CN109933652A - Intelligent answer method, apparatus, computer equipment and storage medium - Google Patents
Intelligent answer method, apparatus, computer equipment and storage medium Download PDFInfo
- Publication number
- CN109933652A CN109933652A CN201910043210.1A CN201910043210A CN109933652A CN 109933652 A CN109933652 A CN 109933652A CN 201910043210 A CN201910043210 A CN 201910043210A CN 109933652 A CN109933652 A CN 109933652A
- Authority
- CN
- China
- Prior art keywords
- question
- target text
- term vector
- answering model
- input
- 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
- 238000000034 method Methods 0.000 title claims abstract description 63
- 239000013598 vector Substances 0.000 claims abstract description 96
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000013528 artificial neural network Methods 0.000 claims abstract description 12
- 238000004590 computer program Methods 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 16
- 238000010276 construction Methods 0.000 claims description 11
- 230000000306 recurrent effect Effects 0.000 claims description 11
- 230000007246 mechanism Effects 0.000 claims description 10
- 238000010586 diagram Methods 0.000 description 15
- 238000012545 processing Methods 0.000 description 8
- 239000000463 material Substances 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 235000021167 banquet Nutrition 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000002996 emotional effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007087 memory ability Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000002688 persistence Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention discloses a kind of intelligent answer method, apparatus, computer equipment and storage mediums.The present invention is applied to the field of neural networks in prediction model.The described method includes: collecting corpus as target text using predetermined manner；Pretreatment is carried out to the target text and is converted by term vector tool to pretreated target text is carried out to obtain term vector；Building is trained using coding and decoding as the Question-Answering Model of frame using the term vector as the input of the Question-Answering Model and to the Question-Answering Model in a manner of loop iteration；If detecting the input of problem, being predicted by the Question-Answering Model after training the answer of described problem and exporting prediction result.Method by implementing the embodiment of the present invention can fully consider the context of question and answer, improve question and answer precision, save human resources and improve user experience.
Description
Technical field
The present invention relates to field of artificial intelligence more particularly to a kind of intelligent answer method, apparatus, computer equipment and
Storage medium.
Background technique
With science and technology and expanding economy, the scale of insurance industry is very grand, and miscellaneous insurance widely goes out
In present people's lives.User is usually carried out by phone contact insurance company when insurance risk occurs in face of Claims Resolution
Processing, and insurance company needs to be arranged and be attended a banquet accordingly to handle the phone that user makes, and reports a case to the security authorities in daily up to ten thousand Claims Resolutions
In the case of, and the processing of each Claims Resolution case also needs multiple telephonic communication, the consuming of human cost is huge.Currently, insurance company
It shares the case attended a banquet using some simple problems of user are answered by setting customer service robot to handle, in face of client's
It puts question to, customer service robot is generally only to return to fixed text by configuring keyword to answer the question, and is not accounted for up and down
The combination of text, does not recognize the true emotional of user yet, and only mechanical answer, user experience are poor.
Summary of the invention
The embodiment of the invention provides a kind of intelligent answer method, apparatus, computer equipment and storage mediums, it is intended to solve
Problem that is mechanical and causing user experience poor is answered in face of puing question to by existing customer service robot.
In a first aspect, the embodiment of the invention provides a kind of intelligent answer methods comprising: language is collected using predetermined manner
Material is used as target text；The target text pre-process and by term vector tool to progress pretreated target text
This is converted to obtain term vector；Building asks the term vector using coding and decoding as the Question-Answering Model of frame as described in
It answers the input of model and the Question-Answering Model is trained in a manner of loop iteration；If detecting the input of problem, pass through
Question-Answering Model after training predicts the answer of described problem and exports prediction result.
Second aspect, the embodiment of the invention also provides a kind of intelligent answer devices comprising: collector unit, for adopting
Predetermined manner is used to collect corpus as target text；Pretreatment unit, for the target text to be pre-processed and passed through
Term vector tool is converted to pretreated target text is carried out to obtain term vector；Construction unit, for constructing to compile
Code is decoded as the Question-Answering Model of frame, using the term vector as the input of the Question-Answering Model and in a manner of loop iteration pair
The Question-Answering Model is trained；Predicting unit, if for detecting the input of problem, by the Question-Answering Model after training to institute
The answer for stating problem is predicted and exports prediction result.
The third aspect, the embodiment of the invention also provides a kind of computer equipments comprising memory and processor, it is described
Computer program is stored on memory, the processor realizes the above method when executing the computer program.
Fourth aspect, it is described computer-readable to deposit the embodiment of the invention also provides a kind of computer readable storage medium
Storage media is stored with computer program, and the computer program can realize the above method when being executed by a processor.
The embodiment of the invention provides a kind of intelligent answer method, apparatus, computer equipment and storage mediums.Wherein, institute
The method of stating includes: to collect corpus as target text using predetermined manner；The target text is pre-processed and passes through word
Vector tool is converted to pretreated target text is carried out to obtain term vector；Construct asking using coding and decoding as frame
Model is answered, is carried out using the term vector as the input of the Question-Answering Model and to the Question-Answering Model in a manner of loop iteration
Training；If detecting the input of problem, being predicted by the Question-Answering Model after training the answer of described problem and exporting and is pre-
Survey result.The embodiment of the present invention can fully consider the upper of question and answer since component Question-Answering Model predicts problem to obtain answer
Hereafter, question and answer precision is improved, human resources are saved and improves user experience.
