CN104281615B - A kind of method and system of complaint handling - Google Patents

A kind of method and system of complaint handling Download PDF

Info

Publication number
CN104281615B
CN104281615B CN201310288121.6A CN201310288121A CN104281615B CN 104281615 B CN104281615 B CN 104281615B CN 201310288121 A CN201310288121 A CN 201310288121A CN 104281615 B CN104281615 B CN 104281615B
Authority
CN
China
Prior art keywords
text
complaint
result
vector
output
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.)
Active
Application number
CN201310288121.6A
Other languages
Chinese (zh)
Other versions
CN104281615A (en
Inventor
武强
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Mobile Group Gansu Co Ltd
Original Assignee
China Mobile Group Gansu Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Mobile Group Gansu Co Ltd filed Critical China Mobile Group Gansu Co Ltd
Priority to CN201310288121.6A priority Critical patent/CN104281615B/en
Publication of CN104281615A publication Critical patent/CN104281615A/en
Application granted granted Critical
Publication of CN104281615B publication Critical patent/CN104281615B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G06F16/355—Class or cluster creation or modification

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Machine Translation (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a kind of method and system of complaint handling, this method includes input and complains text；Analysis complains text to determine sample data with complaining the correlation of solution data；Identification of the training neutral net to complaint text；Identifying system result and actual result, carry out error analysis, and change corresponding weights.Technical scheme is due to using text space vector representation, it will complain text can be with Accurate classification, avoid due to text language expression problem, more staff can be facilitated to be sorted out for complaint situation, analyzed, also allow for system subsequent treatment, and the BP neural network of training sample less, with independent learning ability, for traditional manual tupe, there is complaint to position, and fast, problem amendment is fast, the advantage of full automation.

