CN109446314A - A kind of customer service question processing method and device - Google Patents
A kind of customer service question processing method and device Download PDFInfo
- Publication number
- CN109446314A CN109446314A CN201811354975.9A CN201811354975A CN109446314A CN 109446314 A CN109446314 A CN 109446314A CN 201811354975 A CN201811354975 A CN 201811354975A CN 109446314 A CN109446314 A CN 109446314A
- Authority
- CN
- China
- Prior art keywords
- information
- answer
- similarity
- text
- text information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The embodiment of the invention provides a kind of customer service question processing method and devices, this method remembers LSTM neural network tensor model by the way that extracted keyword is input to shot and long term trained in advance, obtain the feature of problem information, determine the text information in preset text information with obtained characteristic matching, and according to extracted feature, similarity information in computational problem information and determined text information between question text information, using the answer text information in the highest text information of similarity as the answer information of problem information.It can be improved the accuracy rate of customer service issue handling using scheme provided in an embodiment of the present invention.
Description
Technical field
The present invention relates to technical field of information processing, more particularly to a kind of customer service question processing method and device.
Background technique
Currently, customer service robot, mainly using traditional natural language processing techniques processing customer problem.Its processing side
Method is main are as follows: firstly, the keyword of the problem of extraction；Secondly, the problem answer corresponding with each keyword is clustered, and
Determine that the problem of matching with all keywords replies in cluster result；Then, to the calculating above problem and identified problem
Similarity between answer；Finally, being back to user for replying the problem of similarity highest.
However, in practical applications, due to the diversity of language expression, the difficulty of semantic understanding, especially Chinese discrimination
Phenomena such as adopted, is very common, and it is relatively low to result in the problem of handling acquisition using natural language processing technique answer accuracy rate.
Summary of the invention
The embodiment of the present invention is designed to provide a kind of customer service question processing method and device, can be improved customer service problem
The accuracy rate of processing.
Specific technical solution is as follows:
A kind of customer service question processing method, which comprises
Obtain problem information；
Extract the keyword of described problem information；
Extracted keyword is input to shot and long term trained in advance and remembers LSTM neural network tensor model, obtains institute
State the feature of problem information, wherein the LSTM neural network tensor model is: utilizing the keyword pair of sample problem information
The model that LSTM neural network obtains after being trained；
Determine the text information in preset text information with obtained characteristic matching；Wherein, the preset text
Information is: the information summarized to preset question text information sequence and preset answer text information sequence；
According to extracted feature, calculate in described problem information and determined text information between question text information
Similarity information；
Using the answer text information in the highest text information of similarity as the answer information of described problem information.
Further, in the answer text information using in the highest text information of similarity as described problem information
Answer information before, further includes:
Judge in the similarity information being calculated with the presence or absence of the similarity information greater than default value；
If there is the similarity information for being greater than default value, the solution by the highest text information of similarity is executed
The step of answering answer information of the text information as described problem information.
Further, with the presence or absence of the similarity greater than default value in the similarity information for judging to be calculated
After information, further includes:
If there is no the similarity information for being greater than default value, described problem information is sent to client, so that
The client shows described problem information to contact staff；
Receive the answer information of contact staff's input that the client is sent, described；
Answer information to be output is determined from answer information obtained, and exports the answer information to be output.
Further, after the answer information for receiving contact staff's input that the client is sent, described, also
Include:
Feedback information is sent to destination client, wherein the feedback information, for prompting contact staff to receive
The answer information of described problem information, the destination client are as follows: the client of not sent answer information.
A kind of customer service issue handling device, described device include:
Problem obtains module, for obtaining problem information；
Keyword-extraction module, for extracting the keyword of described problem information；
Characteristic extracting module remembers LSTM nerve net for extracted keyword to be input to shot and long term trained in advance
Network tensor model, obtains the feature of described problem information, wherein the LSTM neural network tensor model is: being asked using sample
The model that the keyword of topic information obtains after being trained to LSTM neural network；
First determining module, for determining the text information in preset text information with obtained characteristic matching；Its
In, the preset text information is: carrying out to preset question text information sequence and preset answer text information sequence
The information summarized；
Computing module, for calculating problem in described problem information and determined text information according to extracted feature
Similarity information between text information；
Second determining module, for believing the answer text information in the highest text information of similarity as described problem
The answer information of breath.