Detailed description of the invention
Technical solution in order to illustrate the embodiments of the present invention more clearly, below will be to needed in embodiment description
Attached drawing is briefly described, it should be apparent that, drawings in the following description are some embodiments of the invention, general for this field
For logical technical staff, without creative efforts, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is the application scenarios schematic diagram of intelligent answer method provided in an embodiment of the present invention；
Fig. 2 is the flow diagram of intelligent answer method provided in an embodiment of the present invention；
Fig. 3 is the sub-process schematic diagram of intelligent answer method provided in an embodiment of the present invention；
Fig. 4 be another embodiment of the present invention provides intelligent answer method flow diagram；
Fig. 5 is the sub-process schematic diagram of intelligent answer method provided in an embodiment of the present invention；
Fig. 6 is the sub-process schematic diagram of intelligent answer method provided in an embodiment of the present invention；
Fig. 7 is the schematic block diagram of intelligent answer device provided in an embodiment of the present invention；
Fig. 8 is the schematic block diagram of the specific unit of intelligent answer device provided in an embodiment of the present invention；
Fig. 9 be another embodiment of the present invention provides intelligent answer device schematic block diagram；And
Figure 10 is the schematic block diagram of computer equipment provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair
Embodiment in bright, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, shall fall within the protection scope of the present invention.
It should be appreciated that ought use in this specification and in the appended claims, term " includes " and "comprising" instruction
Described feature, entirety, step, operation, the presence of element and/or component, but one or more of the other feature, whole is not precluded
Body, step, operation, the presence or addition of element, component and/or its set.
It is also understood that mesh of the term used in this description of the invention merely for the sake of description specific embodiment
And be not intended to limit the present invention.As description of the invention and it is used in the attached claims, unless on
Other situations are hereafter clearly indicated, otherwise " one " of singular, "one" and "the" are intended to include plural form.
It will be further appreciated that the term "and/or" used in description of the invention and the appended claims is
Refer to any combination and all possible combinations of one or more of associated item listed, and including these combinations.
Fig. 1 and Fig. 2 are please referred to, Fig. 1 is the application scenarios schematic diagram of intelligent answer method provided in an embodiment of the present invention.Figure
2 be the schematic flow chart of intelligent answer method provided in an embodiment of the present invention.The intelligent answer method is applied in terminal 10,
By interacting realization between terminal 10 and server 20.This method is particularly applicable in intelligence Claims Resolution customer service robot, Yong Hu
When to about Claims Resolution case generation query, problem is proposed to intelligence Claims Resolution customer service robot, passes through intelligence Claims Resolution customer service robot
Realize the question answer dialog with user
Fig. 2 is the flow diagram of intelligent answer method provided in an embodiment of the present invention.As shown, this method include with
Lower step S110-140.
S110, corpus is collected as target text using predetermined manner.
In one embodiment, for training smart Claims Resolution customer service robot, it is necessary first to collect various corpus as training
Data, wherein the collection of corpus mainly includes two kinds of predetermined manners, and one is the modes on line, i.e., collects language from from webpage
Material, another kind is the mode under line, i.e., obtains from the practical call in insurance company's Claims Resolution case.
In one embodiment, as shown in figure 3, the step S110 may include step S111a-S112a.
S111a, target webpage is obtained by way of web crawlers.
In one embodiment, web crawlers refers to one kind according to certain rules, automatically grabs web message
Program mainly includes acquisition, store and three parts of processing specifically select the URL of representative webpage first
Start to grab data from server as initial URL, for example, the official websites such as Pingan Insurance, China Life Insurance and Pacific Ocean insurance；
Then parsed filter will be carried out after the web storage grabbed, includes new URL in the initial URL grabbed, and parsing is just
Beginning URL is filtered selection URL relevant to question and answer to new URL, for example, the URL of FAQ FAQs be put into etc. it is to be captured
In URL queue, remaining incoherent URL is abandoned；The finally selection next step webpage to be grabbed in URL queue to be captured
URL, and repeat the above process, stop when traversing whole network.
S112a, target text is obtained to target webpage progress data cleansing.
In one embodiment, it after getting web page contents, needs to carry out further data cleansing to webpage, because
It also include the useless data such as picture, link other than content of text in the web page contents crawled.Specifically, by just
Then expression formula removes these hashes, and the group of specific character pre-set first or character is combined into matching rule
Then, for example, Chinese, english punctuation mark etc., are matched with the content in target webpage by the matching rule of setting, only stayed
The content of text of lower successful match filters out the unsuccessful picture of matching and link, to obtain target text.
In another embodiment, as shown in figure 4, the step S110 may include step S111A-S112A.
S111A, the calling record that Claims Resolution case is obtained from presetting database.