Description

A kind of method and system of complaint handling
Technical field
The present invention relates to telecommunications industry business support technical field, more particularly to a kind of method and system of complaint handling.
Background technology
As mobile communication business is fast-developing, a large amount of new business Quick threads, the customer complaint quantity occurred also with Rise.This becomes a larger obstacle of lifting user satisfaction.
Currently, the committed step for solving the problems, such as customer complaint is the accurate reply for the identification positioning and particular problem complained. The flow of complaint handling at present is：
1st, when user carry out complaint request during, operator by the request of user according to customer service system " service request by The template of reason " carries out typing.Wherein template is divided into：Communication cost, network quality, roaming service etc..
2nd, a line complaint handling personnel are handled according to the complaint work order of operator's typing.
3rd, difficult complaint is such as run into, a line complains personnel that network class is complained group into network management according to template classification The heart, support class are sent to business support portion.
4th, foreground personnel are complained to turning to send complaint to be reprocessed in two wires, such as also unresolved, and difficulty is being complained group to two Line background maintenance engineer.
Existing complaint handling implementation pattern is a kind of passive, poorly efficient, delay pattern.The complaint handling of upper level Complaint handling personnel help of the process content of personnel to next stage is smaller, and each layer of complaint handling personnel, will read over Complain summary, carry out oneself it is corresponding handle, the duplication of labour is caused, the problem of inefficiency, and according to complaint handling personnel The degree grasped to business is different, also uneven for customer complaint reply quality, reduces user satisfaction, influences user To the cognition degree of mobile communication service.
The content of the invention
In order to solve to handle the technical problem for complaining inefficiency in the prior art, the present invention proposes a kind of complaint handling Method and system, can complain workflow system, automation, intelligence by customer service, greatly save cost of labor.
One aspect of the present invention provides a kind of method of complaint handling, comprises the following steps：
Text is complained in input；
Analysis complains text to determine sample data with complaining the correlation of solution data；
Identification of the training neutral net to complaint text；
Identifying system result and actual result, carry out error analysis, and change corresponding weights.
Another aspect of the present invention provides a kind of system of complaint handling, including text space vector modular converter, BP god Through network core module and result output matching module, wherein,
Text space vector modular converter is used to text will be complained to be converted to text space vector；
BP neural network nucleus module is used to analyze the correlation complained text and complain solution data, determines sample Data, identification of the training neutral net to complaint text, identifying system result and actual result, carry out error analysis, and change Corresponding weights；
As a result export matching module to be used to be screened the complaint response of system archiving user's satisfaction, and user is expired Complaint response of anticipating carries out text vector conversion, forms a complaint response outcome pool R, the vector result of system processing is S, is used As a result Auto-matching algorithm exports the text vector of S to customer service system, receives the customer satisfaction system complaint of customer service system feedback Result vector R, carries out error analysis with system output result S by resulting text vector R and compares.
Technical scheme will complain text to be avoided with Accurate classification due to using text space vector representation Due to text language expression problem (same problem has different Expression of language), it can more facilitate staff for throwing The situation of telling is sorted out, is analyzed, and also allows for system subsequent treatment.And the BP of training sample less, with independent learning ability There is complaint to position fast, problem and correct the fast, advantage of full automation for traditional manual tupe for neutral net.
Brief description of the drawings
Fig. 1 is the flow chart of the complaint text space vectorization in the embodiment of the present invention.
Fig. 2 is the structure diagram of the neutral net in the embodiment of the present invention.
Fig. 3 is that the result in the embodiment of the present invention exports matched flow chart.
Fig. 4 is the structure diagram of the complaint handling system in the embodiment of the present invention.
Embodiment
The embodiment of the present invention is described in detail below in conjunction with the accompanying drawings.
Technical scheme is largely divided into three parts, and Part I is that complaint content text is carried out space vector Change is handled, and Part II is automatic positioning and processing complaint problem, and Part III exports result and Auto-matching.
The expression of text mainly uses vector space model (VSM).The basic thought of vector space model is come with vector Represent text：(W1, W2, W3 ..., Wn), wherein Wi are the weight of ith feature item.Vocabulary is represented as the dimension of vector Text.Initial vector representation is entirely 0,1 form, i.e.,：If occurs the word in text, then the dimension of text vector It is otherwise 0 for 1.
Fig. 1 is the flow chart of the complaint text space vectorization in the embodiment of the present invention.As shown in Figure 1, the complaint text Space vector flow comprises the following steps：
Step 101, establish and complain common words storehouse CK.
Such as vocabulary is included in lexicon：1. quit the subscription of business（Cancellation business）, 2. charge, 3.QQ monthly payments（QQ* is bored）, 4. nothings Method receives（It cannot use）, it is 5. single in detail, 6. guarantee the minimum, the change of 7. set meals, 8. promotion changes, 9. CRBT.
Text is complained in step 102, input.
Step 103, judge current location in lexicon, if has completed, if completed, has gone to step 104, if it is not complete, then going to step 105.
Step 104, terminate to complain text space vectorization flow.
Step 105, judge whether the current location of lexicon occurs in text is complained, if there is then going to step 106, if do not occurred, go to step 107.
The sequence location that step 106, lexicon are currently located fills in 1 in the corresponding dimension of space vector, and goes to step Rapid 108.
The sequence location that step 107, lexicon are currently located fills in 0 in the corresponding dimension of space vector, and goes to step Rapid 108.
Current location in step 108, lexicon adds 1, and goes to step 103.