Further, described device further include:
Judgment module is believed with the presence or absence of the similarity greater than default value in the similarity information for judging to be calculated
Breath；If there is the similarity information for being greater than default value, second determining module is triggered.
Further, described device further include:
If there is no the similarity information for being greater than default value, sending module is triggered；
The sending module, for described problem information to be sent to client, so that the client is to customer service people
Member shows described problem information；
Information receiving module is answered, for receiving the answer information of contact staff's input that the client is sent, described；
Message output module is answered, for determining answer information to be output from answer information obtained, and is exported
The answer information to be output.
Further, described device further include:
Feedback information sending module, for feedback information to be sent to destination client, wherein the feedback information is used
The answer information of described problem information, the destination client are as follows: not sent answer information have been received in prompt contact staff
Client.
At the another aspect implemented of the present invention, additionally provide a kind of electronic equipment, including processor, communication interface, motor,
Memory and communication bus, wherein processor, communication interface, memory complete mutual communication by communication bus；
Memory, for storing computer program；
Processor when for executing the program stored on memory, executes any of the above-described customer service issue handling
Method.
At the another aspect that the present invention is implemented, a kind of computer readable storage medium is additionally provided, it is described computer-readable
It is stored with instruction in storage medium, when run on a computer, executes any of the above-described customer service question processing method.
A kind of customer service question processing method and device provided in an embodiment of the present invention, this method include by extracted key
Word is input to shot and long term memory LSTM trained in advance (Long Short-Term Memory, shot and long term memory) neural network
Model is measured, the feature of problem information is obtained, determines the text information in preset text information with obtained characteristic matching, and
According to extracted feature, similarity information in computational problem information and determined text information between question text information,
Using the answer text information in the highest text information of similarity as the answer information of problem information.Compared with the existing technology,
When handling customer service problem using scheme provided in an embodiment of the present invention, by the keyword input LSTM mind of extracted problem information
The feature of problem information is obtained through network tensor model, the determining text with obtained characteristic matching from preset text information
This information, and include miscellaneous question text information and answer text information, therefore, application in preset text information
When scheme processing customer service problem provided in an embodiment of the present invention, the accuracy rate of customer service issue handling can be improved.Certainly, implement this
Any product or method of invention must be not necessarily required to reach all the above advantage simultaneously.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is the flow chart of the first customer service question processing method of the embodiment of the present invention；
Fig. 2 is the flow chart of second of customer service question processing method provided in an embodiment of the present invention；
Fig. 3 is a kind of structural schematic diagram of customer service issue handling device provided in an embodiment of the present invention；
Fig. 4 is the structural schematic diagram of a kind of electronic 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 only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Referring to Fig. 1, Fig. 1 is the flow chart of the first customer service question processing method provided in an embodiment of the present invention, specific to wrap
It includes:
S101 obtains problem information；
Wherein, problem information can be understood as the problem of user proposes customer service robot.
In addition, problem information can be the complete problem of user's input, it is also possible to lead to the complete problem that user inputs
Cross the information after semantic conversion software is converted；It can also be and obtained after the problem of proposing user speech carries out text conversion
Text information, and the information after being converted to text information by semantic conversion software.
Based on above content, a kind of implementation for obtaining problem information can be with are as follows: the problem of obtaining user's input text,
The keyword of above problem text is extracted, and above problem information is generated according to preset rules to the keyword of extraction.
Wherein, preset rules can be semantic rules.The keyword extraction of question text can be according to deep learning side
Method is extracted to obtain.
As it can be seen that the implementation generates problem information according to preset rules using the keyword of above problem text, improve
The accuracy of robot identification question text, more further improves the accuracy of robot processing customer service problem.
For example, the complete problem that user proposes are as follows: I has had purchased dress, but I repeats to have bought one again
How the clothes of part striking resemblances, return goods? can problem information be with are as follows: how repeat buying clothes returns goods?
S102 extracts the keyword of described problem information；
Wherein, the keyword of problem information can be the position first appeared in problem information using word as spy
Sign, mainly for the corpus of total score general construction.Appear in problem information stem and tail portion word become the probability of keyword compared with
Greatly, thus according to each word problem information for the first time or last bit occur position assign different weights, tie on this basis
Closing TF-IDF, (term frequency-inverse document frequency is a kind of for information retrieval and data digging
Weighting technique is commonly used in pick) and continuous data discretization carry out keyword extraction.