In one embodiment, customary risks company, which can be arranged to attend a banquet accordingly, comes the Claims Resolution phone number of connecting subscribers participating, institute
Have and is required to save with the calling record of user.Since the calling record of such and user session is true communication exchange, use
Happiness, anger, grief and joy, inquiry logic and the communication purpose at family are embodied in calling record, therefore select the call record of Claims Resolution case
Sound can make the question and answer exchange of intelligence Claims Resolution customer service robot more approaching to reality as the target text of training.Specifically,
The calling record storage for case of settling a claim is in the preset database, wherein and telephonograph is named according to time rule, for example,
201809181400, calling record is obtained from presetting database according to the title of telephonograph.
S112A, the calling record is converted by target text by speech tool.
In one embodiment, since what is got is calling record, what is presented is acoustic information, is needed further
Acoustic information is converted into text information.Therefore, by speech tool, for example, news fly or Baidu's speech tool,
Accessed calling record batch input is converted into the speech tool, the text of converted output is
Target text.
S120, the target text pre-process and by term vector tool to the pretreated target text of progress
It is converted to obtain term vector.
In one embodiment, since accessed target text is plain text content, for example, entire article, a section
Fall, several sentences or be one section dialogue, these target texts be can by the mankind identify natural-sounding but computer can not
Identification, it is therefore desirable to pretreatment be carried out to these target texts and vector is converted, to obtain the number that computer is capable of identifying processing
According to.
In one embodiment, as shown in figure 5, the step S120 may include step S121-S122.
S121, the target text is segmented by participle tool.
In one embodiment, pretreatment refers to handling target text into the process segmented, and participle refers to connect
Continuous word sequence is reassembled into the process of word sequence according to certain rules, specifically, first by the stop words of target text,
The removal such as punctuate, symbol and messy code, then analyzes text using participle tool, for example, syntactic analysis, semantic disambiguation
And part-of-speech tagging etc., it is finally segmented to obtain individual word one by one based on the analysis results.Such as: " it may I ask vehicle insurance reason
Compensation needs to prepare any material ", obtain after participle " may I ask/vehicle insurance Claims Resolution/needs/preparation/what/material ".In the present solution,
It is segmented using stammerer participle tool, it is of course possible to understanding, it can also be using other participle tools.
S122, term vector is converted to the participle progress vector in the target text according to term vector tool.
In one embodiment, using word2vec as term vector tool, word2vec is a kind of natural language processing work
Tool, effect are exactly that the words in natural language is switched to the term vector that computer is understood that.Traditional term vector be easy by
The puzzlement of dimension disaster, and between any two word be all it is isolated, the relationship between word and word, therefore this implementation cannot be embodied
Example obtains term vector using word2vec, can be similar between word and word to embody by calculating the distance between vector
Property.Word2Vec mainly uses two kinds of models of Skip-Gram and CBOW, and the present embodiment realizes term vector using Skip-Gram
Conversion, Skip-Gram model is mainly that the word of context is predicted by centre word.Specifically, target text is traversed first
Dictionary is constructed, the binary coding of each node is then generated according to the word frequency construction Hofman tree in dictionary, initializes each n omicronn-leaf
Term vector in the intermediate vector and leaf node of node finally obtains term vector after models for several times repetitive exercise.
S130, building are using coding and decoding as the Question-Answering Model of frame, using the term vector as the defeated of the Question-Answering Model
Enter and the Question-Answering Model is trained in a manner of loop iteration.
In one embodiment, a model framework, specially coding-decoded model (Encoder- end to end are constructed
Decoder), for applied to sequence to the model of sequence problem.Question-Answering Model is built using coding-decoded model as frame,
In Question-Answering Model, problem is list entries, and answer is output sequence, obtains one by the way that problem is input to coding in encoder
Then semantic feature vector is decoded to obtain answer by decoder to the semantic feature vector.It is specific as follows:
X={ x1、x2…xm}
Y={ y1、y2…ym}
C=F (x1、x2…xm)
Yi=g (C, y1、y2…yi-1)
Wherein, X is list entries, and Y is output sequence, and C is semantic feature vector, and encoder is carried out by list entries X
Coding, is translated into intermediate semantic feature vector C, and the semanteme for the expression context that then decoder is generated according to encoder is special
The output that sign vector sum previous moment obtains, which is decoded, generates yi.
In one embodiment, as shown in fig. 6, the step S130 may include step S131-S132.
S131, note is introduced using Recognition with Recurrent Neural Network as encoder and decoder and in the decoding stage of the decoder
Meaning power mechanism construction Question-Answering Model.