In this way, after customer complaint text is inputted, by retrieving the keyword in dictionary（Include keyword synonym）, such as For fruit there are the keyword, which fills in 1 according to the sequence location of place dictionary in the corresponding dimension of space vector, it is on the contrary then Fill in 0.Assuming that CK dictionaries only include following 9 vocabulary, customer complaint content is " user's reflection is unable to normal use QQ blue diamond industry Business, please be handled " by poll search key：First 0, second 0, the 3rd 1, the 4th 1, the 5th to the 9th is all 0, algorithm finally obtains text vector（0,0,1,1,0,0,0,0,0）.
From the point of view of existing customer complaint process experience, when being easiest to reduce the link of user satisfaction, that is, spending Between longest link, be exactly complain be accurately positioned.Therefore clear and definite relation is established between complaining phenomenon and complaining Resolving probiems, For shortening the complaint handling time, repetitive manual work is reduced, is had great importance.
Complaint is automatically processed in real time using the neutral net shown in Fig. 2 in the embodiment of the present invention.Using BP god Through network, establishing one has three-layer network model, including input layer, hidden layer and output layer, is inputted in input layer and complains text （After space vector conversion）, by the effect of the function of hidden layer, in output port output valve,
Output valve is subjected to error calculation, if being not reaching to error requirements, according to correct result, output feedback information arrives Input layer and hidden layer, adjust weights and correlation function, and to have the function that to optimize output valve, that is, system can constantly certainly I learns（Great convenience is provided to later stage system maintenance）.
Input layer is according to the text space vector I for complaining conversion0, in BP neural network, each neuron threshold θj, it is preceding The weights of one layer of neuron to later layer neuron are Wij；
For the input I of hidden layer and output layerj=Σ WijOi+θj
Neuron output in neutral net is calculated via function living is assigned, and assigns function living and uses simoid functions Or logistic functions.The output of neuron is
Error calculation is for each output unit of output layer Errj=Oj（1-Oj）(Tj-Oj）, wherein T is actual result.
Each weights W in neutral netijCorrection formula be
ΔWij=(l) ErrjOi//power increment
Wij=Wij+ΔWijThe renewal of // power
}。
Each threshold value in neutral net（Deviation）θjCorrection formula be
Δθj=(l) Errj// deviation is rised in value
θj=θj+Δθj// deviation updates
}。
System realizes that the Artificial neural network ensemble structure includes three important steps：
1st, analysis complains text and complains the correlation of solution data, determines sample data.Such as complain source： " spoa shows that dream net switch is closed, and user can not use dream network service "（Vector space is complained to represent 0100000111）, finally Complain solution：" dream net switch is please opened in crm system MISC message switchings "（00111101110111）. 0100000111 has correlation with 00111101110111.
2nd, training neutral net（Grader）To customer complaint（The identification of single eigenvalue）Identification.
3rd, identifying system result carries out error analysis with actual result and changes corresponding weights.
The BP neural network algorithm that above structure uses, input can quickly establish clear and definite correspondence with output, count Calculate the limitation that complexity is not high, is complained due to telecommunications（Dictionary capacity is small, without considering that multiple characteristic values input）, systematic training Sample it is little, it is workable during realization.
Part III output result and Auto-matching, as shown in figure 3, comprising the following steps：
Step 301, by system archiving user satisfaction complaint response screened, and by user be satisfied with complaint response into Row text vector is changed, and forms a complaint response outcome pool R.
Step 302, the vector result of system processing are S, using result Auto-matching algorithm by the text vector of S to customer service System exports.
Step 303, the customer satisfaction system complaint result vector R for receiving customer service system feedback, by resulting text vector R with being System output result S carries out error analysis comparison.
In order to realize above-mentioned flow, the embodiment of the present invention also proposed a kind of complaint handling system, and Fig. 4 is of the invention real Apply the structure diagram of the complaint handling system in example.As shown in figure 4, the complaint handling system includes text space vector conversion Module 401, BP neural network nucleus module 402, result output matching module 403 and complaint analysis display module 404.
Text space vector modular converter is connected with client, receives the input for complaining text, and text will be complained to turn Text space vector is changed to, is sent to BP neural network nucleus module.
The analysis of BP neural network nucleus module complains text to determine sample number with complaining the correlation of solution data According to, training neutral net to complaining the identification of text, identifying system result and actual result, carry out error analysis, and change pair The weights answered.
As a result export matching module to be screened the complaint response of system archiving user's satisfaction, and user is satisfied with and is thrown Tell that reply carries out text vector conversion, form a complaint response outcome pool R, the vector result of system processing is S, using result Auto-matching algorithm exports the text vector of S to customer service system, receives the customer satisfaction system complaint result of customer service system feedback Vectorial R, carries out error analysis with system output result S by resulting text vector R and compares.
Analysis display module analysis space vector is complained, analysis displaying is carried out by report, block diagram, cake chart etc., is Complaint problem is presented in business department, background maintenance personnel, and carry out business, system optimization provide effective instrument.
Traditional customer service is complained workflow system, automation, intelligence by the embodiment of the present invention.Greatly save people Work cost.Wherein technological merit has：
1st, using text space vector representation, text will be complained to be avoided with Accurate classification since text language is expressed Problem (same problem has different Expression of language), can more facilitate staff to be sorted out for complaint situation, divide Analysis, also allows for system subsequent treatment.
2nd, the BP neural network of training sample less, with independent learning ability is for traditional manual tupe, tool There is complaint to position fast, problem and correct the fast, advantage of full automation.
It should be noted that：Only to illustrate rather than limitation, the present invention is also not limited to above-mentioned above example Citing, all do not depart from the technical solution of the spirit and scope of the present invention and its improvement, it should all cover the right in the present invention In claimed range.