For example, based on above-mentioned example, the keyword for extracting above problem information can be with are as follows: repetition, clothes, purchase
And the return of goods.
Extracted keyword is input to shot and long term trained in advance and remembers LSTM neural network tensor model by S103,
Obtain the feature of described problem information, wherein the LSTM neural network tensor model is the key that: utilizing sample problem information
The model that word obtains after being trained to LSTM neural network；
Above-mentioned LSTM neural network is a kind of time recurrent neural network, is suitable for being spaced in processing and predicted time sequence
The relatively long critical event with delay.
Traditional neural network can not accomplish persistent memory for above-mentioned LSTM neural network.It is assumed that utilizing
Neural network classifies to the event of each time point in film, it is evident that traditional neural network cannot use previous thing
Part goes the next event of reasoning.And above-mentioned LSTM neural network can solve this problem.Therefore, which is band
There is the neural network of circulation, information is allowed to retain a period of time.
LSTM neural network is trained to obtain LSTM neural network tensor mould using the keyword of sample problem information
The detailed process of type may include:
Obtain sample problem information；
The keyword of sample problem information is extracted, and obtains the feature of sample problem information, as sample characteristics；
LSTM neural network is trained using extracted keyword, the sample for obtaining the output of LSTM neural network is asked
Inscribe the feature of information；
Obtained feature and above-mentioned sample characteristics are compared, and above-mentioned LSTM nerve net is adjusted according to comparing result
The network parameter of network so far completes the process being trained using the keyword of sample problem information to LSTM neural network.
For example, based on above-mentioned example, keyword i.e. repetition, clothes, the return of goods and the purchase of problem information is extracted, benefit
With the feature of above-mentioned LSTM neural network tensor model output above problem information, that is, repeats, buys and return goods.
S104 determines the text information in preset text information with obtained characteristic matching；Wherein, described preset
Text information is: the letter summarized to preset sequence of question text information and preset answer sequence text information
Breath；
Wherein, have between the answer text information in the problems in text information text information and text information specific
Mapping relations one by one.
In addition, since preset content of text messages is more, how to be obtained from a large amount of content of text messages and gained
The text information of the characteristic matching arrived, a kind of implementation can be with are as follows: obtain the feature of above-mentioned text information, calculate obtained
The similarity between aforementioned extracted feature, the similarity that can will be greater than preset threshold correspond to above-mentioned text envelope to feature respectively
Breath is as the text information to match with above extracted feature.
Wherein, the similarity between feature can be calculated using Euclidean distance or cosine similarity mode.
For example, it is based on above-mentioned example, includes: in preset text information
Is problem 1: how repeat buying handled? problem replies: you can choose the return of goods；
Problem 2: how clothes is bought? problem replies: determining a clothes to be bought, shopping cart is added, walk according to corresponding
Rapid payment；
Does problem 3: how repeat buying return goods? problem replies: selecting " return of goods " from my order, clicks after returning goods, ask
According to prompt input return of goods address and contact method；
Problem 4: how goods is sold? problem replies: please send a telegraph 400300209, be responsible for people specially and service to you.
The each text information for including by above-mentioned preset text information matches extracted feature, can obtain it is multiple with
Text information, that is, the problem 1 and problem 3 of obtained characteristic matching.
S105 calculates question text information in described problem information and determined text information according to extracted feature
Between similarity information；
A kind of implementation that similarity information is calculated in above-mentioned S105 can be with are as follows: according to extracted feature, calculates institute
Determine total number of characters of features described above in the presence of question text information in text information divided by total number of characters of the problem information.
For example, it is based on above-mentioned example, above-mentioned problem 1 and problem 3 and features described above is calculated separately and " repeats, purchase
Buy and return goods " between similarity, respectively obtain 57% and 75%.
S106 believes the answer text information in the highest text information of similarity as the answer of described problem information
Breath.
In this step, according to one of the answer text information in the problems in text information text information and text information
Answer text information in the highest text information of similarity is then determined as the answer text envelope of problem information by one mapping relations
Breath, and export determining answer text information.
For example, it is based on above-mentioned example, according to the result of calculating, it is seen that the similarity of problem 3 and extracted feature
Highest, then problem 3 is exactly that problem corresponding to identified text information output problem 3 replies, that is to say, that output is " from mine
" return of goods " are selected in order, are clicked after returning goods, please according to prompt input return of goods address and contact method ".
A kind of implementation further includes step A before S106:
Step A judges in the similarity information being calculated with the presence or absence of the similarity information greater than default value；If
In the presence of the similarity information for being greater than default value, S106 is executed；
In this step, determine that text information determined by S104 is answered as the problem of problem information by default value
It is multiple.It is greater than the similarity information of default value if it exists, then it is assumed that the problem of similarity information corresponds to text information replies can
It is replied as the problem of problem information.
Judge further to mention in similarity with the presence or absence of the similarity greater than default value as it can be seen that this implementation passes through
The accuracy rate that high problem information replies.
In another implementation, it is based on step A, further includes step B~step D:
Described problem information is sent to client if there is no the similarity information for being greater than default value by step B,
So that the client shows described problem information to contact staff；
In this step, if there is no be greater than preset similarity information in identified similarity information, it is meant that,
The accuracy of determining text information is lower, at this time, it may be necessary to using client to profession contact staff's showing problem information, with
Obtain the more accurate answer information that contact staff provides.
It should be noted that client is at least one.
Step C receives the answer information of contact staff's input that the client is sent, described；
In this step, for problem information, if client is one, the answer letter of client transmission is received
Breath；If client be it is multiple, receive multiple client transmission contact staff input answer information.
Step D determines answer information to be output from answer information obtained, and exports the answer to be output
Information.
If receiving the answer information of client transmission, using the answer information as determining solution to be output
Information is answered, and exports the answer information.If receiving the answer information of multiple client transmission, need to believe from multiple answers
An answer information to be output is determined in breath, and exports identified answer information.
As it can be seen that this implementation passes through for there is no the similarity informations greater than default value, by described problem information
It is sent to client, and determines answer information to be output from the answer information that acquired client is sent, output institute is really
Fixed answer information；The answer information that the implementation inputs contact staff is exported as the answer information of problem information, into
One step improves the accuracy rate of customer service issue handling, so as to bring good experience effect for user.
In one implementation, the answer information that client is sent is received in order to reduce repetition, after step c, also
May include step E:
Feedback information is sent to destination client by step E, wherein the feedback information, for having prompted contact staff
Through receiving the answer information of described problem information, the destination client are as follows: the client of not sent answer information.
In this step, a part of contact staff has input answer information in the client in time, receives client and sends
The part contact staff input answer information, at this point it is possible to which being interpreted as can be true from the answer information received
The answer information for determining problem information, without the answer information in waiting other contact staff input, in order to reduce contact staff's
Feedback information is then sent to the contact staff for information of not answering a question by repeated work, so that contact staff sees feedback information
Afterwards, without in the answer information for retransmitting the problem information.
As it can be seen that this implementation, which by the way that feedback information is sent to destination client, can not only reduce repetition, receives visitor
The answer information that family end is sent, additionally it is possible to reduce the workload of contact staff.
It can be seen that keyword of the method provided in an embodiment of the present invention by extraction problem information, by extracted pass
Key word is input in advance trained shot and long term memory LSTM neural network tensor model, obtains the feature of problem information, and according to
Extracted feature, computational problem information and from the text information determined in preset text information between question text information
Similarity information answered using the answer text information in the highest text information of similarity as the answer information of problem information
It can be improved the accuracy rate of customer service issue handling with method provided in an embodiment of the present invention.