In one embodiment, encoder and decoder are used as using Recognition with Recurrent Neural Network (RNN), and added in decoding stage
Enter attention mechanism (Attention) and constitutes Question-Answering Model.Recognition with Recurrent Neural Network is commonly used to processing sequence data, has memory
Ability, allows the persistence of information, and the calculating of next node will rely on the calculated result of present node, can retain from the beginning
All calculated results.Therefore, the Question-Answering Model constituted using Recognition with Recurrent Neural Network, can fully consider contextual information, root
The answer of problem can be effectively predicted according to problem.However, traditional coding and decoding model is special by a fixed semanteme
Sign vector has some limitations to connect encoder and decoder, list entries entire sequence after encoder encodes
Information Compression into the information in the semantic feature vector of a specific length, causing not can completely to indicate entire list entries,
The content first inputted can be override by the content of rear input, lose many detailed information, especially in long sequence.Therefore, it is
Solution this problem introduces attention mechanism in decoder decoding stage, by attention mechanism breaks conventional codec-solution
Code device structure all relies on the limitation of an internal regular length vector in encoding and decoding.Attention mechanism is used for target data
Be weighted variation, by retain encoder to the intermediate output of list entries as a result, then by a Matching Model come pair
Centre output result carries out the study of selectivity, and is associated output sequence therewith in decoder output, wherein
Refer to calculating the model of similarity with model, generally speaking, the generating probability of each single item in output sequence is depended on defeated
Enter to have selected in sequence which.Specifically, list entries is encoded into multiple semantic feature vectors by encoder, is then calculated
The similarity of one output sequence and semantic feature vector distributes weight, selection and the most matched semantic feature of a upper output sequence to
It measures and is decoded to obtain current output sequence as the input of decoder, so as to adequately apply entrained by list entries
Information exports to generate.
S132, using the term vector as the input of the Question-Answering Model and according to the mode of loop iteration to the question and answer
Model is trained.
In one embodiment, after building Question-Answering Model, obtained term vector is input in Question-Answering Model and is instructed
Practice, specifically, term vector is input in encoder first, intermediate semantic feature is obtained according to the Recognition with Recurrent Neural Network of encoder
Then vector calculates the semantic similarity between intermediate semantic feature vector and a upper output, by semantic similarity it is highest in
Between input of the semantic feature vector as decoder, semantic feature vector is solved according to the Recognition with Recurrent Neural Network of decoder
Code, is equivalent to the inverse process of coding, finally obtains output, and the calculating parameter that loop iteration next time is used as after being exported continues
Training pattern, until the true answer of Model approximation.
If S140, the input for detecting problem, the answer of described problem is predicted by the Question-Answering Model after training
And export prediction result.
In one embodiment, when detecting that user goes out to put question to Claims Resolution customer service machine human hair, first by user's input
Problem segment by participle tool and obtains term vector according to word2vec, and the term vector of problem is input to Question-Answering Model
In, prediction is carried out to problem by Question-Answering Model and exports corresponding answer.In Question-Answering Model treatment process, fully considered with
Context in the dialog procedure of family, question-response that will not be mechanical, but context is combined to export more accurate reasonable answer.
The embodiment of the present invention illustrates a kind of intelligent answer method, collects corpus as target text by using predetermined manner
This；The target text pre-process and by term vector tool to the pretreated target text of progress converted with
Obtain term vector；Building is using coding and decoding as the Question-Answering Model of frame, using the term vector as the input of the Question-Answering Model
And the Question-Answering Model is trained in a manner of loop iteration；If detecting the input of problem, pass through the question and answer after training
Prediction result is predicted the answer of described problem and exported to model, can fully consider the context of question and answer, improves question and answer essence
Degree saves human resources and improves user experience.
Fig. 7 is a kind of schematic block diagram of intelligent answer device 200 provided in an embodiment of the present invention.As shown in fig. 7, corresponding
In the above intelligent answer method, the present invention also provides a kind of intelligent answer devices 200.The intelligent answer device 200 includes being used for
Execute the unit of above-mentioned intelligent answer method, the device can be configured in desktop computer, tablet computer, laptop computer, etc. eventually
In end.Specifically, referring to Fig. 7, the intelligent answer device 200 includes collector unit 210, pretreatment unit 220, construction unit
230 and predicting unit 240.
Collector unit 210, for collecting corpus as target text using predetermined manner.
In one embodiment, as shown in figure 8, the collector unit 210 includes subelement: crawler unit 211a and cleaning
Unit 212a.
Crawler unit 211a, for obtaining target webpage by way of web crawlers.
Cleaning unit 212a obtains target text for carrying out data cleansing to the target webpage.
In another embodiment, as shown in figure 9, the collector unit 210 includes subelement: acquiring unit 211A and the
Two converting unit 212A.
Acquiring unit 211A, for obtaining the calling record of Claims Resolution case from presetting database.
Second converting unit 212A, for the calling record to be converted into target text by speech tool.
Pretreatment unit 220, for pre-process and by term vector tool to being located in advance to the target text
Target text after reason is converted to obtain term vector.
In one embodiment, as shown in figure 8, the pretreatment unit 220 includes subelement: participle unit 221 and the
One converting unit 222.
Participle unit 221, for being segmented by participle tool to the target text；
First converting unit 222, for carrying out vector conversion to the participle in the target text according to term vector tool
Obtain term vector.
Construction unit 230 asks the term vector for constructing the Question-Answering Model using coding and decoding as frame as described in
It answers the input of model and the Question-Answering Model is trained in a manner of loop iteration.
In one embodiment, as shown in figure 8, the construction unit 230 includes subelement: building subelement 231 and instruction
Practice unit 232.
Construct subelement 231, for using Recognition with Recurrent Neural Network as encoder and decoder and in the decoder
Decoding stage introduces attention mechanism construction Question-Answering Model；
Training unit 232, for using the term vector as the input of the Question-Answering Model and according to the side of loop iteration
Formula is trained the Question-Answering Model.
Predicting unit 240, if being answered by the Question-Answering Model after training described problem for detecting the input of problem
Case is predicted and exports prediction result.
It should be noted that it is apparent to those skilled in the art that, above-mentioned 200 He of intelligent answer device
The specific implementation process of each unit can refer to the corresponding description in preceding method embodiment, for convenience of description and succinctly,
Details are not described herein.
The embodiment of the present invention illustrates a kind of intelligent answer device, collects corpus as target text by using predetermined manner
This；The target text pre-process and by term vector tool to the pretreated target text of progress converted with
Obtain term vector；Building is using coding and decoding as the Question-Answering Model of frame, using the term vector as the input of the Question-Answering Model
And the Question-Answering Model is trained in a manner of loop iteration；If detecting the input of problem, pass through the question and answer after training
Prediction result is predicted the answer of described problem and exported to model, can fully consider the context of question and answer, improves question and answer essence
Degree saves human resources and improves user experience.
Above-mentioned intelligent answer device can be implemented as a kind of form of computer program, which can such as scheme
It is run in computer equipment shown in 10.
Referring to Fig. 10, Figure 10 is a kind of schematic block diagram of computer equipment provided by the embodiments of the present application.The calculating
Machine equipment 500 can be terminal, be also possible to server, wherein terminal can be smart phone, tablet computer, notebook electricity
Brain, desktop computer, personal digital assistant and wearable device etc. have the electronic equipment of communication function.Server can be independence
Server, be also possible to the server cluster of multiple servers composition.
Refering to fig. 10, which includes processor 502, memory and the net connected by system bus 501
Network interface 505, wherein memory may include non-volatile memory medium 503 and built-in storage 504.
The non-volatile memory medium 503 can storage program area 5031 and computer program 5032.The computer program
5032 include program instruction, which is performed, and processor 502 may make to execute a kind of intelligent answer method.
The processor 502 is for providing calculating and control ability, to support the operation of entire computer equipment 500.
The built-in storage 504 provides environment for the operation of the computer program 5032 in non-volatile memory medium 503, should
When computer program 5032 is executed by processor 502, processor 502 may make to execute a kind of intelligent answer method.
The network interface 505 is used to carry out network communication with other equipment.It will be understood by those skilled in the art that in Figure 10
The structure shown, only the block diagram of part-structure relevant to application scheme, does not constitute and is applied to application scheme
The restriction of computer equipment 500 thereon, specific computer equipment 500 may include more more or fewer than as shown in the figure
Component perhaps combines certain components or with different component layouts.
Wherein, the processor 502 is for running computer program 5032 stored in memory, to realize following step
It is rapid: corpus is collected as target text using predetermined manner；The target text pre-process and by term vector tool
It is converted to pretreated target text is carried out to obtain term vector；It constructs using coding and decoding as the Question-Answering Model of frame,
It is trained using the term vector as the input of the Question-Answering Model and to the Question-Answering Model in a manner of loop iteration；If
The input for detecting problem predicts the answer of described problem by the Question-Answering Model after training and exports prediction result.
In one embodiment, processor 502 collects corpus as target text step in the realization use predetermined manner
When, it is implemented as follows step: obtaining target webpage by way of web crawlers；Data cleansing is carried out to the target webpage
Obtain target text.
In one embodiment, processor 502 collects corpus as target text step in the realization use predetermined manner
When, it is implemented as follows step: obtaining the calling record of Claims Resolution case from presetting database；By speech tool by institute
It states calling record and is converted into target text.
In one embodiment, processor 502 described pre-process the target text and passes through term vector realizing
When tool converts to obtain term vector step progress pretreated target text, it is implemented as follows step: passes through
Participle tool segments the target text；Vector is carried out to the participle in the target text according to term vector tool to turn
Get term vector in return
In one embodiment, processor 502 is realizing that the building, will be described using coding and decoding as the Question-Answering Model of frame
Term vector as the Question-Answering Model input and when being trained step to the Question-Answering Model in a manner of loop iteration, tool
Body realizes following steps: introducing using Recognition with Recurrent Neural Network as encoder and decoder and in the decoding stage of the decoder
Attention mechanism construction Question-Answering Model；Using the term vector as the input of the Question-Answering Model and according to the mode of loop iteration
The Question-Answering Model is trained.
It should be appreciated that in the embodiment of the present application, processor 502 can be central processing unit (Central
Processing Unit, CPU), which can also be other general processors, digital signal processor (Digital
Signal Processor, DSP), specific integrated circuit (Application Specific Integrated Circuit,
ASIC), ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic
Device, discrete gate or transistor logic, discrete hardware components etc..Wherein, general processor can be microprocessor or
Person's processor is also possible to any conventional processor etc..
Those of ordinary skill in the art will appreciate that be realize above-described embodiment method in all or part of the process,
It is that relevant hardware can be instructed to complete by computer program.The computer program includes program instruction, computer journey
Sequence can be stored in a storage medium, which is computer readable storage medium.The program instruction is by the department of computer science
At least one processor in system executes, to realize the process step of the embodiment of the above method.