Claims (8)

  1. A kind of 1. method of complaint handling, it is characterised in that comprise the following steps：
    Text is complained in input；
    Analysis complains text to determine sample data with complaining the correlation of solution data；
    Identification of the training neutral net to complaint text；
    Identifying system result and actual result, carry out error analysis, and change corresponding weights；
    It is further comprising the steps of：
    The complaint response of system archiving user's satisfaction is screened, and user is satisfied with complaint response and carries out text vector turn Change, form a complaint response outcome pool R；
    The vector result of system processing is S, is exported the text vector of S to customer service system using result Auto-matching algorithm；
    The customer satisfaction system complaint result vector R of customer service system feedback is received, by resulting text vector R and system output result S Carry out error analysis comparison.
  2. 2. the method for a kind of complaint handling according to claim 1, it is characterised in that further comprising the steps of：
    Text will be complained to be converted to text space vector.
  3. 3. the method for a kind of complaint handling according to claim 2, it is characterised in that described that text will be complained to be converted to text This space vector, further comprises the steps：
    Establish and complain common words storehouse；
    Keyword is obtained from complaining in text, the search key in lexicon；
    If lexicon is there are the keyword, according to the sequence location of lexicon where the keyword, in space vector pair 1 is arranged in the dimension answered, is otherwise provided as 0.
  4. 4. the method for a kind of complaint handling according to any claim in claim 1-3, it is characterised in that described point Analysis complains text to further comprise the steps with complaining the correlation of solution data：
    Using BP neural network, establishing one has three-layer network model, including input layer, hidden layer and output layer；
    Inputted in input layer and complain text, by the effect of the function of hidden layer, in output port output valve；
    Output valve is subjected to error calculation, if being not reaching to error requirements, according to correct result, output feedback information to input Layer and hidden layer, adjust weights and correlation function.
  5. A kind of 5. method of complaint handling according to claim 4, it is characterised in that
    The text space vector I of text conversion is complained in input layer input0, in BP neural network, each neuron threshold θj, it is previous The weights of layer neuron to later layer neuron are Wij；
    For the input I of hidden layer and output layerj=∑ WijOi+θj；
    Neuron output in neutral net is calculated via function living is assigned, and the output of neuron is
    Error calculation is for each output unit of output layer Errj=Oj(1-Oj)(Tj-Oj), wherein TjFor actual result.
  6. 6. the method for a kind of complaint handling according to claim 5, it is characterised in that the function of living of assigning uses simoid Function or logistic functions.
  7. 7. a kind of system of complaint handling, it is characterised in that including text space vector modular converter, BP neural network core mould Block and result output matching module, wherein,
    Text space vector modular converter is used to text will be complained to be converted to text space vector；
    BP neural network nucleus module is used to analyze the correlation complained text and complain solution data, determines sample number According to, training neutral net to complaining the identification of text, identifying system result and actual result, carry out error analysis, and change pair The weights answered；
    As a result export matching module to be used to be screened the complaint response of system archiving user's satisfaction, and user is satisfied with and is thrown Tell that reply carries out text vector conversion, form a complaint response outcome pool R, the vector result of system processing is S, using result Auto-matching algorithm exports the text vector of S to customer service system, receives the customer satisfaction system complaint result of customer service system feedback Vectorial R, carries out error analysis with system output result S by resulting text vector R and compares.
  8. 8. the system of a kind of complaint handling according to claim 7, it is characterised in that further include complaint analysis displaying mould Block, it is described to complain analysis display module to be used for analysis space vector, carry out analysis displaying.
CN201310288121.6A 2013-07-08 2013-07-08 A kind of method and system of complaint handling Active CN104281615B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310288121.6A CN104281615B (en) 2013-07-08 2013-07-08 A kind of method and system of complaint handling

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310288121.6A CN104281615B (en) 2013-07-08 2013-07-08 A kind of method and system of complaint handling

Publications (2)

Publication Number Publication Date
CN104281615A CN104281615A (en) 2015-01-14
CN104281615B true CN104281615B (en) 2018-05-15

Family

ID=52256491

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310288121.6A Active CN104281615B (en) 2013-07-08 2013-07-08 A kind of method and system of complaint handling

Country Status (1)