Referring to fig. 2, the embodiment of the invention provides the flow chart of second of customer service question processing method, this method packets by Fig. 2
It includes:
S201 obtains problem information；
S202 extracts the keyword of described problem information；
Extracted keyword is input to shot and long term trained in advance and remembers LSTM neural network tensor model by S203,
Obtain the feature of described problem information, wherein the LSTM neural network tensor model is the key that: utilizing sample problem information
The model that word obtains after being trained to LSTM neural network；
S204 determines the text information in preset text information with obtained characteristic matching；Wherein, described preset
Text information is: the letter summarized to preset sequence of question text information and preset answer sequence text information
Breath；
S205 calculates question text information in described problem information and determined text information according to extracted feature
Between similarity information；
S206 judges in the similarity information being calculated with the presence or absence of the similarity information greater than default value；If
In the presence of the similarity information for being greater than default value, S207 is executed；If there is no the similarity information for being greater than default value, execute
S208~S210；
S207 believes the answer text information in the highest text information of similarity as the answer of described problem information
Breath；
Described problem information is sent to client by S208, so that the client is asked to described in contact staff's displaying
Inscribe information；
S209 receives the answer information of contact staff's input that the client is sent, described；
S210 determines answer information to be output from answer information obtained, and exports the answer to be output
Information.
It can be seen that method provided in an embodiment of the present invention, which is directed to exist in the similarity information calculated, is greater than default value
Similarity information, using the answer text information in the highest text information of similarity as the answer information of problem information, needle
To in the similarity information of calculating, there is no the similarity informations greater than default value, then the answer information inputted from contact staff
The determining answer information of middle output, can not only further increase the accuracy rate of customer service issue handling, and can for user with
Carry out good experience effect.
Corresponding with above-mentioned customer service question processing method, the embodiment of the invention also provides a kind of customer service issue handling dresses
It sets.
Referring to Fig. 3, Fig. 3 provides a kind of structural schematic diagram of customer service issue handling device, the dress for present invention implementation
It sets and includes:
Problem obtains module 301, for obtaining problem information；
Keyword-extraction module 302, for extracting the keyword of described problem information；
Characteristic extracting module 303, for extracted keyword to be input to shot and long term memory LSTM mind trained in advance
Through network tensor model, the feature of described problem information is obtained, wherein the LSTM neural network tensor model is: utilizing sample
The model that the keyword of this problem information obtains after being trained to LSTM neural network；
First determining module 304, for determining the text information in preset text information with obtained characteristic matching；
Wherein, the preset text information is: to preset sequence of question text information and preset answer sequence text information into
The information that row summarizes；
Computing module 305, for calculating and being asked in described problem information and determined text information according to extracted feature
Inscribe the similarity information between text information；
Second determining module 306, for asking the answer text information in the highest text information of similarity as described in
Inscribe the answer information of information.
In one implementation, described device can also include:
Judgment module is believed with the presence or absence of the similarity greater than default value in the similarity information for judging to be calculated
Breath；If there is the similarity information for being greater than default value, the second determining module 306 is triggered.
In one implementation, described device can also include:
If there is no the similarity information for being greater than default value, sending module is triggered；
The sending module, for described problem information to be sent to client, so that the client is to customer service people
Member shows described problem information；
Information receiving module is answered, for receiving the answer information of contact staff's input that the client is sent, described；
Message output module is answered, for determining answer information to be output from answer information obtained, and is exported
The answer information to be output.
In one implementation, described device can also include:
Feedback information sending module, for feedback information to be sent to destination client, wherein the feedback information is used
The answer information of described problem information, the destination client are as follows: not sent answer information have been received in prompt contact staff
Client.
It can be seen that keyword of the device provided in an embodiment of the present invention by extraction problem information, by extracted pass
Key word is input in advance trained shot and long term memory LSTM neural network tensor model, obtains the feature of problem information, and according to
Extracted feature, computational problem information and from the text information determined in preset text information between question text information
Similarity information answered using the answer text information in the highest text information of similarity as the answer information of problem information
It can be improved the accuracy rate of customer service issue handling with method provided in an embodiment of the present invention.
The embodiment of the invention also provides a kind of electronic equipment, as shown in figure 4, electronic equipment includes processor 401, communication
Interface 402, memory 403 and communication bus 404, wherein processor 401, communication interface 402, memory 403 are total by communication
Line 404 completes mutual communication,
Memory 403, for storing computer program；
Processor 401 when for executing the program stored on memory 403, realizes provided in an embodiment of the present invention one
Kind customer service question processing method.
Specifically, a kind of above-mentioned customer service question processing method, this method comprises:
Obtain problem information；
Extract the keyword of described problem information；
Extracted keyword is input to shot and long term trained in advance and remembers LSTM neural network tensor model, obtains institute
State the feature of problem information, wherein the LSTM neural network tensor model is: utilizing the keyword pair of sample problem information
The model that LSTM neural network obtains after being trained；
Determine the text information in preset text information with obtained characteristic matching；Wherein, the preset text
Information is: the information summarized to preset question text information sequence and preset answer text information sequence；
According to extracted feature, calculate in described problem information and determined text information between question text information
Similarity information；
Using the answer text information in the highest text information of similarity as the answer information of described problem information.
