CN112258016A - Customer service intelligent quality inspection analysis method based on ASR - Google Patents
Customer service intelligent quality inspection analysis method based on ASR Download PDFInfo
- Publication number
- CN112258016A CN112258016A CN202011116925.4A CN202011116925A CN112258016A CN 112258016 A CN112258016 A CN 112258016A CN 202011116925 A CN202011116925 A CN 202011116925A CN 112258016 A CN112258016 A CN 112258016A
- Authority
- CN
- China
- Prior art keywords
- quality inspection
- data
- call
- quality
- customer service
- 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.)
- Withdrawn
Links
- 238000007689 inspection Methods 0.000 title claims abstract description 305
- 238000004458 analytical method Methods 0.000 title claims abstract description 27
- 238000013499 data model Methods 0.000 claims abstract description 29
- 238000000034 method Methods 0.000 claims abstract description 24
- 238000011156 evaluation Methods 0.000 claims abstract description 19
- 238000013515 script Methods 0.000 claims abstract description 17
- 238000004141 dimensional analysis Methods 0.000 claims abstract description 7
- 230000001360 synchronised effect Effects 0.000 claims description 22
- 238000004422 calculation algorithm Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 15
- 238000012372 quality testing Methods 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 12
- 238000006243 chemical reaction Methods 0.000 claims description 10
- 230000010365 information processing Effects 0.000 claims description 9
- 230000007246 mechanism Effects 0.000 claims description 9
- 238000013518 transcription Methods 0.000 claims description 9
- 230000035897 transcription Effects 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 8
- 238000012423 maintenance Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 4
- 241000512668 Eunectes Species 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 238000007621 cluster analysis Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 230000001960 triggered effect Effects 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 7
- 238000004806 packaging method and process Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 6
- 230000000007 visual effect Effects 0.000 description 4
- 238000007726 management method Methods 0.000 description 3
- 238000012550 audit Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000003908 quality control method Methods 0.000 description 2
- 238000007670 refining Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/216—Parsing using statistical methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0281—Customer communication at a business location, e.g. providing product or service information, consulting
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- General Engineering & Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Data Mining & Analysis (AREA)
- Game Theory and Decision Science (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- General Business, Economics & Management (AREA)
- Educational Administration (AREA)
- Marketing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an intelligent customer service quality inspection analysis method based on ASR, which is characterized by comprising the following steps of 1: generating a natural language text which accords with a quality inspection rule, splitting and packaging the natural language text into a quality inspection data model, and performing quality inspection by a quality inspection processor; step 2, intelligent quality inspection, presetting a quality inspection strategy and quality inspection factors into a quality inspection processor, dynamically analyzing quality inspection data in a quality inspection data model, carrying out multi-dimensional analysis according to the quality inspection strategy and the quality inspection factors, and generating a quality inspection result; and step 3: generating a quality inspection report, and generating a related quality inspection report for each ticket information after the quality inspection manager finishes quality inspection of the data; and 4, step 4: generating knowledge base script information, and storing scripts generated by natural language texts which reach the preset evaluation grade standard of the system into a knowledge base; and 5: and maintaining the customer service guide flow. The method achieves the effects of objectivity, standardization, comprehensiveness and real-time performance, unifies the assessment standards among different quality inspectors, solves the limitation of the self service capability of the quality inspectors, and solves the problem of the hysteresis of customer service quality inspection.
Description
Technical Field
The invention relates to the technical field of internet, in particular to an intelligent customer service quality inspection analysis method based on ASR.
Background
The customer is the survival basis of the enterprise, along with the improvement of the service quality requirement of the customer, the enterprise pays more attention to the customer service, many enterprises introduce manual quality inspection posts, but along with the rapid increase of the business volume, the manual quality inspection cost is huge, secondly, the huge recording data of the call center is too heavy for limited manual quality inspection personnel, and the quality inspection proportion only accounts for about 0.5% -2% of all seats. The assessment standards of different quality inspectors are not uniform, and the quality inspectors are easily influenced by subjective consciousness and cannot be objective and fair. Quality control is also subject to non-specification due to limitations in the business capabilities of the quality control personnel themselves.
Expensive quality inspection results with low labor cost and accuracy cannot meet the real needs of enterprises, and loss of customers is finally caused by reputation of enterprise brands if problems in the customer service process cannot be analyzed and found timely and accurately or the customer service cannot identify the real needs of the customers correctly.
Therefore, an efficient customer service quality inspection analysis method is needed.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide an intelligent customer service quality inspection analysis method based on ASR.
In order to solve the technical problem, the invention provides an intelligent customer service quality inspection analysis method based on ASR, which is characterized by comprising the following steps:
step 1: generating a natural language text according with a quality inspection rule, wherein the quality inspection rule is a rule meeting a quality inspection data model of a quality inspection processor, the quality inspection processor consists of quality inspection data, a quality inspection strategy and a quality inspection factor, the quality inspection data model consists of a quality inspection object, a natural language text and an object text relation, the quality inspection object comprises customer service and a client, the natural language text is the text content of the conversation between the customer service and the client, the object text relation comprises the text relation corresponding to the customer service and the client, the quality inspection rule comprises conversation voice, the conversation voice comprises a bill attribute, a conversation object and a conversation object, the bill attribute comprises a main bill key related to the customer service, and the main bill key comprises: the method comprises the following steps of binding a ticket and a customer service by using a unique identifier, voice starting time, voice ending time, voice call duration, voice connection waiting duration and customer evaluation; the call object is a calling object; the called object is a called object; the natural language text is a natural language text which can be subjected to quality inspection and is generated by the voice transcription system according to the ASR system which automatically matches the recording file and supports transcription, the natural language text is split and packaged into a quality inspection data model, and a quality inspection processor performs quality inspection;
step 2, intelligent quality inspection, presetting a quality inspection strategy and quality inspection factors into a quality inspection processor, dynamically analyzing quality inspection data in a quality inspection data model, carrying out multi-dimensional analysis according to the quality inspection strategy and the quality inspection factors, and generating a quality inspection result;
and step 3: generating a quality inspection report, generating a related quality inspection report for each call bill information after the quality inspection manager finishes quality inspection of data, wherein the quality inspection report shows a current call bill quality inspection result, the call bill quality inspection result comprises score information, a total score, a sensitive word set and customer intention keywords of three quality inspection standard item templates, and the system performs cluster analysis according to batch call bills to generate a trend graph of initial score and total score change of customer service at stage time and the triggering times of sensitive words, so that a supervisor can conveniently know service quality change of the customer service at stage time and know the quality condition of the customer service information in multiple directions;
and 4, step 4: generating knowledge base script information, wherein the knowledge base script information is preset by a knowledge base, collecting the call bill texts which reach the system preset evaluation level standard into a high-quality call bill base table according to the call bill quality inspection result, carrying out secondary analysis on the natural language texts in the high-quality call bill base table, refining the dialogues in the customer service natural language texts through a quality inspection tool bag, storing the scripts generated by the natural language texts which reach the system preset evaluation level standard into the knowledge base, and enabling the scripts to enter the knowledge base after the scripts pass the audit of a supervisor.
And 5: and (4) customer service guide flow maintenance, namely filing the natural language text data and the quality inspection report which reach the preset evaluation level standard of the system and are subjected to secondary analysis, recording the natural language text which reaches the preset evaluation level standard of the system and is subjected to secondary analysis, and performing customer service guide flow maintenance according to the recorded information of selection, filtration and filing of the natural language text.
In step 1, the specific steps of generating the natural language text are as follows:
step 1-1, acquiring call data of a third party call center system according to call ticket id in a database table through a scheduling task according to a timing scheduling preset strategy, wherein the timing scheduling preset strategy is a time strategy for acquiring the call data of the third party call center system, an acquisition period is configured in a form of configuring a corner expression, the acquisition period is 1 hour or half a day, the call data corresponds to the call ticket id according to the call ticket data generated in real time in a call system by customer service, the call ticket data is data in the database table corresponding to the third party call center system, the call data is acquired from the database of the third party call center system according to the call ticket id, the call data meeting a quality inspection rule is synchronized and converted into a model needed by the system to be put in storage, a call ticket data synchronization identifier in the database table is updated to be 1, and for the call data not meeting the quality inspection rule, updating the ticket data synchronization identifier in the database table to 0 to indicate failure, recording the number of updating synchronization +1, wherein the synchronization threshold is 50, and compensating the acquired failed or non-generated voice data through a compensation mechanism, wherein the compensation mechanism is as follows: after the synchronous data flow is executed for the first time in each scheduling, the customer service call bill data with the call bill data synchronous identification of 0 and the synchronous times not exceeding the threshold value in the database are obtained again, the call data are synchronized in the call center database, and the synchronous failure data exceeding the threshold value are marked with abnormal labels and are processed offline by service personnel;
step 1-2: the voice file intelligently converts a natural language text, the ASR system is asynchronously called to perform voice conversion according to the call ticket data which is successfully synchronized in the steps, each call ticket data can independently open one thread, the data conversion between the threads is not mutually influenced, the voice conversion can be rapidly completed, the converted text is analyzed and packaged according to a quality inspection data model, the quality inspection data model is matched according to quality inspection factor items in a quality inspection processor, and the packaged quality inspection data model is stored in a database.
In the step 2, the intelligent quality inspection is specifically performed according to the following steps:
step 2-1: presetting quality inspection factors, wherein the quality inspection factors are divided into three different attribute dimensions by factor items: the system comprises a data factor, a client factor and a customer service factor, wherein the data factor comprises call starting time, call ending time, call duration, client queuing duration and a service queue selected by a client; the client factor comprises client intention keywords and client evaluation; the customer service factors comprise customer service expression capacity, service expression, service attitude, opening white, statement words, closing words and bad sensitive words; and the quality inspection processor scores according to scores of different standards corresponding to the factor items in the quality inspection standard item template of the quality inspection strategy in the step 2-2. The factor item determines the quality inspection dimension, the background visual dynamic configuration is realized, a quality inspector can configure specific factor items in a quality inspection factor module of a background quality inspection processor, the configuration is finished and stored in a warehouse, the visual dynamic configuration needs to be configured before the next period of the system quality inspection, and the next quality inspection period can take effect at the beginning.
Step 2-2: presetting a quality inspection strategy, wherein the quality inspection strategy consists of a quality inspection standard item template and a quality inspection chain, more than one quality inspection standard item template is provided, the quality inspection standard item template determines a grading standard of a specific dimension by factor items, the quality inspection standard item template calculates scores of different dimensions according to attributes and rules of the quality inspection standard item template, quality inspection is carried out on quality inspection data according to the quality inspection standard item template, the quality inspection score is calculated and presented by a two-dimensional matrix, the quality inspection chain comprises a score calculation chain and an information processing chain, the score calculation chain is a calculation chain formed by quality inspection score rules of different types of quality inspection standard item templates, the information processing chain comprises a sensitive word template and a customer intention information extraction template, the sensitive word template is matched with natural language text contents of a service object in a data object according to a preset sensitive word bag, and a customer obtains malicious words in the sensitive word bag, triggering a sensitive word early warning mechanism in real time, wherein the wind control system carries out early warning according to a sensitive word template, and the customer intention information extraction template extracts customer intention product keywords or question keywords according to keywords in a customer intention keyword library table;
step 2-3: configuring a quality inspection toolkit and a matching algorithm, wherein the quality inspection process specifically comprises the following steps: matching, analyzing, and quality testing the generated natural language text, the quality testing tool package being a Gensim tool package, centrally installed on a server using anaconda tools for processing natural language text, matching text content according to sensitive words in sensitive word bags using kmp matching algorithms, the quality testing tool package and kmp matching algorithms being automatically initialized by the system.
Step 2-4: the quality inspection processor performs quality inspection, and performs multi-dimensional quality inspection grading, early warning of sensitive words and extraction of client intention keywords according to a quality inspection data model, quality inspection factors and a quality inspection strategy preset by the system; the method comprises the steps that a preset quality inspection factor item is initially changed, a model in a quality inspection standard template is related according to the factor item, a grading standard is preset in the model, and a quality inspection processor analyzes and calculates the value of the model according to a value calculation chain in a quality inspection chain and on the basis of a genim open source toolkit according to a quality inspection standard template class. The method comprises the steps of importing a pre-prepared industry service corpus into a language model of genim, producing expected word bags by the Gensim according to the imported corpus, enabling each word in each word bag to correspond to a unique id, cutting each sentence in the corpus by utilizing jieba participles, producing word bags by a cut collection, forming a key, namely a value dictionary, enabling elements in each list to have unique id marks, cutting the imported conversation text by a quality inspection processor, matching the cut participle collection with the elements in the word bags, and calculating corresponding scores according to matching results and standard word bags. Obtaining initial scores according to multi-dimensional analysis and grading of different score templates, after processing of each template class is finished, calculating an average value according to each initial score to obtain a final quality inspection total score, finishing calculation of the total score, after execution of a current score chain is finished, extracting early warning sensitive words and customer intention information according to an information processing chain, matching the sensitive words of a preset sensitive word bag in a customer service object text according to an kmp algorithm, sending a matched set to a wind control system through a message system for early warning, wherein the customer information extraction refers to matching keywords of a customer intention keyword library according to a kmp algorithm, and extracting data.
In the step 2-2: the quality inspection standard item template comprises an etiquette standard template, a flow standard template and a service skill template; each quality inspection standard item template comprises more than one basic standard item, each basic standard item has a corresponding grading level which corresponds to A, B + and B, C, D, E grades respectively, each grade corresponds to a ten-degree score, the quality inspection processor analyzes the dialog text corresponding to the customer service object in the data model according to a corpus preset by the basic standard items, and the corresponding score is given according to the matching degree of the analysis result and the grading model.
The invention achieves the following beneficial effects:
objectivity: the assessment standards among different quality inspectors are unified, manual judgment is reduced, subjective influence is avoided, objective and fair effects are achieved, and meanwhile labor cost is saved.
And (3) specification: the method solves the limitation of the self service capability of the quality inspector, improves the assessment effect of the quality inspector, and eliminates the non-standard quality inspection of the quality inspection process by external factors.
Comprehensive: the method solves the limitation of quality inspection resources, seamlessly covers each customer service telephone, performs 100% full quality inspection, and effectively avoids the sampling mode from watching the telephone.
Real-time: the method solves the problem of lag of customer service quality inspection, monitors customer service calls in time, and finds and warns in time, so that the supervisor can find and solve problems in time.
Drawings
FIG. 1 is a simplified process flow diagram of an exemplary embodiment of the present invention;
FIG. 2 is a system architecture diagram in an exemplary embodiment of the invention;
FIG. 3 is a schematic diagram of a quality inspection processor in an exemplary embodiment of the invention;
FIG. 4 is a schematic diagram of implementing collaboration in an exemplary embodiment of the invention;
FIG. 5 is a diagram of base criteria items in an exemplary embodiment of the invention;
FIG. 6 is a flow diagram illustrating a customer service guide in an exemplary embodiment of the invention.
Detailed Description
The invention will be further described with reference to the drawings and the exemplary embodiments:
as shown in fig. 1, an ASR-based customer service intelligent quality inspection analysis method is characterized by comprising the following steps:
step 1: generating a natural language text according with a quality inspection rule, wherein the quality inspection rule is a rule meeting a quality inspection data model of a quality inspection processor, the quality inspection processor shown in fig. 3 comprises quality inspection data, a quality inspection strategy and a quality inspection factor, the quality inspection data model comprises a quality inspection object, a natural language text and an object text relation, the quality inspection object comprises customer service and a customer, the natural language text is text content of conversation between the customer service and the customer, the object text relation comprises text relation corresponding to the customer service and the customer, the quality inspection rule comprises call voice, the call voice comprises call ticket attributes, a call object and a called object, the call ticket attributes comprise a call ticket main key with the associated customer service, and the call ticket main key comprises: the method comprises the following steps of binding a ticket and a customer service by using a unique identifier, voice starting time, voice ending time, voice call duration, voice connection waiting duration and customer evaluation; the call object is a calling object; the called object is a called object; the natural language text is a natural language text which can be subjected to quality inspection and is generated by the voice transcription system according to the ASR system which automatically matches the recording file and supports transcription, the natural language text is split and packaged into a quality inspection data model, and a quality inspection processor performs quality inspection;
in step 1, the specific steps of generating the natural language text are as follows:
step 1-1, as shown in FIG. 4, a cooperation flow of an ASR-based intelligent customer service quality inspection analysis method, which obtains call data of a third party call center system according to a call ticket id in a database table by a scheduling task according to a timing scheduling preset strategy, wherein the timing scheduling preset strategy is a time strategy for obtaining the call data of the third party call center system, an obtaining period is configured in a form of configuring a corner expression, the obtaining period is 1 hour or half day, the call data is the data in the database table corresponding to the third party call center system according to the call ticket id corresponding to the call ticket data generated by the customer service in the call system in real time, the call data is obtained from the database of the third party call center system according to the call ticket id, the call data meeting the quality inspection rule is synchronized and converted into a model required by the system to be put in storage, and the synchronous identification of the call data in the database table is updated to be 1, which shows success, for the call data which do not meet the quality inspection rule, updating the ticket data synchronization identification in the database table to 0 to indicate failure, recording the number of updating synchronization times +1, wherein the synchronization time threshold is 50 times, and compensating the acquired failure or non-generated voice data through a compensation mechanism, wherein the compensation mechanism is as follows: after the synchronous data flow is executed for the first time in each scheduling, the customer service call bill data with the call bill data synchronous identification of 0 and the synchronous times not exceeding the threshold value in the database are obtained again, the call data are synchronized in the call center database, and the synchronous failure data exceeding the threshold value are marked with abnormal labels and are processed offline by service personnel;
step 1-2: intelligent conversion of natural language text from voice file
And asynchronously calling an ASR system to perform voice conversion according to the call ticket data which is successfully synchronized in the steps, wherein each call ticket data can independently start a thread, the data conversion between the threads is not mutually influenced, the voice conversion can be quickly completed, the converted text is analyzed and packaged according to a quality inspection data model, the quality inspection data model is matched according to quality inspection factor items in a quality inspection processor, and the packaged quality inspection data model is stored in a database.
Step 2, intelligent quality inspection, presetting a quality inspection strategy and quality inspection factors into a quality inspection processor, dynamically analyzing quality inspection data in a quality inspection data model, carrying out multi-dimensional analysis according to the quality inspection strategy and the quality inspection factors, and generating a quality inspection result;
in the step 2, the intelligent quality inspection is specifically performed according to the following steps:
step 2-1: presetting quality inspection factors, wherein the quality inspection factors are divided into three different attribute dimensions by factor items: the system comprises a data factor, a client factor and a customer service factor, wherein the data factor comprises call starting time, call ending time, call duration, client queuing duration and a service queue selected by a client; the client factor comprises client intention keywords and client evaluation; the customer service factors comprise customer service expression capacity, service expression, service attitude, opening white, statement words, closing words and bad sensitive words; and the quality inspection processor scores according to scores of different standards corresponding to the factor items in the quality inspection standard item template of the quality inspection strategy in the step 2-2. The factor item determines the quality inspection dimension, the background visual dynamic configuration is realized, a quality inspector can configure specific factor items in a quality inspection factor module of a background quality inspection processor, the configuration is finished and stored in a warehouse, the visual dynamic configuration needs to be configured before the next period of the system quality inspection, and the next quality inspection period can take effect at the beginning.
Step 2-2: presetting a quality inspection strategy, wherein the quality inspection strategy consists of a quality inspection standard item template and a quality inspection chain, more than one quality inspection standard item template is provided, the quality inspection standard item template determines a grading standard of a specific dimension by factor items, the quality inspection standard item template calculates scores of different dimensions according to attributes and rules of the quality inspection standard item template, quality inspection is carried out on quality inspection data according to the quality inspection standard item template, the quality inspection score is calculated and presented by a two-dimensional matrix, the quality inspection chain comprises a score calculation chain and an information processing chain, the score calculation chain is a calculation chain formed by quality inspection score rules of different types of quality inspection standard item templates, the information processing chain comprises a sensitive word template and a customer intention information extraction template, the sensitive word template is matched with natural language text contents of a service object in a data object according to a preset sensitive word bag, and a customer obtains malicious words in the sensitive word bag, triggering a sensitive word early warning mechanism in real time, wherein the wind control system carries out early warning according to a sensitive word template, and the customer intention information extraction template extracts customer intention product keywords or question keywords according to keywords in a customer intention keyword library table;
in the step 2-2: as shown in fig. 5, the quality inspection standard item template includes an etiquette normative template, a flow normative template, and a service skill template; each quality inspection standard item template comprises more than one basic standard item, each basic standard item has a corresponding grading level which corresponds to A, B + and B, C, D, E grades respectively, each grade corresponds to a ten-degree score, the quality inspection processor analyzes the dialog text corresponding to the customer service object in the data model according to a corpus preset by the basic standard items, and the corresponding score is given according to the matching degree of the analysis result and the grading model.
