain the first reply data in response to the read statement data;Obtain the second reply data in response to the focus information;Third reply data is generated based on first reply data and second reply data;And the output third reply data.The disclosure additionally provides a kind of intelligent conversational system.
Description
Technical field
This disclosure relates to a kind of answer generation method and intelligent conversational system.
Background technique
Question answering system is with a wide range of applications in fields such as electric business customer services, and easily scale to information consumption other
Related fields.User inputs question demand in question answering system, it is desirable that it exports the reasonable correct of the question demand for being directed to user
Answer.
In existing question answering system, answer usually is determined according only to the intent information of question sentence, is had the following problems: question answering system
For the same answer for being intended to have fixed, but since user's question formulation is flexible and changeable, can be had based on same intention
A variety of question formulations, in this case, although intention assessment is correct, answer cannot comply fully with the expectation of user's enquirement.
Summary of the invention
An aspect of this disclosure provides a kind of answer generation method, is applied in intelligent conversational system, the intelligence meeting
Telephone system can respond the input information received and provide feedback information.The above method includes: acquisition read statement
Data, identification characterize the focus information of the read statement data, obtain the first response in response to the read statement data
Data are obtained the second reply data in response to the focus information, are answered based on first reply data and described second
Answer is according to generation third reply data and exports.
Optionally, above-mentioned focus information is for characterizing the informational content that user most pays close attention in read statement data.
Optionally, it includes: building focus information rule base that above-mentioned identification, which characterizes the focus information of the read statement data,;
Believe from the one or more segments to match with the rule in the focus information rule base are extracted in the read statement data
Breath, when extracting a piece segment information, using the piece segment information as the focus information of the characterization read statement data, with
And when extracting multiple segment informations, highest segment information of confidence level is chosen as the characterization read statement data
Focus information.
Optionally, above-mentioned acquisition in response to the first reply data of the read statement data include: identification with it is described defeated
Enter the corresponding intent information of phrase data, and, corresponding original answer data is obtained based on the intent information, by the original
Beginning answer data is as first reply data.
Optionally, before above-mentioned acquisition is in response to the second reply data of the focus information, the above method further include:
Identify question sentence type belonging to the read statement data.Second reply data packet of the above-mentioned acquisition in response to the focus information
It includes: obtaining the original art rule that should answer corresponding with the question sentence type, the original art rule characterization that should answer is asked with described
The answer form of sentence type adaptation;Focus amendment is carried out to the original art rule that should answer using the focus information, is obtained
The art data that should answer corresponding with the read statement data should answer art data as second reply data by described in.
Optionally, question sentence type belonging to the above-mentioned identification read statement data includes: trained Question Classification model, benefit
Question sentence type belonging to the read statement data described in the Question Classification model prediction;And/or establish Question Classification rule
Then library, according to the classifying rules in the Question Classification rule base determine the read statement data belonging to question sentence type.
Optionally, above-mentioned that third reply data packet is generated based on first reply data and second reply data
It includes: based on preset fusion rule, art data should be answered and the original answer data merges to described, obtain the third
Reply data.
Optionally, it is above-mentioned to it is described should answer art data and the original answer data to carry out fusion include: to obtain institute
After stating original answer data, the art data that should answer are carried out making processing clear according to the original answer data;It will be through
It crosses and makes the described of processing clear and should answer art data and the original answer data group is combined into third reply data.
Optionally, the above method further include: in the third answer number corresponding with read statement data for obtaining predetermined quantity
According to later, model training is carried out based on the read statement data and the third reply data, wherein with the read statement
The corresponding original answer of the intent information of data, question sentence type and focus information as training data, with the read statement
Label of the corresponding third reply data of data as the training data, training obtain question sentence answer correction model.And/or
Person pre-processes the read statement data before the focus information that identification characterizes the read statement data.
Another aspect of the present disclosure provides a kind of answer generating means, is applied in intelligent conversational system, the intelligence
Conversational system can be responded and be provided feedback information to the input information received, the device include: the first acquisition module,
First identification module, second obtain module, third acquisition module, integrated treatment module and output module.First, which obtains module, uses
In acquisition read statement data.First identification module characterizes the focus information of the read statement data for identification.Second obtains
Modulus block is used to obtain the first reply data in response to the read statement data.Third obtain module for obtain in response to
Second reply data of the focus information.Integrated treatment module is used to be based on first reply data and second response
Data generate third reply data.Output module is for exporting the third reply data.
Optionally, above-mentioned focus information is for characterizing the informational content that user most pays close attention in the read statement data.
Optionally, the first identification module includes: building submodule, extracting sub-module and the first determining submodule.Building
Module is for constructing focus information rule base.Extracting sub-module is used to extract from the read statement data to be believed with the focus
One or more piece segment informations that rule in breath rule base matches.First, which determines that submodule is used to work as, extracts a segment
When information, using the piece segment information as the focus information for characterizing the read statement data;When extracting multiple segment informations,
Highest segment information of confidence level is chosen as the focus information for characterizing the read statement data.
Optionally, the second acquisition module includes: that identification submodule and second determine submodule.Identify submodule for identification
Intent information corresponding with the read statement data.Second determines that submodule is used to obtain accordingly based on the intent information
Original answer data, using the original answer data as first reply data.
Optionally, above-mentioned apparatus further includes the second identification module, is obtained for obtaining module in third in response to the coke
Before second reply data of point information, question sentence type belonging to the read statement data is identified.Third obtains module
Acquisition submodule and amendment submodule.Acquisition submodule is for obtaining the original art rule that should answer corresponding with the question sentence type
Then, the answer form of the original art rule characterization and the question sentence type adaptation that should answer.Submodule is corrected to be used to utilize institute
It states focus information and focus amendment is carried out to the original art rule that should answer, obtain response corresponding with the read statement data
Art data are talked about, art data should be answered as second reply data by described in.
Optionally, the second identification module is specifically used for training Question Classification model, utilizes the Question Classification model prediction
Question sentence type belonging to the read statement data;And/or Question Classification rule base is established, according to the Question Classification
Classifying rules in rule base determines question sentence type belonging to the read statement data.
Optionally, integrated treatment module is specifically used for being based on preset fusion rule, to the art data and described of should answering
Original answer data is merged, and the third reply data is obtained.
Optionally, integrated treatment module includes processing submodule and fusion submodule.Submodule is handled to be used to obtain institute
After stating original answer data, the art data that should answer are carried out making processing clear according to the original answer data.Fusion
Submodule is used to that described art data will should to be answered and the original answer data group is combined into third response by make processing clear
Data.