Therefore, the present invention also provides a kind of storage mediums.The storage medium can be computer readable storage medium.This is deposited
Storage media is stored with computer program, and wherein computer program includes program instruction.The program instruction makes when being executed by processor
Processor executes following steps: collecting corpus as target text using predetermined manner；The target text is pre-processed
And it is converted by term vector tool to pretreated target text is carried out to obtain term vector；It constructs with coding and decoding and is
The Question-Answering Model of frame, to the question and answer using the term vector as the input of the Question-Answering Model and in a manner of loop iteration
Model is trained；If detecting the input of problem, the answer of described problem is predicted by the Question-Answering Model after training
And export prediction result.
In one embodiment, the processor is realized described using predetermined manner collection language in the instruction of execution described program
When material is as target text step, it is implemented as follows step: obtains target webpage by way of web crawlers；To the mesh
Mark webpage carries out data cleansing and obtains target text.
In one embodiment, the processor is realized described using predetermined manner collection language in the instruction of execution described program
When material is as target text step, it is implemented as follows step: obtaining the calling record of Claims Resolution case from presetting database；It is logical
It crosses speech tool and the calling record is converted into target text.
In one embodiment, the processor is realized described to target text progress in the instruction of execution described program
It pre-processes and passes through term vector tool to when carrying out pretreated target text and being converted to obtain term vector step, specifically
It realizes following steps: the target text being segmented by participle tool；According to term vector tool to the target text
In participle carry out vector be converted to term vector.
In one embodiment, the processor realizes the building using coding and decoding as frame executing described program instruction
The Question-Answering Model of frame, to the question and answer mould using the term vector as the input of the Question-Answering Model and in a manner of loop iteration
When type is trained step, it is implemented as follows step: using Recognition with Recurrent Neural Network as encoder and decoder and described
The decoding stage of decoder introduces attention mechanism construction Question-Answering Model；Using the term vector as the input of the Question-Answering Model
And the Question-Answering Model is trained according to the mode of loop iteration.
The storage medium can be USB flash disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), magnetic disk
Or the various computer readable storage mediums that can store program code such as CD.
Those of ordinary skill in the art may be aware that list described in conjunction with the examples disclosed in the embodiments of the present disclosure
Member and algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware
With the interchangeability of software, each exemplary composition and step are generally described according to function in the above description.This
A little functions are implemented in hardware or software actually, the specific application and design constraint depending on technical solution.Specially
Industry technical staff can use different methods to achieve the described function each specific application, but this realization is not
It is considered as beyond the scope of this invention.
In several embodiments provided by the present invention, it should be understood that disclosed device and method can pass through it
Its mode is realized.For example, the apparatus embodiments described above are merely exemplary.For example, the division of each unit, only
Only a kind of logical function partition, there may be another division manner in actual implementation.Such as multiple units or components can be tied
Another system is closed or is desirably integrated into, or some features can be ignored or not executed.
The steps in the embodiment of the present invention can be sequentially adjusted, merged and deleted according to actual needs.This hair
Unit in bright embodiment device can be combined, divided and deleted according to actual needs.In addition, in each implementation of the present invention
Each functional unit in example can integrate in one processing unit, is also possible to each unit and physically exists alone, can also be with
It is that two or more units are integrated in one unit.
If the integrated unit is realized in the form of SFU software functional unit and when sold or used as an independent product,
It can store in one storage medium.Based on this understanding, technical solution of the present invention is substantially in other words to existing skill
The all or part of part or the technical solution that art contributes can be embodied in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be a
People's computer, terminal or network equipment etc.) it performs all or part of the steps of the method described in the various embodiments of the present invention.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can readily occur in various equivalent modifications or replace
It changes, these modifications or substitutions should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with right
It is required that protection scope subject to.
Claims (10)
1. a kind of intelligent answer method characterized by comprising
Corpus is collected as target text using predetermined manner；
Pretreatment is carried out to the target text and is converted by term vector tool to pretreated target text is carried out
To obtain term vector；
Building is using coding and decoding as the Question-Answering Model of frame, using the term vector as the input of the Question-Answering Model and with circulation
The mode of iteration is trained the Question-Answering Model；
If detecting the input of problem, the answer of described problem is predicted by the Question-Answering Model after training and exports prediction
As a result.
2. intelligent answer method according to claim 1, which is characterized in that described to collect corpus conduct using predetermined manner
Target text includes:
Target webpage is obtained by way of web crawlers；
Data cleansing is carried out to the target webpage and obtains target text.
3. intelligent answer method according to claim 1, which is characterized in that described to collect corpus conduct using predetermined manner
Target text includes:
The calling record of Claims Resolution case is obtained from presetting database；
The calling record is converted into target text by speech tool.
4. intelligent answer method according to claim 1, which is characterized in that described to be pre-processed to the target text
And it is converted by term vector tool to pretreated target text is carried out to obtain term vector, comprising:
The target text is segmented by participle tool；
Vector is carried out to the participle in the target text according to term vector tool and is converted to term vector.