Country Link
CN (1) CN104281615B (en)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104866550A (en) * 2015-05-12 2015-08-26 湖北光谷天下传媒股份有限公司 Text filtering method based on simulation of neural network
CN108021582B (en) * 2016-11-04 2020-12-04 中国移动通信集团湖南有限公司 Internet public opinion monitoring method and device
CN107247792B (en) * 2017-06-16 2021-01-15 中国电子技术标准化研究院 Method and device for matching functional departments and computer equipment
CN109345262A (en) * 2017-08-01 2019-02-15 兰州大学 It is a kind of intelligently to complain classification and processing system
CN107729919A (en) * 2017-09-15 2018-02-23 国网山东省电力公司电力科学研究院 In-depth based on big data technology is complained and penetrates analysis method
CN108563791A (en) * 2018-04-29 2018-09-21 华中科技大学 A kind of construction quality complains the method and system of text classification
CN109684475B (en) * 2018-11-21 2021-03-30 斑马网络技术有限公司 Complaint processing method, complaint processing device, complaint processing equipment and storage medium
CN111340323B (en) * 2018-12-19 2023-09-05 中国移动通信集团湖南有限公司 Automatic dispatch method and system for complaint service request
CN110308244B (en) * 2019-06-26 2022-03-04 深圳市宇驰检测技术股份有限公司 Air monitoring and early warning method and system of unmanned aerial vehicle and storage medium
CN110704619B (en) * 2019-09-24 2022-06-10 支付宝(杭州)信息技术有限公司 Text classification method and device and electronic equipment
CN112584407B (en) * 2020-12-04 2022-07-22 重庆玖舆博泓科技有限公司 LTE user complaint qualitative method and device based on space-time combination

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101047935A (en) * 2007-04-27 2007-10-03 中国移动通信集团福建有限公司 Customer complaint process analysis method
CN101320451A (en) * 2008-07-04 2008-12-10 中国移动通信集团福建有限公司 Customer complaining method using decision tree
CN101458676A (en) * 2007-12-11 2009-06-17 中国移动通信集团公司 Quality testing determining method, quality testing system and neural net
CN102567405A (en) * 2010-12-31 2012-07-11 北京安码科技有限公司 Hotspot discovery method based on improved text space vector representation

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6269351B1 (en) * 1999-03-31 2001-07-31 Dryken Technologies, Inc. Method and system for training an artificial neural network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101047935A (en) * 2007-04-27 2007-10-03 中国移动通信集团福建有限公司 Customer complaint process analysis method
CN101458676A (en) * 2007-12-11 2009-06-17 中国移动通信集团公司 Quality testing determining method, quality testing system and neural net
CN101320451A (en) * 2008-07-04 2008-12-10 中国移动通信集团福建有限公司 Customer complaining method using decision tree
CN102567405A (en) * 2010-12-31 2012-07-11 北京安码科技有限公司 Hotspot discovery method based on improved text space vector representation

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
面向电信客户投诉和建议的智能分析模型;李济汉 等;《现代电信科技》;20130525(第5期);4-7 *

Also Published As

Publication number Publication date
CN104281615A (en) 2015-01-14

Similar Documents

Publication Publication Date Title
CN104281615B (en) A kind of method and system of complaint handling
CN110020426B (en) Method and device for distributing user consultation to customer service group
CN104572998B (en) Question and answer order models update method and device for automatically request-answering system
CN110427466A (en) Training method and device for the matched neural network model of question and answer
CN106303113B (en) A kind of shunt method and equipment
CN108021934A (en) The method and device of more key element identifications
CN111199272A (en) Adaptive scheduling method for intelligent workshop
CN110309967A (en) Prediction technique, system, equipment and the storage medium of customer service session grading system
CN107993019A (en) A kind of resume appraisal procedure and device
CN112419096A (en) Automatic user power demand worksheet transferring method based on NLP information extraction and few-sample self-learning
CN112417159A (en) Cross-language entity alignment method of context alignment enhanced graph attention network
CN109345262A (en) It is a kind of intelligently to complain classification and processing system
CN109993543A (en) A kind of complaint handling method and system
US20220414687A1 (en) Method, device, equipment and medium for determining customer tabs based on deep learning
CN109657792A (en) Construct the method, apparatus and computer-readable medium of neural network
CN110147905A (en) Information processing method, device, system and storage medium
CN116777148B (en) Intelligent distribution processing system for service work orders based on data analysis
CN110532394A (en) The processing method and system of Order Remarks text
CN109693244A (en) The method and device of optimization dialogue robot
CN116757837A (en) Credit wind control method and system applied to winning bid
Tapia et al. Implementing Operational AI in telecom environments
CN108009735A (en) A kind of resume appraisal procedure and device
CN116703046A (en) Control method and system for real-time dispatching sequence, electronic equipment and storage medium
Avdagić-Golub et al. Optimization of agent-user matching process using a machine learning algorithms
CN111402004A (en) Personalized telecom package recommendation method for telecom operator

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant