CN112036705A - Quality inspection result data acquisition method, device and equipment - Google Patents
Quality inspection result data acquisition method, device and equipment Download PDFInfo
- Publication number
- CN112036705A CN112036705A CN202010779770.6A CN202010779770A CN112036705A CN 112036705 A CN112036705 A CN 112036705A CN 202010779770 A CN202010779770 A CN 202010779770A CN 112036705 A CN112036705 A CN 112036705A
- Authority
- CN
- China
- Prior art keywords
- quality
- tested
- file
- text
- quality inspection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000007689 inspection Methods 0.000 title claims abstract description 86
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000001514 detection method Methods 0.000 claims abstract description 113
- 238000005516 engineering process Methods 0.000 claims abstract description 30
- 238000013135 deep learning Methods 0.000 claims abstract description 19
- 238000012360 testing method Methods 0.000 claims abstract description 14
- 239000013598 vector Substances 0.000 claims description 34
- 230000008451 emotion Effects 0.000 claims description 26
- 238000004422 calculation algorithm Methods 0.000 claims description 25
- 238000004458 analytical method Methods 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 14
- 238000013145 classification model Methods 0.000 claims description 12
- 238000010801 machine learning Methods 0.000 claims description 11
- 239000011159 matrix material Substances 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 7
- 230000004044 response Effects 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 230000001902 propagating effect Effects 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 abstract description 17
- 238000003058 natural language processing Methods 0.000 abstract description 13
- 230000008569 process Effects 0.000 description 24
- 238000010586 diagram Methods 0.000 description 13
- 230000006870 function Effects 0.000 description 6
- 238000012986 modification Methods 0.000 description 6
- 230000004048 modification Effects 0.000 description 6
- 238000004590 computer program Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000013473 artificial intelligence Methods 0.000 description 4
- 230000006854 communication Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000008909 emotion recognition Effects 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000009191 jumping Effects 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013016 damping Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 238000004579 scanning voltage microscopy Methods 0.000 description 1
- 238000013519 translation Methods 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
- G06Q10/06398—Performance of employee with respect to a job function
-
- 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
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06Q30/01—Customer relationship services
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/26—Speech to text systems
Abstract
The invention discloses a quality inspection result data acquisition method, device and equipment, and belongs to the technical field of natural language processing and deep learning. The method comprises the following steps: acquiring a voice file to be quality tested; converting the voice file to be subjected to quality inspection into a text to be subjected to quality inspection by an automatic voice recognition technology; and determining a file to be quality-tested corresponding to the selected detection dimension, analyzing and calculating to obtain quality-test result data, wherein the file to be quality-tested comprises the voice file to be quality-tested and the text to be quality-tested. According to the invention, the quality inspection result data is obtained by utilizing the corresponding automatic voice recognition technology, the natural language processing technology and the deep learning technology to replace manual quality inspection, and detection mechanisms such as keyword detection and the like are introduced, so that the quality inspection efficiency is comprehensively improved, and the labor cost is reduced.
Description
Technical Field
The invention relates to the technical field of natural language processing and deep learning, in particular to a quality inspection result data acquisition method, device and equipment.
Background
The traditional customer service quality inspection method adopts a manual sampling inspection mode, and a quality inspector mainly listens to customer service recording records, records the recording results in a form and analyzes the data of the form so as to judge whether the quality of the inspected customer service meets the requirements. However, the quality inspection in this way is heavy in workload, low in efficiency and coverage rate, and difficult to effectively evaluate the overall service quality. For example, in the scenario of collection hastening in the financial industry, collection hastening is an important means for returning funds of modern financial companies, manual collection hastening still occupies an irreplaceable position in collection hastening, but the collection hastening members are not good enough in good quality and are easy to make mistakes in the process of communicating with customers, the trial and error of each collection hastening member is damage to the enterprise image, and meanwhile, the collection hastening effect is not good, so that quality inspection needs to be performed on the collection hastening process, the management and control of collection hastening are enhanced, but after all, the traditional manual quality inspection efficiency is too low, and the requirements are increasingly difficult to meet.
Natural Language Processing (NLP) is a field of computer science, artificial intelligence, linguistics that focuses on the interaction between computers and human (natural) language. Natural language processing is an important direction in the fields of computer science and artificial intelligence, and it is an important direction to study various theories and methods that can realize effective communication between people and computers using natural language. Natural language processing is a science integrating linguistics, computer science and mathematics. Therefore, the research in this field will relate to natural language, i.e. the language that people use everyday, so it is closely related to the research of linguistics, but has important difference. Natural language processing is not a general study of natural language but is directed to developing computer systems, and particularly software systems therein, that can efficiently implement natural language communications, and thus is part of computer science.
Deep Learning (DL) is a new research direction in the field of Machine Learning (ML), which is introduced into Machine Learning to make it closer to the original target, Artificial Intelligence (AI). Deep learning is the intrinsic law and expression level of the learning sample data, and the information obtained in the learning process is very helpful for the interpretation of data such as characters, images and sounds. The final aim of the method is to enable the machine to have the analysis and learning capability like a human, and to recognize data such as characters, images and sounds. Deep learning is a complex machine learning algorithm, and achieves the effect in speech and image recognition far exceeding the prior related art. Deep learning has achieved many achievements in search technology, data mining, machine learning, machine translation, natural language processing, multimedia learning, speech, recommendation and personalization technologies, and other related fields. The deep learning enables the machine to imitate human activities such as audio-visual and thinking, solves a plurality of complex pattern recognition problems, and makes great progress on the artificial intelligence related technology.
Disclosure of Invention
In order to solve the problems in the prior art, embodiments of the present invention provide a quality inspection result data obtaining method, apparatus, and device, in which corresponding automatic speech recognition technology, natural language processing technology, and deep learning technology are used to obtain quality inspection result data to replace manual quality inspection, and detection mechanisms such as keyword detection are introduced, so that quality inspection efficiency is comprehensively improved, and labor cost is reduced. The technical scheme is as follows:
in one aspect, a method for acquiring quality inspection result data is provided, the method comprising:
acquiring a voice file to be quality tested;
converting the voice file to be subjected to quality inspection into a text to be subjected to quality inspection by an automatic voice recognition technology;
and determining a file to be quality-tested corresponding to the selected detection dimension, analyzing and calculating to obtain quality-test result data, wherein the file to be quality-tested comprises the voice file to be quality-tested and the text to be quality-tested.
Preferably, determining the file to be quality-tested corresponding to the selected detection dimension for analysis and calculation, and obtaining quality-test result data, includes:
extracting corresponding keywords from the text to be tested through a TextRank algorithm or a TF-IDF algorithm, and determining a keyword detection result according to a preset keyword comparison rule.
Preferably, extracting corresponding keywords from the text to be quality-checked through a preset TextRank algorithm or a preset TF-IDF algorithm, includes:
segmenting the text to be quality tested to obtain a plurality of sentences;
performing word segmentation and part-of-speech tagging and stop word filtering on each sentence, and only retaining words with specified part-of-speech so as to obtain candidate words;
constructing a candidate keyword graph according to the node set formed by the candidate words and the relationship between the nodes;
iteratively propagating the weight of each node through a corresponding calculation formula until convergence;
the weights of all the nodes are sorted in a reverse order, and a preset number of candidate keywords are obtained according to a preset screening rule;
and marking the text to be subjected to quality inspection according to the candidate keywords.
Preferably, determining the file to be quality-tested corresponding to the selected detection dimension for analysis and calculation, and obtaining quality-test result data, includes:
and constructing a forbidden word tree by determining a finite automaton algorithm, and matching the text to be quality tested with the forbidden word tree to determine a forbidden word detection result.
Preferably, the forbidden word tree is constructed by determining a finite automaton algorithm, which comprises:
constructing tree nodes, wherein the nodes are null;
searching predetermined forbidden words from each level of nodes, and if the searching result is existence, continuing searching; and if the search result is not existed, constructing a new sub-tree until the search of the last forbidden word is finished.
Preferably, determining the file to be quality-tested corresponding to the selected detection dimension for analysis and calculation, and obtaining quality-test result data, includes:
and converting the text to be tested into vector data to be tested, inputting the vector data to be tested into a pre-trained deep learning classification model, and obtaining an emotion detection result.