5. intelligent answer method according to claim 1, which is characterized in that building the asking using coding and decoding as frame
Model is answered, is carried out using the term vector as the input of the Question-Answering Model and to the Question-Answering Model in a manner of loop iteration
Training, comprising:
Attention mechanism is introduced using Recognition with Recurrent Neural Network as encoder and decoder and in the decoding stage of the decoder
Construct Question-Answering Model；
The Question-Answering Model is carried out using the term vector as the input of the Question-Answering Model and according to the mode of loop iteration
Training.
6. a kind of intelligent answer device characterized by comprising
Collector unit, for collecting corpus as target text using predetermined manner；
Pretreatment unit, for pre-process and by term vector tool to the pretreated mesh of progress to the target text
Mark text is converted to obtain term vector；
Construction unit, for constructing the Question-Answering Model using coding and decoding as frame, using the term vector as the Question-Answering Model
Input and the Question-Answering Model is trained in a manner of loop iteration；
Predicting unit, if being carried out by the Question-Answering Model after training to the answer of described problem for detecting the input of problem
It predicts and exports prediction result.
7. intelligent answer device according to claim 1 characterized by comprising
Participle unit, for being segmented by participle tool to the target text；
First converting unit, for according to term vector tool in the target text participle carry out vector be converted to word to
Amount.
8. intelligent answer device according to claim 1 characterized by comprising
Construct subelement, for using Recognition with Recurrent Neural Network as encoder and decoder and in the decoding stage of the decoder
Introduce attention mechanism construction Question-Answering Model；
Training unit, for using the term vector as the input of the Question-Answering Model and according to the mode of loop iteration to described
Question-Answering Model is trained.
9. a kind of computer equipment, which is characterized in that the computer equipment includes memory and processor, on the memory
It is stored with computer program, the processor is realized as described in any one of claim 1-5 when executing the computer program
Method.
10. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has computer journey
Sequence, the computer program can realize method according to any one of claims 1 to 5 when being executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910043210.1A CN109933652A (en) | 2019-01-17 | 2019-01-17 | Intelligent answer method, apparatus, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910043210.1A CN109933652A (en) | 2019-01-17 | 2019-01-17 | Intelligent answer method, apparatus, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109933652A true CN109933652A (en) | 2019-06-25 |
Family
ID=66985112
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910043210.1A Pending CN109933652A (en) | 2019-01-17 | 2019-01-17 | Intelligent answer method, apparatus, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109933652A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110502623A (en) * | 2019-07-08 | 2019-11-26 | 平安科技（深圳）有限公司 | Intelligent answer method, electronic device, computer equipment and readable storage medium storing program for executing |
| CN110597975A (en) * | 2019-10-18 | 2019-12-20 | 淄博矿业集团有限责任公司 | Coal mine material management dialogue method, system and all-in-one machine based on Self-extension mechanism |
| CN110765246A (en) * | 2019-09-29 | 2020-02-07 | 平安直通咨询有限公司上海分公司 | Question answering method and device based on intelligent robot, storage medium and intelligent equipment |
| CN110825855A (en) * | 2019-09-18 | 2020-02-21 | 平安科技（深圳）有限公司 | Response method and device based on artificial intelligence, computer equipment and storage medium |
| CN111061847A (en) * | 2019-11-22 | 2020-04-24 | 中国南方电网有限责任公司 | Dialogue generation and corpus expansion method and device, computer equipment and storage medium |
| CN111368064A (en) * | 2020-03-26 | 2020-07-03 | 平安医疗健康管理股份有限公司 | Survey information processing method, device, equipment and storage medium |
| CN111930914A (en) * | 2020-08-14 | 2020-11-13 | 工银科技有限公司 | Question generation method and device, electronic equipment and computer-readable storage medium |
| CN112257434A (en) * | 2019-07-02 | 2021-01-22 | Tcl集团股份有限公司 | Unmanned aerial vehicle control method, system, mobile terminal and storage medium |
| CN113569572A (en) * | 2021-02-09 | 2021-10-29 | 腾讯科技（深圳）有限公司 | Text entity generation method, model training method and device |
| CN114416213A (en) * | 2022-03-29 | 2022-04-29 | 北京沃丰时代数据科技有限公司 | Word vector file loading method and device and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107832310A (en) * | 2017-11-27 | 2018-03-23 | 首都师范大学 | Structuring argument generation method and system based on seq2seq models |
| US20180174020A1 (en) * | 2016-12-21 | 2018-06-21 | Microsoft Technology Licensing, Llc | Systems and methods for an emotionally intelligent chat bot |
| CN108345585A (en) * | 2018-01-11 | 2018-07-31 | 浙江大学 | A kind of automatic question-answering method based on deep learning |
| CN108763284A (en) * | 2018-04-13 | 2018-11-06 | 华南理工大学 | A kind of question answering system implementation method based on deep learning and topic model |
-
2019
- 2019-01-17 CN CN201910043210.1A patent/CN109933652A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180174020A1 (en) * | 2016-12-21 | 2018-06-21 | Microsoft Technology Licensing, Llc | Systems and methods for an emotionally intelligent chat bot |
| CN107832310A (en) * | 2017-11-27 | 2018-03-23 | 首都师范大学 | Structuring argument generation method and system based on seq2seq models |
| CN108345585A (en) * | 2018-01-11 | 2018-07-31 | 浙江大学 | A kind of automatic question-answering method based on deep learning |
| CN108763284A (en) * | 2018-04-13 | 2018-11-06 | 华南理工大学 | A kind of question answering system implementation method based on deep learning and topic model |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112257434A (en) * | 2019-07-02 | 2021-01-22 | Tcl集团股份有限公司 | Unmanned aerial vehicle control method, system, mobile terminal and storage medium |
| CN112257434B (en) * | 2019-07-02 | 2023-09-08 | Tcl科技集团股份有限公司 | Unmanned aerial vehicle control method, unmanned aerial vehicle control system, mobile terminal and storage medium |
| CN110502623A (en) * | 2019-07-08 | 2019-11-26 | 平安科技（深圳）有限公司 | Intelligent answer method, electronic device, computer equipment and readable storage medium storing program for executing |
| CN110825855A (en) * | 2019-09-18 | 2020-02-21 | 平安科技（深圳）有限公司 | Response method and device based on artificial intelligence, computer equipment and storage medium |
| WO2021051615A1 (en) * | 2019-09-18 | 2021-03-25 | 平安科技（深圳）有限公司 | Response method and apparatus based on artificial intelligence, computer device, and storage medium |
| CN110765246A (en) * | 2019-09-29 | 2020-02-07 | 平安直通咨询有限公司上海分公司 | Question answering method and device based on intelligent robot, storage medium and intelligent equipment |
| CN110597975A (en) * | 2019-10-18 | 2019-12-20 | 淄博矿业集团有限责任公司 | Coal mine material management dialogue method, system and all-in-one machine based on Self-extension mechanism |
| CN111061847A (en) * | 2019-11-22 | 2020-04-24 | 中国南方电网有限责任公司 | Dialogue generation and corpus expansion method and device, computer equipment and storage medium |
| CN111368064B (en) * | 2020-03-26 | 2023-04-07 | 深圳平安医疗健康科技服务有限公司 | Survey information processing method, device, equipment and storage medium |
| CN111368064A (en) * | 2020-03-26 | 2020-07-03 | 平安医疗健康管理股份有限公司 | Survey information processing method, device, equipment and storage medium |
| CN111930914A (en) * | 2020-08-14 | 2020-11-13 | 工银科技有限公司 | Question generation method and device, electronic equipment and computer-readable storage medium |
| CN111930914B (en) * | 2020-08-14 | 2023-09-12 | 中国工商银行股份有限公司 | Problem generation method and device, electronic equipment and computer readable storage medium |
| CN113569572A (en) * | 2021-02-09 | 2021-10-29 | 腾讯科技（深圳）有限公司 | Text entity generation method, model training method and device |
| CN113569572B (en) * | 2021-02-09 | 2024-05-24 | 腾讯科技（深圳）有限公司 | Text entity generation method, model training method and device |
| CN114416213A (en) * | 2022-03-29 | 2022-04-29 | 北京沃丰时代数据科技有限公司 | Word vector file loading method and device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109933652A (en) | Intelligent answer method, apparatus, computer equipment and storage medium | |
| CN112000791B (en) | Motor fault knowledge extraction system and method | |
| CN109657051A (en) | Text snippet generation method, device, computer equipment and storage medium | |
| CN111143576A (en) | Event-oriented dynamic knowledge graph construction method and device | |
| CN113127624B (en) | Question-answer model training method and device | |
| CN111783394A (en) | Training method of event extraction model, event extraction method, system and equipment | |
| CN111324721A (en) | Method for constructing intelligent question-answering knowledge base | |
| CN117349275B (en) | Text structuring method and system based on large language model | |
| CN111428104B (en) | Epilepsy auxiliary medical intelligent question-answering method based on viewpoint type reading understanding | |
| CN113159187B (en) | Classification model training method and device and target text determining method and device | |
| CN109918477A (en) | A kind of distributed search resources bank selection method based on variation self-encoding encoder | |
| CN116578688A (en) | Text processing method, device, equipment and storage medium based on multiple rounds of questions and answers | |
| CN117708274A (en) | Search question-answering system based on large model, method and electronic equipment thereof | |
| KR102418260B1 (en) | Method for analyzing customer consultation record | |
| CN117828024A (en) | Plug-in retrieval method, device, storage medium and equipment | |
| Muangnak et al. | The neural network conversation model enables the commonly asked student query agents | |
| CN116702743A (en) | Text similarity detection method and device, electronic equipment and storage medium | |
| CN116561284A (en) | Intelligent response method, device, electronic equipment and medium | |
| CN116578697A (en) | Finance-oriented language emotion analysis and labeling method | |
| CN116432662A (en) | Training method of text processing model, text processing method and device | |
| CN116645961A (en) | Speech recognition method, speech recognition device, electronic apparatus, and storage medium | |
| CN114611529B (en) | Intention recognition method and device, electronic equipment and storage medium | |
| CN115310429A (en) | Data compression and high-performance calculation method in multi-turn listening dialogue model | |
| CN114970733A (en) | Corpus generation method, apparatus, system, storage medium and electronic device | |
| CN115270746A (en) | Question sample generation method and device, electronic equipment and storage medium |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190625 |