It can be seen that electronic equipment provided in this embodiment is executed, by the keyword by extracting problem information, by institute
The keyword of extraction is input to shot and long term memory LSTM neural network tensor model trained in advance, obtains the spy of problem information
Sign, and according to extracted feature, computational problem information and the problem text from the text information determined in preset text information
Similarity information between this information, using the answer text information in the highest text information of similarity as the solution of problem information
Information is answered, can be improved the accuracy rate of customer service issue handling using method provided in an embodiment of the present invention.
The visitor that the embodiment of above-mentioned related content customer service question processing method and preceding method embodiment part provide
Take that issue handling mode is identical, and which is not described herein again.
The communication bus that above-mentioned electronic equipment is mentioned can be Peripheral Component Interconnect standard (Peripheral Component
Interconnect, PCI) bus or expanding the industrial standard structure (Extended Industry Standard
Architecture, EISA) bus etc..The communication bus can be divided into address bus, data/address bus, control bus etc..For just
It is only indicated with a thick line in expression, figure, it is not intended that an only bus or a type of bus.
Communication interface is for the communication between above-mentioned electronic equipment and other equipment.
Memory may include random access memory (Random Access Memory, RAM), also may include non-easy
The property lost memory (Non-Volatile Memory, NVM), for example, at least a magnetic disk storage.Optionally, memory may be used also
To be storage device that at least one is located remotely from aforementioned processor.
Above-mentioned processor can be general processor, including central processing unit (Central Processing Unit,
CPU), network processing unit (Network Processor, NP) etc.；It can also be digital signal processor (Digital Signal
Processing, DSP), it is specific integrated circuit (Application Specific Integrated Circuit, ASIC), existing
It is field programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic device, discrete
Door or transistor logic, discrete hardware components.
In another embodiment provided by the invention, a kind of computer readable storage medium is additionally provided, which can
It reads to be stored with instruction in storage medium, when run on a computer, so that computer executes any institute in above-described embodiment
The customer service question processing method stated.
In another embodiment provided by the invention, a kind of computer program product comprising instruction is additionally provided, when it
When running on computers, so that computer executes any customer service question processing method in above-described embodiment.
In the above-described embodiments, can come wholly or partly by software, hardware, firmware or any combination thereof real
It is existing.When implemented in software, it can entirely or partly realize in the form of a computer program product.The computer program
Product includes one or more computer instructions.When loading on computers and executing the computer program instructions, all or
It partly generates according to process or function described in the embodiment of the present invention.The computer can be general purpose computer, dedicated meter
Calculation machine, computer network or other programmable devices.The computer instruction can store in computer readable storage medium
In, or from a computer readable storage medium to the transmission of another computer readable storage medium, for example, the computer
Instruction can pass through wired (such as coaxial cable, optical fiber, number from a web-site, computer, server or data center
Equity investment people line (DSL)) or wireless (such as infrared, wireless, microwave etc.) mode to another web-site, computer, clothes
Business device or data center are transmitted.The computer readable storage medium can be any available Jie that computer can access
Matter either includes the data storage devices such as one or more usable mediums integrated server, data center.Described use is situated between
Matter can be magnetic medium, (for example, floppy disk, hard disk, tape), optical medium (for example, DVD) or semiconductor medium (such as it is solid
State hard disk Solid State Disk (SSD)) etc..
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
Each embodiment in this specification is all made of relevant mode and describes, same and similar portion between each embodiment
Dividing may refer to each other, and each embodiment focuses on the differences from other embodiments.Especially for device,
For electronic equipment, computer readable storage medium and program product embodiment, since it is substantially similar to the method embodiment, institute
To be described relatively simple, the relevent part can refer to the partial explaination of embodiments of method.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the scope of the present invention.It is all
Any modification, equivalent replacement, improvement and so within the spirit and principles in the present invention, are all contained in protection scope of the present invention
It is interior.
Claims (10)
1. a kind of customer service question processing method, which is characterized in that the described method includes:
Obtain problem information；
Extract the keyword of described problem information；
Extracted keyword is input to shot and long term trained in advance and remembers LSTM neural network tensor model, obtains described ask
Inscribe the feature of information, wherein the LSTM neural network tensor model is: using the keyword of sample problem information to LSTM mind
The model obtained after network is trained；
Determine the text information in preset text information with obtained characteristic matching；Wherein, the preset text information
It is: the information that preset question text information sequence and preset answer text information sequence are summarized；
According to extracted feature, calculate similar between described problem information and question text information in determined text information
Spend information；
Using the answer text information in the highest text information of similarity as the answer information of described problem information.
2. the method as described in claim 1, which is characterized in that in the answer text by the highest text information of similarity
Before this information is as the answer information of described problem information, further includes:
Judge in the similarity information being calculated with the presence or absence of the similarity information greater than default value；
If there is the similarity information for being greater than default value, the answer text by the highest text information of similarity is executed
The step of answer information of this information as described problem information.
3. method according to claim 2, which is characterized in that judge whether deposit in the similarity information being calculated described
After the similarity information for being greater than default value, further includes:
If there is no the similarity information for being greater than default value, described problem information is sent to client, so that described
Client shows described problem information to contact staff；
Receive the answer information of contact staff's input that the client is sent, described；
Answer information to be output is determined from answer information obtained, and exports the answer information to be output.
4. method as claimed in claim 3, which is characterized in that receive the customer service people that the client is sent, described described
After the answer information of member's input, further includes:
Feedback information is sent to destination client, wherein the feedback information, it is described for prompting contact staff to receive
The answer information of problem information, the destination client are as follows: the client of not sent answer information.
5. a kind of customer service issue handling device, which is characterized in that described device includes:
Problem obtains module, for obtaining problem information；
Keyword-extraction module, for extracting the keyword of described problem information；
Characteristic extracting module, for extracted keyword to be input to shot and long term memory LSTM neural network trained in advance
Model is measured, obtains the feature of described problem information, wherein the LSTM neural network tensor model is: being believed using sample problem
The model that the keyword of breath obtains after being trained to LSTM neural network；
First determining module, for determining the text information in preset text information with obtained characteristic matching；Wherein, institute
Stating preset text information is: summarize to preset question text information sequence and preset answer text information sequence
The information arrived；
Computing module, for calculating question text in described problem information and determined text information according to extracted feature
Similarity information between information；
Second determining module, for using the answer text information in the highest text information of similarity as described problem information
Answer information.
6. device as claimed in claim 5, which is characterized in that described device further include:
Judgment module, with the presence or absence of the similarity information greater than default value in the similarity information for judging to be calculated；
If there is the similarity information for being greater than default value, second determining module is triggered.
7. device as claimed in claim 6, which is characterized in that described device further include:
If there is no the similarity information for being greater than default value, sending module is triggered；
The sending module, for described problem information to be sent to client, so that the client is to contact staff's exhibition
Show described problem information；
Information receiving module is answered, for receiving the answer information of contact staff's input that the client is sent, described；
Message output module is answered, for determining answer information to be output from answer information obtained, and described in output
Answer information to be output.
8. device as claimed in claim 7, which is characterized in that described device further include:
Feedback information sending module, for feedback information to be sent to destination client, wherein the feedback information, for mentioning
Show that contact staff has received the answer information of described problem information, the destination client are as follows: the visitor of not sent answer information
Family end.
9. a kind of electronic equipment, which is characterized in that including processor, communication interface, motor, memory and communication bus, wherein
Processor, communication interface, memory complete mutual communication by communication bus；
Memory, for storing computer program；
Processor when for executing the program stored on memory, realizes any method and step of Claims 1 to 4.
10. a kind of computer readable storage medium, which is characterized in that instruction is stored in the computer readable storage medium,
When run on a computer, any method and step of Claims 1 to 4 is realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811354975.9A CN109446314A (en) | 2018-11-14 | 2018-11-14 | A kind of customer service question processing method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811354975.9A CN109446314A (en) | 2018-11-14 | 2018-11-14 | A kind of customer service question processing method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109446314A true CN109446314A (en) | 2019-03-08 |
Family
ID=65554340
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811354975.9A Pending CN109446314A (en) | 2018-11-14 | 2018-11-14 | A kind of customer service question processing method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109446314A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110110095A (en) * | 2019-04-29 | 2019-08-09 | 国网上海市电力公司 | A kind of power command text matching technique based on shot and long term memory Recognition with Recurrent Neural Network |
| CN111275463A (en) * | 2020-02-14 | 2020-06-12 | 厦门快商通科技股份有限公司 | Intelligent customer service method, device and equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103294777A (en) * | 2013-05-13 | 2013-09-11 | 沈文策 | Question answering method and system |
| CN106649561A (en) * | 2016-11-10 | 2017-05-10 | 复旦大学 | Intelligent question-answering system for tax consultation service |
| CN108292305A (en) * | 2015-12-04 | 2018-07-17 | 三菱电机株式会社 | Method for handling sentence |
| CN108491433A (en) * | 2018-02-09 | 2018-09-04 | 平安科技（深圳）有限公司 | Chat answer method, electronic device and storage medium |
-
2018
- 2018-11-14 CN CN201811354975.9A patent/CN109446314A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103294777A (en) * | 2013-05-13 | 2013-09-11 | 沈文策 | Question answering method and system |
| CN108292305A (en) * | 2015-12-04 | 2018-07-17 | 三菱电机株式会社 | Method for handling sentence |
| CN106649561A (en) * | 2016-11-10 | 2017-05-10 | 复旦大学 | Intelligent question-answering system for tax consultation service |
| CN108491433A (en) * | 2018-02-09 | 2018-09-04 | 平安科技（深圳）有限公司 | Chat answer method, electronic device and storage medium |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110110095A (en) * | 2019-04-29 | 2019-08-09 | 国网上海市电力公司 | A kind of power command text matching technique based on shot and long term memory Recognition with Recurrent Neural Network |
| CN111275463A (en) * | 2020-02-14 | 2020-06-12 | 厦门快商通科技股份有限公司 | Intelligent customer service method, device and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11816440B2 (en) | Method and apparatus for determining user intent | |
| Umer et al. | Sentiment analysis of tweets using a unified convolutional neural network‐long short‐term memory network model | |
| WO2020073673A1 (en) | Text analysis method and terminal | |
| CN107341716A (en) | A kind of method, apparatus and electronic equipment of the identification of malice order | |
| CN111371767B (en) | Malicious account identification method, malicious account identification device, medium and electronic device | |
| CN110598070B (en) | Application type identification method and device, server and storage medium | |
| JP6661754B2 (en) | Content distribution method and apparatus | |
| CN110362665B (en) | Question-answering system and method based on semantic similarity | |
| CN110717009A (en) | Method and equipment for generating legal consultation report | |
| CN110913354A (en) | Short message classification method and device and electronic equipment | |
| CN109446314A (en) | A kind of customer service question processing method and device | |
| CN110532562B (en) | Neural network training method, idiom misuse detection method and device and electronic equipment | |
| WO2017173063A1 (en) | Updating messaging data structures to include predicted attribute values associated with recipient entities | |
| CN110096578A (en) | A kind of brush amount user identification method, device and the server of intelligent customer service | |
| CN114386435A (en) | Intelligent chat channel processor | |
| CN115269768A (en) | Element text processing method and device, electronic equipment and storage medium | |
| CN111694949B (en) | Multi-text classification method and device | |
| CN113139558A (en) | Method and apparatus for determining a multi-level classification label for an article | |
| JP2023554210A (en) | Sort model training method and apparatus for intelligent recommendation, intelligent recommendation method and apparatus, electronic equipment, storage medium, and computer program | |
| CN114048383A (en) | Information recommendation method and device, electronic equipment and medium | |
| CN113886543A (en) | Method, apparatus, medium, and program product for generating an intent recognition model | |
| CN114048315A (en) | Method and device for determining document tag, electronic equipment and storage medium | |
| CN113807920A (en) | Artificial intelligence based product recommendation method, device, equipment and storage medium | |
| CN113360672A (en) | Methods, apparatus, devices, media and products for generating a knowledge graph | |
| CN113821592A (en) | Data processing method, device, 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 |
Application publication date: 20190308 |
|
| WD01 | Invention patent application deemed withdrawn after publication |