Optionally, above-mentioned apparatus further includes model training module, for obtain predetermined quantity with read statement data
After corresponding third reply data, model training is carried out based on the read statement data and the third reply data,
In, using the corresponding original answer of the intent information of the read statement data, question sentence type and focus information as training data,
Using third reply data corresponding with the read statement data as the label of the training data, training obtains question sentence answer
Correction model.And/or above-mentioned apparatus further includes preprocessing module, for described defeated in the first identification module identification characterization
Before the focus information for entering phrase data, the read statement data are pre-processed.
Another aspect of the present disclosure provides a kind of intelligent conversational system, and the intelligence conversational system can be to receiving
Input information is responded and is provided feedback information, and the intelligence conversational system includes memory, processor and is stored in storage
On device and the computer program that can run on a processor, for realizing as described above when the processor executes described program
Method.
Another aspect of the present disclosure provides a kind of computer readable storage medium, is stored with computer executable instructions,
Described instruction is when executed for realizing method as described above.
Another aspect of the present disclosure provides a kind of computer program, and the computer program, which includes that computer is executable, to be referred to
It enables, described instruction is when executed for realizing method as described above.
Detailed description of the invention
In order to which the disclosure and its advantage is more fully understood, referring now to being described below in conjunction with attached drawing, in which:
Fig. 1 diagrammatically illustrates the applied field of answer generation method according to an embodiment of the present disclosure and intelligent conversational system
Scape;
Fig. 2 diagrammatically illustrates the flow chart of answer generation method according to an embodiment of the present disclosure;
Fig. 3 diagrammatically illustrates the schematic diagram of answer generating process according to an embodiment of the present disclosure;
Fig. 4 diagrammatically illustrates the block diagram of answer generating means according to an embodiment of the present disclosure;
Fig. 5 diagrammatically illustrates the block diagram of answer generating means according to another embodiment of the present disclosure;And
Fig. 6 diagrammatically illustrates the block diagram of intelligent conversational system according to an embodiment of the present disclosure.
Specific embodiment
Hereinafter, will be described with reference to the accompanying drawings embodiment of the disclosure.However, it should be understood that these descriptions are only exemplary
, and it is not intended to limit the scope of the present disclosure.In the following detailed description, to elaborate many specific thin convenient for explaining
Section is to provide the comprehensive understanding to the embodiment of the present disclosure.It may be evident, however, that one or more embodiments are not having these specific thin
It can also be carried out in the case where section.In addition, in the following description, descriptions of well-known structures and technologies are omitted, to avoid
Unnecessarily obscure the concept of the disclosure.
Term as used herein is not intended to limit the disclosure just for the sake of description specific embodiment.It uses herein
The terms "include", "comprise" etc. show the presence of the feature, step, operation and/or component, but it is not excluded that in the presence of
Or add other one or more features, step, operation or component.
There are all terms (including technical and scientific term) as used herein those skilled in the art to be generally understood
Meaning, unless otherwise defined.It should be noted that term used herein should be interpreted that with consistent with the context of this specification
Meaning, without that should be explained with idealization or excessively mechanical mode.
It, in general should be according to this using statement as " at least one in A, B and C etc. " is similar to
Field technical staff is generally understood the meaning of the statement to make an explanation (for example, " system at least one in A, B and C "
Should include but is not limited to individually with A, individually with B, individually with C, with A and B, with A and C, have B and C, and/or
System etc. with A, B, C).Using statement as " at least one in A, B or C etc. " is similar to, generally come
Saying be generally understood the meaning of the statement according to those skilled in the art to make an explanation (for example, " having in A, B or C at least
One system " should include but is not limited to individually with A, individually with B, individually with C, with A and B, have A and C, have
B and C, and/or the system with A, B, C etc.).
Shown in the drawings of some block diagrams and/or flow chart.It should be understood that some sides in block diagram and/or flow chart
Frame or combinations thereof can be realized by computer program instructions.These computer program instructions can be supplied to general purpose computer,
The processor of special purpose computer or other programmable data processing units, so that these instructions are when executed by this processor can be with
Creation is for realizing function/operation device illustrated in these block diagrams and/or flow chart.The technology of the disclosure can be hard
The form of part and/or software (including firmware, microcode etc.) is realized.In addition, the technology of the disclosure, which can be taken, is stored with finger
The form of computer program product on the computer readable storage medium of order, the computer program product is for instruction execution system
System uses or instruction execution system is combined to use.
Embodiment of the disclosure provides a kind of answer generation method and can apply the intelligent conversational system of this method.
This method includes input acquisition stage, identifying processing stage and answer output stage.It is defeated to input acquisition stage acquisition characterization user
The read statement data of the sentence entered, subsequently into the identifying processing stage.In the identifying processing stage, identification characterizes the input language
The focus information of sentence data obtains the first reply data in response to read statement data and second in response to focus information and answers
Answer evidence, then third reply data is generated based on the first reply data and the second reply data.Answer output stage output the
Three reply datas.
Fig. 1 diagrammatically illustrates the applied field of answer generation method according to an embodiment of the present disclosure and intelligent conversational system
Scape.It should be noted that being only the example that can apply the scene of the embodiment of the present disclosure shown in Fig. 1, to help art technology
Personnel understand the technology contents of the disclosure, but are not meant to that the embodiment of the present disclosure may not be usable for other equipment, system, environment
Or scene.
As shown in Figure 1, the application scenarios may include terminal device 101, network 102 and server/server cluster
103.Network 102 between terminal device 101 and server/server cluster 103 to provide the medium of communication link.Network
102 may include various connection types, such as wired, wireless communication link or fiber optic cables etc..
User can be used terminal device 101 and be interacted by network 102 with server/server cluster 103, be asked with input
It inscribes and receives answer.Terminal device 101 can be the various electronic equipments with input/output function, including but not limited to intelligently
Mobile phone, tablet computer, pocket computer on knee and desktop computer etc..
Server/server cluster 103 can be to provide the server or server cluster of various services, back-stage management clothes
Business device or server cluster can carry out analyzing etc. to data such as the user's input problems received processing, and by corresponding answer
Feed back to terminal device.
It should be noted that answer generation method provided by the embodiment of the present disclosure generally can be by server/server
Cluster 103 executes.Correspondingly, answer generating means provided by the embodiment of the present disclosure generally can be set in servers/services
In device cluster 103.Answer generation method provided by the embodiment of the present disclosure can also be by being different from server/server cluster
103 and the server or server cluster that can be communicated with terminal device 101, and/or server/server cluster 103 hold
Row.Correspondingly, answer generating means provided by the embodiment of the present disclosure also can be set in different from server/server cluster
103 and the server or server cluster that can be communicated with terminal device 101, and/or server/server cluster 103 in.
It should be understood that the terminal device, network and server/server cluster number in Fig. 1 are only schematical.
According to needs are realized, any number of terminal device, network and server/server cluster can have.