Step 2-3: configuring a quality inspection toolkit and a matching algorithm, wherein the quality inspection process specifically comprises the following steps: matching, analyzing, and quality testing the generated natural language text, the quality testing tool package being a Gensim tool package, centrally installed on a server using anaconda tools for processing natural language text, matching text content according to sensitive words in sensitive word bags using kmp matching algorithms, the quality testing tool package and kmp matching algorithms being automatically initialized by the system.
Step 2-4: quality testing processor quality testing
The quality inspection processor performs multi-dimensional quality inspection grading, early warning of sensitive words and extraction of client intention keywords according to a quality inspection data model, quality inspection factors and a quality inspection strategy preset by the system; the method comprises the steps that a preset quality inspection factor item is initially changed, a model in a quality inspection standard template is related according to the factor item, a grading standard is preset in the model, and a quality inspection processor analyzes and calculates the value of the model according to a value calculation chain in a quality inspection chain and on the basis of a genim open source toolkit according to a quality inspection standard template class. The method comprises the steps of importing a pre-prepared industry service corpus into a language model of genim, producing expected word bags by the Gensim according to the imported corpus, enabling each word in each word bag to correspond to a unique id, cutting each sentence in the corpus by utilizing jieba participles, producing word bags by a cut collection, forming a key, namely a value dictionary, enabling elements in each list to have unique id marks, cutting the imported conversation text by a quality inspection processor, matching the cut participle collection with the elements in the word bags, and calculating corresponding scores according to matching results and standard word bags. Obtaining initial scores according to multi-dimensional analysis and grading of different score templates, after processing of each template class is finished, calculating an average value according to each initial score to obtain a final quality inspection total score, finishing calculation of the total score, after execution of a current score chain is finished, extracting early warning sensitive words and customer intention information according to an information processing chain, matching the sensitive words of a preset sensitive word bag in a customer service object text according to an kmp algorithm, sending a matched set to a wind control system through a message system for early warning, wherein the customer information extraction refers to matching keywords of a customer intention keyword library according to a kmp algorithm, and extracting data.
And step 3: generating a quality inspection report, generating a related quality inspection report for each call bill information after the quality inspection manager finishes quality inspection of data, wherein the quality inspection report shows a current call bill quality inspection result, the call bill quality inspection result comprises score information, a total score, a sensitive word set and customer intention keywords of three quality inspection standard item templates, and the system performs cluster analysis according to batch call bills to generate a trend graph of initial score and total score change of customer service at stage time and the triggering times of sensitive words, so that a supervisor can conveniently know service quality change of the customer service at stage time and know the quality condition of the customer service information in multiple directions;
and 4, step 4: generating knowledge base script information, wherein the knowledge base script information is preset by a knowledge base, collecting the call bill texts which reach the system preset evaluation level standard into a high-quality call bill base table according to the call bill quality inspection result, carrying out secondary analysis on the natural language texts in the high-quality call bill base table, refining the dialogues in the customer service natural language texts through a quality inspection tool bag, storing the scripts generated by the natural language texts which reach the system preset evaluation level standard into the knowledge base, and enabling the scripts to enter the knowledge base after the scripts pass the audit of a supervisor.
And 5: customer service guidance flow maintenance, as shown in fig. 6, a customer guidance flow chart of an ASR-based customer service intelligent quality inspection analysis method, which files natural language text data and quality inspection reports after secondary analysis when the customer service meets a system preset evaluation level standard, records natural language texts after secondary analysis when the customer service meets the system preset evaluation level standard, and performs guidance flow maintenance according to information of selection, filtering and filing of the recorded natural language texts.
As shown in fig. 2, an embodiment based on the method of the present invention, an intelligent quality inspection analysis system for customer service based on ASR, includes a voice transcription system, an ASR system, a quality inspection system, a wind control system, and a background management system, which are connected to each other;
the voice transcription system regularly acquires voice files in batches, calls an ASR system to convert the voice files into natural language texts and stores the natural language texts into a database;
the ASR system belongs to a three-party open source interface and transcribes the voice into a natural text system;
the quality inspection system initializes quality inspection factor item information, presets a quality inspection strategy, analyzes text data transmitted by the voice transcription system in terms, evaluates text quality according to quality inspection item standards, extracts high-quality dialogue text content to generate knowledge base script information, and displays a quality inspection report to a background management system
The wind control system is used for receiving the service early warning information and sending the early warning information to the staff according to a preset early warning template
The background management system displays quality inspection item results, provides a manual quality inspection result inspection interface, optimizes customer service guiding flow, imports a knowledge base script, and establishes files for customer service and periodically archives.
The invention achieves the following beneficial effects:
objectivity: the assessment standards among different quality inspectors are unified, manual judgment is reduced, subjective influence is avoided, objective and fair effects are achieved, and meanwhile labor cost is saved.
And (3) specification: the method solves the limitation of the self service capability of the quality inspector, improves the assessment effect of the quality inspector, and eliminates the non-standard quality inspection of the quality inspection process by external factors.
Comprehensive: the method solves the limitation of quality inspection resources, seamlessly covers each customer service telephone, performs 100% full quality inspection, and effectively avoids the sampling mode from watching the telephone.
Real-time: the method solves the problem of lag of customer service quality inspection, monitors customer service calls in time, and finds and warns in time, so that the supervisor can find and solve problems in time.
The above embodiments do not limit the present invention in any way, and all other modifications and applications that can be made to the above embodiments in equivalent ways are within the scope of the present invention.
Claims (4)
1. An intelligent customer service quality inspection analysis method based on ASR is characterized by comprising the following steps:
step 1: generating a natural language text according with a quality inspection rule, wherein the quality inspection rule is a rule meeting a quality inspection data model of a quality inspection processor, the quality inspection processor consists of quality inspection data, a quality inspection strategy and a quality inspection factor, the quality inspection data model consists of a quality inspection object, a natural language text and an object text relation, the quality inspection object comprises customer service and a client, the natural language text is the text content of the conversation between the customer service and the client, the object text relation comprises the text relation corresponding to the customer service and the client, the quality inspection rule comprises conversation voice, the conversation voice comprises a bill attribute, a conversation object and a conversation object, the bill attribute comprises a main bill key related to the customer service, and the main bill key comprises: the method comprises the following steps of binding a ticket and a customer service by using a unique identifier, voice starting time, voice ending time, voice call duration, voice connection waiting duration and customer evaluation; the call object is a calling object; the called object is a called object; the natural language text is a natural language text which can be subjected to quality inspection and is generated by the voice transcription system according to the ASR system which automatically matches the recording file and supports transcription, the natural language text is split and packaged into a quality inspection data model, and a quality inspection processor performs quality inspection;
step 2, intelligent quality inspection, presetting a quality inspection strategy and quality inspection factors into a quality inspection processor, dynamically analyzing quality inspection data in a quality inspection data model, carrying out multi-dimensional analysis according to the quality inspection strategy and the quality inspection factors, and generating a quality inspection result;
and step 3: generating a quality inspection report, generating a related quality inspection report for each call bill information after the quality inspection manager finishes quality inspection of data, wherein the quality inspection report shows a current call bill quality inspection result, the call bill quality inspection result comprises score information, a total score and a sensitive word set of three quality inspection standard item templates and client intention keywords, and a system carries out cluster analysis according to batch call bills to generate a service initial score, a trend graph of the total score change and the sensitive word triggering times of service at stage time;
and 4, step 4: generating knowledge base script information, wherein the knowledge base script information is preset by a knowledge base, a system collects call ticket texts reaching a system preset evaluation level standard into a high-quality call ticket base table according to a call ticket quality inspection result, performs secondary analysis on natural language texts in the high-quality call ticket base table, refines dialogues in customer service natural language texts through a quality inspection tool bag, and stores scripts generated by the natural language texts reaching the system preset evaluation level standard into the knowledge base;
and 5: and (4) customer service guide flow maintenance, namely filing the natural language text data and the quality inspection report which reach the preset evaluation level standard of the system and are subjected to secondary analysis, recording the natural language text which reaches the preset evaluation level standard of the system and is subjected to secondary analysis, and performing customer service guide flow maintenance according to the recorded information of selection, filtration and filing of the natural language text.
2. The ASR-based customer service intelligent quality inspection analysis method of claim 1, wherein: in step 1, the specific steps of generating the natural language text are as follows:
step 1-1, acquiring call data of a third party call center system according to call ticket id in a database table through a scheduling task according to a timing scheduling preset strategy, wherein the timing scheduling preset strategy is a time strategy for acquiring the call data of the third party call center system, an acquisition period is configured in a form of configuring a corner expression, the acquisition period is 1 hour or half a day, the call data corresponds to the call ticket id according to the call ticket data generated in real time in a call system by customer service, the call ticket data is data in the database table corresponding to the third party call center system, the call data is acquired from the database of the third party call center system according to the call ticket id, the call data meeting a quality inspection rule is synchronized and converted into a model needed by the system to be put in storage, a call ticket data synchronization identifier in the database table is updated to be 1, and for the call data not meeting the quality inspection rule, updating the ticket data synchronization identifier in the database table to 0 to indicate failure, recording the number of updating synchronization +1, wherein the synchronization threshold is 50, and compensating the acquired failed or non-generated voice data through a compensation mechanism, wherein the compensation mechanism is as follows: after the synchronous data flow is executed for the first time in each scheduling, the customer service call ticket data with the call ticket data synchronous identification of 0 and the synchronous times of no more than the threshold value in the database is obtained again, the call data is synchronized from the call center database, and an abnormal label is marked for the synchronous failure data exceeding the threshold value;
step 1-2: the voice file intelligently converts a natural language text, the ASR system is asynchronously called to perform voice conversion according to the call ticket data which is successfully synchronized in the steps, each call ticket data can independently open one thread, the data conversion between the threads is not mutually influenced, the voice conversion can be rapidly completed, the converted text is analyzed and packaged according to a quality inspection data model, the quality inspection data model is matched according to quality inspection factor items in a quality inspection processor, and the packaged quality inspection data model is stored in a database.
3. The ASR-based customer service intelligent quality inspection analysis method of claim 2, wherein: in the step 2, the intelligent quality inspection is specifically performed according to the following steps:
step 2-1: presetting quality inspection factors, wherein the quality inspection factors are divided into three different attribute dimensions by factor items: the system comprises a data factor, a client factor and a customer service factor, wherein the data factor comprises call starting time, call ending time, call duration, client queuing duration and a service queue selected by a client; the client factor comprises client intention keywords and client evaluation; the customer service factors comprise customer service expression capacity, service expression, service attitude, opening white, statement words, closing words and bad sensitive words;
step 2-2: presetting a quality inspection strategy, wherein the quality inspection strategy consists of a quality inspection standard item template and a quality inspection chain, more than one quality inspection standard item template is arranged, the quality inspection standard item template determines the scoring standard of specific dimensionality by factor items, calculates scores of different dimensionalities according to the attributes and rules of the quality inspection standard item template, the quality inspection chain comprises a score calculation chain and an information processing chain, wherein the score calculation chain is a calculation chain formed by quality inspection score rules of quality inspection standard item templates of different classes, the information processing chain comprises a sensitive word template and a client intention information extraction template, the sensitive word template is matched with the natural language text content of a client object in a data object according to a preset sensitive word bag to obtain malicious words in the sensitive word bag, a sensitive word early warning mechanism is triggered in real time, the customer intention information extraction template extracts customer intention product keywords or question keywords according to keywords in a customer intention keyword library table;
step 2-3: configuring a quality inspection toolkit and a matching algorithm, wherein the quality inspection process specifically comprises the following steps: matching, analyzing and quality testing the quality testing data, and performing quality testing on the generated natural language text, wherein the quality testing tool package is a Gensim tool package, the quality testing tool package is centrally installed on a server by using an anaconda tool to process the natural language text, the text content is matched according to sensitive words in sensitive word bags by using kmp matching algorithm, and the quality testing tool package and kmp matching algorithm are automatically initialized by the system;
step 2-4: the quality inspection processor performs quality inspection, and performs multi-dimensional quality inspection grading, early warning of sensitive words and extraction of client intention keywords according to a quality inspection data model, quality inspection factors and a quality inspection strategy preset by the system; the method comprises the steps that a preset quality inspection factor item is initially changed, a model in a quality inspection standard template is related according to the factor item, a grading standard is preset in the model, and a quality inspection processor analyzes and calculates a score based on a genim open source toolkit from a score calculation chain in a quality inspection chain sequentially according to a quality inspection standard template class; importing a pre-prepared industry service corpus into a language model of genim, producing expected word bags by the Gensim according to the imported corpus, wherein each word in each word bag corresponds to a unique id, cutting each sentence in the corpus by utilizing jieba participle, producing word bags by a cut collection, forming a key, namely a value dictionary, elements in each list have unique id marks, cutting the imported conversation text by the quality inspection processor, matching the cut participle collection with the elements in the word bags, and calculating corresponding scores according to a matching result and standard word bags; obtaining initial scores according to multi-dimensional analysis and grading of different score templates, after processing of each template class is finished, calculating an average value according to each initial score to obtain a final quality inspection total score, finishing calculation of the total score, after execution of a current score chain is finished, extracting early warning sensitive words and customer intention information according to an information processing chain, matching the sensitive words of a preset sensitive word bag in a customer service object text according to an kmp algorithm, sending a matched set to a wind control system through a message system for early warning, wherein the customer information extraction refers to matching keywords of a customer intention keyword library according to a kmp algorithm, and extracting data.
4. The ASR-based customer service intelligent quality inspection analysis method of claim 3, characterized in that: in the step 2-2: the quality inspection standard item template comprises an etiquette standard template, a flow standard template and a service skill template; each quality inspection standard item template comprises more than one basic standard item, each basic standard item has a corresponding grading level which corresponds to A, B + and B, C, D, E grades respectively, each grade corresponds to a ten-degree score, the quality inspection processor analyzes the dialog text corresponding to the customer service object in the data model according to a corpus preset by the basic standard items, and the corresponding score is given according to the matching degree of the analysis result and the grading model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011116925.4A CN112258016A (en) | 2020-10-19 | 2020-10-19 | Customer service intelligent quality inspection analysis method based on ASR |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011116925.4A CN112258016A (en) | 2020-10-19 | 2020-10-19 | Customer service intelligent quality inspection analysis method based on ASR |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112258016A true CN112258016A (en) | 2021-01-22 |
Family
ID=74245395
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011116925.4A Withdrawn CN112258016A (en) | 2020-10-19 | 2020-10-19 | Customer service intelligent quality inspection analysis method based on ASR |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112258016A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113361969A (en) * | 2021-07-07 | 2021-09-07 | 北京容联七陌科技有限公司 | Intelligent quality inspection system capable of flexibly configuring templates |
| CN113468296A (en) * | 2021-09-02 | 2021-10-01 | 杭州实在智能科技有限公司 | Model self-iteration type intelligent customer service quality inspection system and method capable of configuring business logic |
| CN113808574A (en) * | 2021-10-28 | 2021-12-17 | 平安消费金融有限公司 | AI voice quality inspection method, device, equipment and storage medium based on voice information |
| CN114049895A (en) * | 2022-01-12 | 2022-02-15 | 北京荣达天下信息科技有限公司 | ASR-based voice quality inspection analysis method and system |
| WO2022206719A1 (en) * | 2021-03-31 | 2022-10-06 | 北京智齿博创科技有限公司 | Outbound call failure result detection method based on freeswitch and asr |
| CN116361429A (en) * | 2023-01-19 | 2023-06-30 | 北京伽睿智能科技集团有限公司 | Business exception employee management method, system, equipment and storage medium |
| CN116980522A (en) * | 2023-09-22 | 2023-10-31 | 湖南三湘银行股份有限公司 | System and method for notifying customer image based on intelligent quality inspection |
-
2020
- 2020-10-19 CN CN202011116925.4A patent/CN112258016A/en not_active Withdrawn
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022206719A1 (en) * | 2021-03-31 | 2022-10-06 | 北京智齿博创科技有限公司 | Outbound call failure result detection method based on freeswitch and asr |
| CN113361969A (en) * | 2021-07-07 | 2021-09-07 | 北京容联七陌科技有限公司 | Intelligent quality inspection system capable of flexibly configuring templates |
| CN113361969B (en) * | 2021-07-07 | 2022-03-08 | 北京容联七陌科技有限公司 | Intelligent quality inspection system capable of flexibly configuring templates |
| CN113468296A (en) * | 2021-09-02 | 2021-10-01 | 杭州实在智能科技有限公司 | Model self-iteration type intelligent customer service quality inspection system and method capable of configuring business logic |
| CN113468296B (en) * | 2021-09-02 | 2022-05-03 | 杭州实在智能科技有限公司 | Model self-iteration type intelligent customer service quality inspection system and method capable of configuring business logic |
| CN113808574A (en) * | 2021-10-28 | 2021-12-17 | 平安消费金融有限公司 | AI voice quality inspection method, device, equipment and storage medium based on voice information |
| CN114049895A (en) * | 2022-01-12 | 2022-02-15 | 北京荣达天下信息科技有限公司 | ASR-based voice quality inspection analysis method and system |
| CN116361429A (en) * | 2023-01-19 | 2023-06-30 | 北京伽睿智能科技集团有限公司 | Business exception employee management method, system, equipment and storage medium |
| CN116361429B (en) * | 2023-01-19 | 2024-02-02 | 北京伽睿智能科技集团有限公司 | Business exception employee management method, system, equipment and storage medium |
| CN116980522A (en) * | 2023-09-22 | 2023-10-31 | 湖南三湘银行股份有限公司 | System and method for notifying customer image based on intelligent quality inspection |
| CN116980522B (en) * | 2023-09-22 | 2024-01-09 | 湖南三湘银行股份有限公司 | System and method for notifying customer image based on intelligent quality inspection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112258016A (en) | Customer service intelligent quality inspection analysis method based on ASR | |
| CN112804400B (en) | Customer service call voice quality inspection method and device, electronic equipment and storage medium | |
| US11004013B2 (en) | Training of chatbots from corpus of human-to-human chats | |
| CN107093431B (en) | Method and device for quality inspection of service quality | |
| CN109327632A (en) | Intelligent quality inspection system, method and the computer readable storage medium of customer service recording | |
| CN113468296B (en) | Model self-iteration type intelligent customer service quality inspection system and method capable of configuring business logic | |
| CN111917924A (en) | Customer service voice quality inspection method and related equipment | |
| CN111522947A (en) | Method and system for processing complaint work order | |
| TW201935370A (en) | System and method for evaluating customer service quality from text content | |
| CN109389418A (en) | Electric service client's demand recognition methods based on LDA model | |
| CN110598008A (en) | Data quality inspection method and device for recorded data and storage medium | |
| US20230252978A1 (en) | Systems and methods of material compliance | |
| CN111835925A (en) | Off-line voice quality inspection and analysis system for call center | |
| CN115062117A (en) | Method for automatically generating and classifying documents based on natural language processing technology | |
| CN110931016A (en) | Voice recognition method and system for offline quality inspection | |
| CN113037934A (en) | Hot word analysis system based on call recording of call center | |
| CN116828109A (en) | Intelligent evaluation method and system for telephone customer service quality | |
| CN113837545A (en) | Electric power customer service system with real-time monitoring service quality | |
| CN113505606B (en) | Training information acquisition method and device, electronic equipment and storage medium | |
| CN114492436A (en) | Processing method, device and system for auditing interview information | |
| US20240162743A1 (en) | Power Demand Side Speech Interaction Method and System | |
| CN111508529A (en) | Dynamic extensible voice quality inspection scoring method | |
| CN115080732A (en) | Complaint work order processing method and device, electronic equipment and storage medium | |
| CN111881272A (en) | Real-time early warning system and method for telephone service | |
| CN111756938A (en) | Call center telephone intelligent analysis monitoring method and system |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210122 |