Fig. 2 diagrammatically illustrates the flow chart of answer generation method according to an embodiment of the present disclosure.This method is applied to
In intelligent conversational system, which can respond the input information received and provide feedback information.
As shown in Fig. 2, this method includes operation S201~S206.
In operation S201, read statement data are obtained.
Wherein, the read statement data are used to characterize the information of user's input.
In operation S202, identification characterizes the focus information of the read statement data.
Wherein, focus (Focus) information for characterizing read statement data embodies the focus of read statement data.
In operation S203, the first reply data in response to the read statement data is obtained.
In operation S204, the second reply data in response to the focus information is obtained.
In operation S205, third reply data is generated based on first reply data and second reply data.
In operation S206, the third reply data is exported.
As it can be seen that method shown in Fig. 2 after getting read statement data, is on the one hand obtained in response to the read statement data
The first reply data is obtained, on the other hand obtains the second reply data in response to characterizing the focus information of the read statement data, then
Third reply data is generated based on the first reply data and the second reply data.Since focus information embodies read statement data
Focus, with read statement data to embody form related, can obtain adapting to read statement in conjunction with focus data
It is different but corresponding to same constructed answers to avoid the occurrence of read statement data for the third reply data of data focused particularly on a little
Problem carries out direct response to the specific input of user as much as possible, realizes more intelligent conversation procedure, meets user's need
It asks.
In embodiment of the disclosure, focus information be used to characterize user most pays close attention in read statement data information at
Point, be meaning informational content most outstanding in read statement data, be send the read statement data sender wish to receive
The part of square great care, and the prompt and reference of response that user is gone for.It usually can use various modes
Label and expression focus information.For example, Chinese can mark focus information with "Yes", it is " small to compare two read statement data
King is to go on business tomorrow " and " being that Xiao Wang will go on business tomorrow ", although the purpose for issuing the sender of the two read statement data is identical,
It is to inform that Xiao Wang will go on business this event tomorrow, but the former focus information is " tomorrow ", it is emphasised that the time, and the coke of the latter
Point information is " Xiao Wang ", it is emphasised that personage.As can be seen that the minor change of read statement data may make that focus information is complete
Difference, the focus information of answer generation method combination read statement data shown in Fig. 2 obtain for read statement data
Three reply datas so that third reply data more fine granularity be a little adapted with focusing particularly on for read statement data, for
Personalized under identical intention of family put question to generate height meet it is user's concern, meeting user's term habit, more smooth,
Natural response.
Method shown in Fig. 2 is compared with the prior art, prior art answer generated cannot comply fully with use
The expectation that family is putd question to, such as: read statement data are the problem of user inputs.The problem of user inputs 1 are as follows: " Can I
Cancel my order? " (" I can cancel an order? "), user input the problem of 2 are as follows: " how to cancel my
Order? " (" how cancelling an order? ").
The intent information of problem 1 and problem 2 is " cancel order " (" cancelling an order "), and the prior art is based on the meaning
Figure information selects constructed answers: " You may be able to cancel the order if you contact us at
XXX-XXX-XXX Within an hour after placing it. " is (" if you dial XXX- within the latter hour that places an order
XXX-XXX contacts us, you can cancel an order.").As can be seen that the intent information of problem 1 and problem 2 is although identical, but
The focus of user is different, and the prior art answer without the positive enquirement for directly replying user with same answer, needs to use
Family makes inferences the answer that can just obtain oneself, and user experience is bad.And the scheme according to the embodiment of the present disclosure, believe in conjunction with focus
Breath generates third reply data, since problem 1 is different with the focus information of problem 2, so that being directed to the third reply data of problem 1
It is different with for the third reply data of problem 2.
In one embodiment of the present disclosure, operation S202 identification characterizes the focus information packet of the read statement data
Include: building focus information rule base matches from extracting in read statement data with the rule in the focus information rule base
One or more piece segment informations, when extracting a piece segment information, using the piece segment information as characterizing the read statement
It is described defeated as characterizing to choose highest segment information of confidence level when extracting multiple segment informations for the focus information of data
Enter the focus information of phrase data.
Wherein it is possible to construct focus information rule based on the marking convention of focus information in linguistics and expression way
Library.Based on the focus information rule base, when read statement data correspond to interrogative sentence, extracted from the read statement data with
Corresponding segment information of phrase with interrogative is as focus information, when read statement data correspond to declarative sentence, from this
It is extracted in read statement data with corresponding noun, verb, nominal phrase, and/or verb character phrase segment information as burnt
Point information.When extracting multiple segment informations, using the multiple segment informations extracted as multiple candidate focus informations, calculate
The confidence level of multiple candidate's focus informations is simultaneously ranked up, and chooses the highest candidate focus information of confidence level as characterization input language
The focus information of sentence data.
Specifically, it calculates the confidence level of candidate focus information and the algorithm being ranked up may include word frequency-inverse file frequency
Rate (TF-IDF) algorithm and/or text sequence (Text Rank) algorithm.
When TF-IDF algorithm is used alone, first according to focus space predetermined, candidate focus information is matched.So
(process needs to guarantee that each candidate focus information is divided into a word) is segmented to read statement data afterwards, is calculated each
Word frequency (TF) of the candidate focus information relative to read statement data.Each candidate focus letter is calculated by corpus under a large amount of lines again
The inverse document frequency (IDF) of breath calculates the first scoring of each candidate focus information using TF-IDF algorithm.According to each candidate burnt
First scoring of point information is ranked up multiple candidate focus informations, chooses the highest candidate focus information conduct of the first scoring
Focus information.
When Text Rank algorithm is used alone, the second of each candidate focus information is calculated using Text Rank algorithm
Scoring.Multiple candidate focus informations are ranked up according to the second scoring of each candidate focus information, choose the second scoring highest
Candidate focus information as focus information.
When TF-IDF algorithm and Text Rank algorithm is used in combination, for each candidate focus information, to candidate coke
The first scoring and the second scoring of point information are weighted summation, obtain the comprehensive score of candidate's focus information.According to each time
It selects the comprehensive score of focus information to be ranked up multiple candidate focus informations, chooses the highest candidate focus information of comprehensive score
As focus information.
In one embodiment of the present disclosure, operation S203 obtains the first answer number in response to the read statement data
According to including: identification intent information corresponding with read statement data, corresponding original answer number is obtained based on the intent information
According to using the original answer data as the first reply data.
Wherein, (Intent) information that is intended to corresponding with read statement data is to send the sender of the read statement data
The operation wished to carry out gone out by the read statement data representation or the wanting completion of the task.With the defeated of identical intent information
Entering phrase data can be consistent based on original answer data accessed by identical intent information there are many expression-form
, the original answer data be intended to expressed by read statement data go out the operation wished to carry out or want complete task into
Row response, and and be not concerned with read statement data embody form.Such as intent information " cancel order " (" is taken
Disappear order "), corresponding original answer data are as follows: " You may be able to cancel the order if you
Contact us at XXX-XXX-XXX within an hour after placing it. " (" if you place an order it is latter
It dials XXX-XXX-XXX in hour to contact us, you can cancel an order.").
Scheme according to the present embodiment, the first reply data obtained based on intent information and read statement data entirety institute
The operation faced or task are related, the second reply data and the specific focus of read statement data obtained based on focus information
Correlation, in conjunction with the two it is available can reach customer objective adapt to again user expression-form for read statement data
Fining customization third reply data.
In one embodiment of the present disclosure, the second reply data in response to the focus information is obtained in operation S204
Before, method shown in Fig. 2 further include: question sentence type belonging to identification read statement data.On this basis, operation S204 is obtained
Taking the second reply data in response to the focus information includes following operation.
Firstly, obtaining original art (Script) rule that should answer corresponding with the question sentence type.
Wherein, question sentence type belonging to read statement data may include one or more of: alternative question, general
Interrogative sentence, the sentence for asking the time, the sentence for asking size, the sentence for asking model, asks express delivery at the sentence asked in reply interrogative sentence, ask price
Sentence, the sentence for asking place etc..The original art rule characterization that should answer corresponding to question sentence type is suitable with corresponding question sentence type
The answer form matched.
For example, read statement data " Is the phone in my order black or red? " (" in my order
Mobile phone is black or red? ") belonging to alternative question, the original art rule that should answer corresponding with the alternative question is
" ' is black ' or ' is red ' " (" ' being black ' or ' being red ' "), illustrate that the answer form being adapted to alternative question is
Answer one of the project selected.Read statement data " Is my order cancelled? " (" my cancellation of order
? ") belonging to general question, the original art rule that should answer corresponding with the general question is " ' yes ' or ' no ' " (" ' being '
Or 'No' "), illustrate that the answer form being adapted to general question is to answer yes/no.
Then, after obtaining the original art rule that should answer corresponding with the question sentence type, the focus information is utilized
Focus amendment is carried out to the original art rule that should answer, obtains the art data that should answer corresponding with the read statement data,
Art data should be answered as second reply data by described in.
Example above is continued to use, for read statement data " Is the phone in my order black or
Red? " (" mobile phone in my order is black or red? "), the focus information that identification characterizes the read statement data is " the
Phone " (" mobile phone ") identifies that question sentence type belonging to the read statement data is alternative question.It is doubted obtaining with the selection
The corresponding original art rule that should answer of question sentence is benefit after " ' is black ' or ' is red ' " (" ' being black ' or ' being red ' ")
Focus amendment is carried out to the original art rule that should answer with focus information, obtains the art that should answer corresponding with the read statement data
Data are as follows: " ' The phone is black ' or ' The phone is red ' " (" ' mobile phone is black ' or ' mobile phone is red
Color ' ").
As can be seen that being corrected by the focus to the original art rule that should answer corresponding with question sentence type, in the form of answer
Under the premise of still adapting to question sentence type, the focus of user is embodied in the art data that should answer, in the defeated of answer user
Enter the psychological needs for meeting user while the query in phrase data.
Specifically, question sentence type belonging to above-mentioned identification read statement data includes: trained Question Classification model, utilizes institute
State question sentence type belonging to read statement data described in Question Classification model prediction.And/or establish Question Classification rule
Library, according to the classifying rules in the Question Classification rule base determine the read statement data belonging to question sentence type.
In one embodiment of the present disclosure, operation S205 is based on the first reply data and the second reply data and generates the
Three reply datas include: that should answer art data and the original answer data merges to described based on preset fusion rule,
Obtain the third reply data.
Wherein, it may include: the art data that should answer that above-mentioned corresponding answer art data and original answer data, which carry out fusion,
Third reply data is combined into original answer data group.Alternatively, being defined according to the corresponding answer art data of original answer data
Change processing, by through making clear processing should answer art data and original answer data group is combined into third reply data.
For example, for read statement data " Is the phone in my order black or red? " (" my order
In mobile phone be black or red? "), corresponding original answer data are as follows: " red II " (" No. 2 red "), corresponding response
Talk about art data are as follows: " ' The phone is black ' or ' The phone is red ' ", (" ' mobile phone is black ' or ' mobile phone was
It is red ' ").Know that mobile phone is red after obtaining original answer data, therefore according to the corresponding answer art number of the original answer data
It is " The phone is red " (" mobile phone is red ") according to processing is made clear.It will make that treated clear again and should answer art data
With original answer data merged to obtain third reply data " The phone is red, red II. " (" mobile phone be it is red,
No. 2 red.").
Scheme according to the present embodiment is generated in the original answer data based on read statement data and the art data that should answer
When third reply data, when the art data that should answer are unique, directly original answer data and the art data that should answer can be carried out
Fusion can carry out making processing clear when the art data that should answer are not unique according to the corresponding answer art data of original answer data,
The two is merged again, to export to user, result is unique, defines, can answer customer problem and meet user conversation habit
Response.
In one embodiment of the present disclosure, method shown in Fig. 2 further include: obtain predetermined quantity and read statement
After the corresponding third reply data of data, model instruction is carried out based on the read statement data and the third reply data
Practice, training obtains question sentence answer correction model.Wherein, with the corresponding original answer of the intent information of the read statement data,
Question sentence type and focus information are as training data, using third reply data corresponding with the read statement data described in
The label of training data.That is, obtaining enough read statement data and third corresponding with read statement data
After reply data, question sentence answer amendment mould can be obtained based on the read statement data and third reply data training
Type.It then, can be corresponding by the intent information of the read statement data when getting the read statement data of unknown answer
Original answer, question sentence type and focus information input question sentence answer correction model, using question sentence answer correction model prediction with
The corresponding third reply data of the read statement data.
In one embodiment of the present disclosure, method shown in Fig. 2 further include: characterize the input in operation S202 identification
Before the focus information of phrase data, read statement data are pre-processed, as removed the noise, superfluous in read statement data
It is remaining etc..
Below with reference to Fig. 3, method shown in Fig. 2 is illustrated in conjunction with specific embodiments:
Fig. 3 diagrammatically illustrates the schematic diagram of answer generating process according to an embodiment of the present disclosure.
As shown in figure 3, obtaining read statement data after the answer generating process starts, read statement data are in this example
The problem of user inputs.
Focus information identification, question sentence type identification and intent information is first carried out respectively to the problem to extract.
In focus information cognitive phase, identify that focus information, focus information express the focus of user from problem,
It is the tutorial message for answer extracted from problem.It can determine what user wanted by analysis focus information
Answer.Focus information can be obtained by syntactic analysis, key-phrase extraction, band is typically chosen for special question and is had a question
Syntactic analysis can be used as focus information, for declarative sentence in the phrase of word, extracts noun, verb, nominal phrase or dynamic
Part of speech word is as candidate focus information.Confidence evaluation is carried out to candidate focus information, select the highest word of confidence level or
Focus information of the phrase as this.
For example, problem " How to cancel my order? " the focus information of (" order for how cancelling me ") is
" how to " (" how "), problem " How much is my phone? " the focus information of (" my mobile phone how much ") is
" how much " (" how much "), the focus letter of problem " I want to cancel my order. " (" I wants to cancel an order ")
Breath is " cancel my order " (" cancelling an order ").
In the question sentence type identification stage, classify to question sentence clause type belonging to problem.By to Question Classification energy
Enough positioning and analyze the information that user wants to obtain is what, while may determine that from question sentence sentence pattern and releasing answer form.It builds
Vertical question sentence classification of type system, excavates the clause sentence pattern information in question sentence.The foundation of Question Classification system combines common sentence pattern, and
It needs to sort out special business classification according to business and is used as entire classification system.Classification method take support vector machines (SVM),
The disaggregated models such as the text classification (TextCNN) based on convolutional neural networks, while analyzing corpus feature and establish rule base and being used for
Assist in identifying question sentence type.Sentence pattern clause using the corresponding answer of question sentence of identical sentence pattern clause be it is similar, according to asking
Sentence sentence pattern inquires corresponding answer words art from words art knowledge base.
Since special question is covered than broad, special question is carried out according to specific business system in the method
Subdivision.The citing of sentence pattern classification: alternative question, general question ask in reply interrogative sentence, ask price, ask size, ask model, ask
Express delivery asks arrival time, asks place etc..
For example, problem " Is the phone in my order black or red? " (" mobile phone in my order is black
Color or red? ") belong to alternative question, problem " Is my order cancelled? " (" my cancellation of order? ")
Belong to general question, problem " How much is my phone? " (" my mobile phone how much? ") belong to and ask price, problem
" What ' s the size of the screen? " (" size of screen is how many? ") belong to and ask size, problem " When
Will I receive my package? " (" when I receive package? ") belong to and ask the time.
Be intended to the information extraction stage, the user expressed in extraction problem is intended to, user be intended to refer to user with intelligence
The thing really felt like doing, the information wanted are thought in conversational system interactive process.Identify user be intended to after according to
Family is intended to release corresponding answer.For example, problem problem " How to cancel my order? " (" how to cancel my order
Intent information singly ") is " cancel order " (" cancelling an order ").Intent information extraction process is regarded in this programme and is divided
Generic task is handled, and is pre-defined a lot of and is intended to classification, passes through convolutional neural networks (CNN), support vector machines (SVM) equal part
Class model is realized.
Then, it on the one hand carries out words art to extract and correct, the process is according to focus information and question sentence type from words art library
Inference Search goes out corresponding revised words art.Specifically, corresponding original response is extracted from words art library based on question sentence type
Art rule is talked about, the original art rule that should answer is modified according to focus information, the art data that should answer is obtained, is shown in Fig. 3
For revised words art, corresponding to the second reply data in Fig. 2.On the other hand answer library searching is carried out, intent information is based on
Corresponding original answer data is retrieved from answer library, original answer is shown as in Fig. 3, corresponding to the first answer number in Fig. 2
According to.
Then, it is based on fusion rule library, answer mixing operation is carried out to revised words art and original answer, is obtained final
Answer, the final result correspond to the third reply data in Fig. 2.
So far entire answer generating process terminates, and the above process has used three kinds of knowledge bases: words art library, answer library and fusion
Rule base.Art may include art if problem types and user focus correspondence in words art library, have in answer library for not
It is intended to the answer editted together, has the various rules of words art and answer fusion in fusion rule library.
In a specific example, user input the problem of are as follows: " I found I chose the wrong color
Of my phone, can I cancel my order? " (" I found that selecting mistake mobile phone color, I can cancel my order
? ").It for the problem, is first pre-processed, pretreatment includes segmenting, removing stop words etc..Pretreated question sentence is carried out
Focus information identifies that focus information is " cancel my order " (" order for cancelling me ") in this sentence.Pretreated is asked
Sentence carries out question sentence type identification: corresponding to the general question of " ' yes ' or ' no ' " (" ' being ' or 'No' ") words art.To pretreatment
The question sentence crossed carries out intent information extraction, and obtaining intent information is " cancel order " (" cancelling an order "), and according to intention
Information inquiry answer library obtains original answer data: " You may be able to cancel the order if you
Contact us at XXX-XXX-XXX within an hour after placing it. " (" if you place an order it is latter
It dials XXX-XXX-XXX in hour to contact us, you can cancel an order.").Focus information and question sentence classification information are carried out
Talk about art reasoning, obtain revised words art: " Yes you can. " (" be you can be with.").By original answer data and amendment
Art carries out answer fusion afterwards, obtains final result by answer create-rule or generation model: " Yes you
can.You may be able tO cancel the order if you contact us at XXX-XXX-XXX
Within an hour after placing it. " (" be you can be with.If you dial XXX- within the latter hour that places an order
XXX-XXX contacts us, you can cancel an order.").The final result is exported to user.
The answer that the embodiment of the present disclosure provides generates scheme and not only wants the information that lets the user know that user's output, also from
Expectation of the user to answer from the point of view of family.Question sentence type and focus in the problem of sufficiently excavating and user's input is utilized
Information.These information make generate answer expression-form it is more smooth, naturally, being more in line with user on sentence pattern and content
Enquirement.For the enquirement of various personalizations, the desired answer of user can be provided, user's think time is reduced, to make
Obtain entire question answering process remarkable fluency.Wherein, the identification of question sentence classification combines common clause sentence pattern, and increases for practical business
Business correlation question sentence type is added, so that Question Classification is more in line with practical business, and has been combined first in answer generating process
It tests knowledge base and deep neural network model generates answer jointly, meet the accuracy and diversity of answer.
Fig. 4 diagrammatically illustrates the block diagram of answer generating means according to an embodiment of the present disclosure.The answer generating means
400 are applied in intelligent conversational system, and the intelligence conversational system can be responded and be provided to the input information received
Feedback information.
As shown in figure 4, answer generating means 400 include: the first acquisition module 410, the first identification module 420, second obtain
Modulus block 430, third obtain module 440, integrated treatment module 450 and output module 460.
First acquisition module 410 is for obtaining read statement data.
First identification module 420 characterizes the focus information of the read statement data for identification.
Second acquisition module 430 is used to obtain the first reply data in response to the read statement data.
Third obtains module 440 and is used to obtain the second reply data in response to the focus information.
Integrated treatment module 450 is used to generate third response based on first reply data and second reply data
Data.
Output module 460 is for exporting the third reply data.
In embodiment of the disclosure, above-mentioned focus information is for characterizing what user in the read statement data most paid close attention to
Informational content.
Fig. 5 diagrammatically illustrates the block diagram of answer generating means according to another embodiment of the present disclosure.As shown in figure 5,
Answer generating means 500 include: the first acquisition module 410, the first identification module 420, second obtains module 430, third obtains
Module 440, integrated treatment module 450 and output module 460.
In one embodiment of the present disclosure, the first identification module 420 includes: building submodule 421, extracting sub-module
422 and first determine submodule 423.
Building submodule 421 is for constructing focus information rule base.Extracting sub-module 422 is used for from the read statement number
The one or more piece segment informations to match according to middle extraction with the rule in the focus information rule base.First determines submodule
423 are used for when extracting a piece segment information, using the piece segment information as the focus information for characterizing the read statement data;
When extracting multiple segment informations, highest segment information of confidence level is chosen as the focus for characterizing the read statement data
Information.
In one embodiment of the present disclosure, the second acquisition module 430 includes: that identification submodule 431 and second determine son
Module 432.Identify the intent information corresponding with the read statement data for identification of submodule 431.Second determines submodule
432 for obtaining corresponding original answer data based on the intent information, using the original answer data as described first
Reply data.
Further, as an optional embodiment, answer generating means 500 further include the second identification module 470.
Second identification module 470 is used to obtain the acquisition of module 440 in response to the second response of the focus information in third
Before data, question sentence type belonging to the read statement data is identified.
On this basis, it includes acquisition submodule 441 and amendment submodule 442 that third, which obtains module 440,.
Acquisition submodule 441 is regular for obtaining the original art that should answer corresponding with the question sentence type, described original to answer
The answer form of art rule of answering characterization and the question sentence type adaptation.Submodule 442 is corrected to be used to utilize the focus information
Focus amendment is carried out to the original art rule that should answer, obtains the art data that should answer corresponding with the read statement data,
Art data should be answered as second reply data by described in.
Specifically, the second identification module 470 utilizes the Question Classification model prediction institute for training Question Classification model
State question sentence type belonging to read statement data;And/or Question Classification rule base is established, is advised according to the Question Classification
Then the classifying rules in library determines question sentence type belonging to the read statement data.
In one embodiment of the present disclosure, integrated treatment module 450 is specifically used for being based on preset fusion rule, to described
The art that should answer data and the original answer data are merged, and the third reply data is obtained.
Specifically, integrated treatment module 450 includes processing submodule 451 and fusion submodule 452.
Processing submodule 451 to the art data that should answer according to the original answer data for carrying out making place clear
Reason.Fusion submodule 452 is used for should answer art data and the original answer data combination by making processing clear
For third reply data.
In one embodiment of the present disclosure, answer generating means 500 further include model training module 480, for obtaining
After the third reply data corresponding with read statement data of predetermined quantity, based on the read statement data and described the
Three reply datas carry out model training, wherein with the corresponding original answer of the intent information of the read statement data, question sentence class
Type and focus information are as training data, using third reply data corresponding with the read statement data as the trained number
According to label, training obtain question sentence answer correction model.
In one embodiment of the present disclosure, answer generating means 500 further include preprocessing module 490, for first
Before the identification of identification module 420 characterizes the focus information of the read statement data, the read statement data are located in advance
Reason.
It should be noted that in device section Example each module/unit/subelement etc. embodiment, the skill of solution
Art problem, the function of realization and the technical effect reached respectively with the implementation of corresponding step each in method section Example
Mode, the technical issues of solving, the function of realization and the technical effect that reaches are same or like, and details are not described herein.
It is module according to an embodiment of the present disclosure, submodule, unit, any number of or in which any more in subelement
A at least partly function can be realized in a module.It is single according to the module of the embodiment of the present disclosure, submodule, unit, son
Any one or more in member can be split into multiple modules to realize.According to the module of the embodiment of the present disclosure, submodule,
Any one or more in unit, subelement can at least be implemented partly as hardware circuit, such as field programmable gate
Array (FPGA), programmable logic array (PLA), system on chip, the system on substrate, the system in encapsulation, dedicated integrated electricity
Road (ASIC), or can be by the hardware or firmware for any other rational method for integrate or encapsulate to circuit come real
Show, or with any one in three kinds of software, hardware and firmware implementations or with wherein any several appropriately combined next reality
It is existing.Alternatively, can be at least by part according to one or more of the module of the embodiment of the present disclosure, submodule, unit, subelement
Ground is embodied as computer program module, when the computer program module is run, can execute corresponding function.
For example, first obtains module 410, the first identification module 420, second obtains module 430, third obtains module 440,
In integrated treatment module 450, output module 460, the second identification module 470, model training module 480 and preprocessing module 490
Any number of may be incorporated in a module realize or any one module therein can be split into multiple moulds
Block.Alternatively, at least partly function of one or more modules in these modules can be at least partly function of other modules
It combines, and is realized in a module.In accordance with an embodiment of the present disclosure, first obtain module 410, the first identification module 420,
Second obtain module 430, third obtain module 440, integrated treatment module 450, output module 460, the second identification module 470,
At least one of model training module 480 and preprocessing module 490 can at least be implemented partly as hardware circuit, such as
Field programmable gate array (FPGA) programmable logic array (PLA), system on chip, the system on substrate, in encapsulation is
System, specific integrated circuit (ASIC), or can be hard by carrying out any other rational method that is integrated or encapsulating etc. to circuit
Part or firmware realize, or with any one in three kinds of software, hardware and firmware implementations or with wherein any several
It is appropriately combined to realize.Alternatively, first obtains module 410, the first identification module 420, second obtains module 430, third obtains
Module 440, integrated treatment module 450, output module 460, the second identification module 470, model training module 480 and pretreatment mould
At least one of block 490 can at least be implemented partly as computer program module, when the computer program module is transported
When row, corresponding function can be executed.
Fig. 6 diagrammatically illustrates the intelligent session system according to an embodiment of the present disclosure for being adapted for carrying out method as described above
The block diagram of system.Intelligent conversational system shown in Fig. 6 is only an example, function to the embodiment of the present disclosure and should not use model
Shroud carrys out any restrictions.
As shown in fig. 6, intelligent conversational system 600 includes processor 610 and computer readable storage medium 620.The intelligence
Conversational system 600 can execute the method according to the embodiment of the present disclosure.
Specifically, processor 610 for example may include general purpose microprocessor, instruction set processor and/or related chip group
And/or special microprocessor (for example, specific integrated circuit (ASIC)), etc..Processor 610 can also include using for caching
The onboard storage device on way.Processor 610 can be the different movements for executing the method flow according to the embodiment of the present disclosure
Single treatment unit either multiple processing units.
Computer readable storage medium 620, such as can be non-volatile computer readable storage medium, specific example
Including but not limited to: magnetic memory apparatus, such as tape or hard disk (HDD);Light storage device, such as CD (CD-ROM);Memory, such as
Random access memory (RAM) or flash memory;Etc..
Computer readable storage medium 620 may include computer program 621, which may include generation
Code/computer executable instructions execute processor 610 according to the embodiment of the present disclosure
Method or its any deformation.
Computer program 621 can be configured to have the computer program code for example including computer program module.Example
Such as, in the exemplary embodiment, the code in computer program 621 may include one or more program modules, for example including
621A, module 621B ....It should be noted that the division mode and number of module are not fixation, those skilled in the art can
To be combined according to the actual situation using suitable program module or program module, when these program modules are combined by processor 610
When execution, processor 610 is executed according to the method for the embodiment of the present disclosure or its any deformation.
According to an embodiment of the invention, first obtains module 410, the first identification module 420, second obtains module 430, the
Three obtain modules 440, integrated treatment module 450, output module 460, the second identification module 470, model training module 480 and pre-
At least one of processing module 490 can be implemented as the computer program module with reference to Fig. 6 description, by processor 610
When execution, answer generation method described above may be implemented.
The disclosure additionally provides a kind of computer readable storage medium, which can be above-mentioned reality
It applies included in equipment/device/system described in example;Be also possible to individualism, and without be incorporated the equipment/device/
In system.Above-mentioned computer readable storage medium carries one or more program, when said one or multiple program quilts
When execution, the method according to the embodiment of the present disclosure is realized.
In accordance with an embodiment of the present disclosure, computer readable storage medium can be non-volatile computer-readable storage medium
Matter, such as can include but is not limited to: portable computer diskette, hard disk, random access storage device (RAM), read-only memory
(ROM), erasable programmable read only memory (EPROM or flash memory), portable compact disc read-only memory (CD-ROM), light
Memory device, magnetic memory device or above-mentioned any appropriate combination.In the disclosure, computer readable storage medium can
With to be any include or the tangible medium of storage program, the program can be commanded execution system, device or device use or
Person is in connection.
Flow chart and block diagram in attached drawing are illustrated according to the system of the various embodiments of the disclosure, method and computer journey
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part of one module, program segment or code of table, a part of above-mentioned module, program segment or code include one or more
Executable instruction for implementing the specified logical function.It should also be noted that in some implementations as replacements, institute in box
The function of mark can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are practical
On can be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.Also it wants
It is noted that the combination of each box in block diagram or flow chart and the box in block diagram or flow chart, can use and execute rule
The dedicated hardware based systems of fixed functions or operations is realized, or can use the group of specialized hardware and computer instruction
It closes to realize.
It will be understood by those skilled in the art that the feature recorded in each embodiment and/or claim of the disclosure can
To carry out multiple combinations and/or combination, even if such combination or combination are not expressly recited in the disclosure.Particularly, exist
In the case where not departing from disclosure spirit or teaching, the feature recorded in each embodiment and/or claim of the disclosure can
To carry out multiple combinations and/or combination.All these combinations and/or combination each fall within the scope of the present disclosure.
Although the disclosure, art technology has shown and described referring to the certain exemplary embodiments of the disclosure
Personnel it should be understood that in the case where the spirit and scope of the present disclosure limited without departing substantially from the following claims and their equivalents,
A variety of changes in form and details can be carried out to the disclosure.Therefore, the scope of the present disclosure should not necessarily be limited by above-described embodiment,
But should be not only determined by appended claims, also it is defined by the equivalent of appended claims.
Claims (10)
1. a kind of answer generation method is applied in intelligent conversational system, the intelligence conversational system can be defeated to what is received
Enter information to be responded and feedback information is provided, which comprises
Obtain read statement data;
Identification characterizes the focus information of the read statement data;
Obtain the first reply data in response to the read statement data;
Obtain the second reply data in response to the focus information;
Third reply data is generated based on first reply data and second reply data;And
Export the third reply data.
2. according to the method described in claim 1, wherein, the focus information is for characterizing user in the read statement data
The informational content most paid close attention to.
3. according to the method described in claim 1, wherein, the identification characterizes the focus information packet of the read statement data
It includes:
Construct focus information rule base;
From one or more pieces that extraction matches with the rule in the focus information rule base in the read statement data
Segment information;
When extracting a piece segment information, using the piece segment information as the focus information for characterizing the read statement data;With
And
When extracting multiple segment informations, highest segment information of confidence level is chosen as the characterization read statement data
Focus information.
4. according to the method described in claim 1, wherein, first answer number obtained in response to the read statement data
According to including:
Identify intent information corresponding with the read statement data;And
Corresponding original answer data is obtained based on the intent information, using the original answer data as first response
Data.
5. method according to claim 1 or 4, in which:
Before the acquisition is in response to the second reply data of the focus information, the method also includes: identification is described defeated
Enter question sentence type belonging to phrase data;
Described obtain include: in response to the second reply data of the focus information
The original art rule that should answer corresponding with the question sentence type is obtained, the original art rule characterization that should answer is asked with described
The answer form of sentence type adaptation;And
Focus amendment is carried out to the original art rule that should answer using the focus information, is obtained and the read statement data
The corresponding art data that should answer should answer art data as second reply data by described in.
6. according to the method described in claim 5, wherein, question sentence type packet belonging to the identification read statement data
It includes:
Training Question Classification model, utilizes question sentence type belonging to read statement data described in the Question Classification model prediction;
And/or
Question Classification rule base is established, the read statement data are determined according to the classifying rules in the Question Classification rule base
Affiliated question sentence type.
7. described to be based on first reply data and second reply data according to the method described in claim 5, wherein
Generating third reply data includes:
Based on preset fusion rule, art data should be answered and the original answer data merges to described, obtain described
Three reply datas.
8. according to the method described in claim 7, wherein, it is described to it is described should answer art data and the original answer data into
Row merges
The art data that should answer are carried out making processing clear according to the original answer data;And
Art data will should be answered described in processing and the original answer data group is combined into third reply data by making clear.
9. according to the method described in claim 5, further include:
After the third reply data corresponding with read statement data for obtaining predetermined quantity, it is based on the read statement data
Model training is carried out with the third reply data, wherein original is answered so that the intent information of the read statement data is corresponding
Case, question sentence type and focus information as training data, using third reply data corresponding with the read statement data as
The label of the training data, training obtain question sentence answer correction model;And/or
Before the focus information that identification characterizes the read statement data, the read statement data are pre-processed.
10. a kind of intelligence conversational system, the intelligence conversational system can be responded and be provided to the input information received
Feedback information, the intelligence conversational system include memory, processor and storage on a memory and can run on a processor
Computer program, for realizing answer such as according to any one of claims 1 to 9 when the processor executes described program
Generation method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811653405.XA CN109739969A (en) | 2018-12-29 | 2018-12-29 | Answer generation method and intelligent conversational system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811653405.XA CN109739969A (en) | 2018-12-29 | 2018-12-29 | Answer generation method and intelligent conversational system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109739969A true CN109739969A (en) | 2019-05-10 |
Family
ID=66363013
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811653405.XA Pending CN109739969A (en) | 2018-12-29 | 2018-12-29 | Answer generation method and intelligent conversational system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109739969A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110287325A (en) * | 2019-06-28 | 2019-09-27 | 南方电网科学研究院有限责任公司 | A kind of power grid customer service recommended method and device based on intelligent sound analysis |
CN110457661A (en) * | 2019-08-16 | 2019-11-15 | 腾讯科技(深圳)有限公司 | Spatial term method, apparatus, equipment and storage medium |
CN111506715A (en) * | 2020-04-13 | 2020-08-07 | 深圳追一科技有限公司 | Query method and device, electronic equipment and storage medium |
CN112084767A (en) * | 2020-07-22 | 2020-12-15 | 联想(北京)有限公司 | Information response processing method, intelligent voice device and storage medium |
CN112883177A (en) * | 2021-02-18 | 2021-06-01 | 联想(北京)有限公司 | Intelligent reply method and device |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100268716A1 (en) * | 2009-04-17 | 2010-10-21 | Degaugue Fabien G | System and methods of providing interactive expertized communications responses using single and multi-channel site-specific integration |
CN104573028A (en) * | 2015-01-14 | 2015-04-29 | 百度在线网络技术(北京)有限公司 | Intelligent question-answer implementing method and system |
CN106649786A (en) * | 2016-12-28 | 2017-05-10 | 北京百度网讯科技有限公司 | Deep question answer-based answer retrieval method and device |
US20180025075A1 (en) * | 2016-07-20 | 2018-01-25 | International Business Machines Corporation | Evaluating Temporal Relevance in Question Answering |
CN107908803A (en) * | 2017-12-26 | 2018-04-13 | 上海智臻智能网络科技股份有限公司 | The response method and device, storage medium, terminal of question and answer interaction |
CN108846063A (en) * | 2018-06-04 | 2018-11-20 | 北京百度网讯科技有限公司 | Determine the method, apparatus, equipment and computer-readable medium of problem answers |
-
2018
- 2018-12-29 CN CN201811653405.XA patent/CN109739969A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100268716A1 (en) * | 2009-04-17 | 2010-10-21 | Degaugue Fabien G | System and methods of providing interactive expertized communications responses using single and multi-channel site-specific integration |
CN104573028A (en) * | 2015-01-14 | 2015-04-29 | 百度在线网络技术(北京)有限公司 | Intelligent question-answer implementing method and system |
US20180025075A1 (en) * | 2016-07-20 | 2018-01-25 | International Business Machines Corporation | Evaluating Temporal Relevance in Question Answering |
CN106649786A (en) * | 2016-12-28 | 2017-05-10 | 北京百度网讯科技有限公司 | Deep question answer-based answer retrieval method and device |
CN107908803A (en) * | 2017-12-26 | 2018-04-13 | 上海智臻智能网络科技股份有限公司 | The response method and device, storage medium, terminal of question and answer interaction |
CN108846063A (en) * | 2018-06-04 | 2018-11-20 | 北京百度网讯科技有限公司 | Determine the method, apparatus, equipment and computer-readable medium of problem answers |
Non-Patent Citations (1)
Title |
---|
周洪宇 等: "《长江教育论丛》", 29 February 2016 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110287325A (en) * | 2019-06-28 | 2019-09-27 | 南方电网科学研究院有限责任公司 | A kind of power grid customer service recommended method and device based on intelligent sound analysis |
CN110457661A (en) * | 2019-08-16 | 2019-11-15 | 腾讯科技(深圳)有限公司 | Spatial term method, apparatus, equipment and storage medium |
CN110457661B (en) * | 2019-08-16 | 2023-06-20 | 腾讯科技(深圳)有限公司 | Natural language generation method, device, equipment and storage medium |
CN111506715A (en) * | 2020-04-13 | 2020-08-07 | 深圳追一科技有限公司 | Query method and device, electronic equipment and storage medium |
CN112084767A (en) * | 2020-07-22 | 2020-12-15 | 联想(北京)有限公司 | Information response processing method, intelligent voice device and storage medium |
CN112883177A (en) * | 2021-02-18 | 2021-06-01 | 联想(北京)有限公司 | Intelligent reply method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109739969A (en) | Answer generation method and intelligent conversational system | |
US11328017B2 (en) | Method and system for generating a conversational agent by automatic paraphrase generation based on machine translation | |
US9582757B1 (en) | Scalable curation system | |
CA2929018C (en) | Natural expression processing method, processing and response method, device and system | |
US11010700B2 (en) | Identifying task and personality traits | |
WO2022095380A1 (en) | Ai-based virtual interaction model generation method and apparatus, computer device and storage medium | |
CN111428010B (en) | Man-machine intelligent question-answering method and device | |
CN106202301A (en) | A kind of intelligent response system based on degree of depth study | |
CN106407178A (en) | Session abstract generation method and device | |
US11756532B2 (en) | Intelligence-driven virtual assistant for automated idea documentation | |
US20140081643A1 (en) | System and method for determining expertise through speech analytics | |
CN110046230B (en) | Method for generating recommended speaking collection, and recommended speaking method and device | |
JP2020071869A (en) | Video-based job provider and job seeker matching server and method | |
KR102169382B1 (en) | Artificial Intelligence-Based Personalized Expert Cross Matching and Proposal System | |
CN110347863A (en) | Talk about art recommended method and device and storage medium | |
CN116235177A (en) | Systems and methods related to robotic authoring by mining intent from dialogue data using known intent of an associated sample utterance | |
CN110019742A (en) | Method and apparatus for handling information | |
US20190295199A1 (en) | Intelligent legal simulator | |
CN109460503A (en) | Answer input method, device, storage medium and electronic equipment | |
JP7096172B2 (en) | Devices, programs and methods for generating dialogue scenarios, including utterances according to character. | |
CN108595609A (en) | Generation method, system, medium and equipment are replied by robot based on personage IP | |
CN108959529A (en) | Determination method, apparatus, equipment and the storage medium of problem answers type | |
US11114092B2 (en) | Real-time voice processing systems and methods | |
US20220392434A1 (en) | Reducing biases of generative language models | |
CN116383027B (en) | Man-machine interaction data processing method and server |
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 |