Preferably, the converting the text to be quality-checked into vector data to be quality-checked, inputting the vector data to be quality-checked into a trained deep learning classification model, and obtaining an emotion detection result, including:
preparing a corpus marked with emotion classification, preprocessing the corpus, and converting a sentence of corpus into a vocabulary set;
converting the vocabulary set into a word vector set in a matrix form;
inputting the word vector set into a preset LTSM network model to obtain a sentence vector;
inputting the sentence vectors into a preset machine learning classification model for supervised training until the model converges;
and re-executing the following steps on the text to be quality checked: preparing linguistic data marked with emotion classification, preprocessing the linguistic data, converting a sentence of the linguistic data into a vocabulary set, and repeating the steps to obtain an emotion detection result.
Preferably, determining the file to be quality-tested corresponding to the selected detection dimension for analysis and calculation, and obtaining quality-test result data, includes:
judging whether the time difference between the customer service response time and the time when the customer speaks to finish in the voice file to be quality tested exceeds a preset threshold value or not, and obtaining a mute detection result; and/or the presence of a gas in the gas,
calculating the average speech speed of customer service and the speech speed of a single sentence according to the voice file to be quality checked in a statistical manner, and obtaining a speech speed detection result; and/or the presence of a gas in the gas,
and judging whether time overlap exists between the customer service single sentence time period and the customer single sentence time period according to the voice file to be quality tested, and obtaining a voice overlap detection result.
In another aspect, there is provided a quality inspection result data acquiring apparatus including:
a data acquisition module to: acquiring a voice file to be quality tested;
a file conversion module to: converting the voice file to be subjected to quality inspection into a text to be subjected to quality inspection by an automatic voice recognition technology;
an analysis computation module to: and determining a file to be quality-tested corresponding to the selected detection dimension, analyzing and calculating to obtain quality-test result data, wherein the file to be quality-tested comprises the voice file to be quality-tested and the text to be quality-tested.
In still another aspect, there is provided a quality inspection result data acquiring apparatus including:
a processor;
a memory for storing executable instructions of the processor;
wherein the processor is configured to execute the steps of the quality inspection result data acquisition method according to any one of the above aspects via the executable instructions.
The technical scheme provided by the embodiment of the invention has the following beneficial effects:
1. converting the voice file to be inspected into a text to be inspected by an automatic voice recognition technology; then, analyzing and calculating the voice file to be quality-checked or the text to be quality-checked from at least one detection dimension of a keyword detection dimension, a forbidden word detection dimension, an emotion detection dimension, a silence detection dimension, a speech speed detection dimension and a speech overlapping detection dimension so as to obtain quality-check result data, and realizing a quality-check process by using a preset automatic speech recognition technology, an NLP technology and a deep learning technology to replace manual quality check, so that the quality-check efficiency is improved on the whole, and the labor cost is reduced;
2. by introducing a keyword detection mechanism, whether the customer service covers a key process or not can be detected quickly, the quality inspection coverage rate and the automation rate are improved, a large number of forbidden words can be detected quickly by detecting the forbidden words, the detection quality is improved, the emotion detection mechanism adopts a corresponding emotion recognition algorithm, more semantic information is kept by dimension reduction operation, the algorithm accuracy is improved, and the quality inspection detection quality and the detection efficiency are further improved by a silence detection mechanism, a speech speed detection mechanism and a speech overlapping detection mechanism.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a flowchart of a quality inspection result data acquisition method according to embodiment 1 of the present invention;
FIG. 2 is a flow diagram of sub-steps of step 103 of FIG. 1;
FIG. 3 is a flowchart illustrating example substeps 1032 of FIG. 2;
FIG. 4 is an example of a matching process between a text to be inspected and a forbidden word tree;
FIG. 5 is a flowchart illustrating exemplary substeps 1033 of FIG. 2;
fig. 6 is a schematic structural diagram of a quality inspection result data acquisition apparatus according to embodiment 2 of the present invention;
fig. 7 is a schematic structural diagram of quality inspection result data acquisition equipment provided in embodiment 3 of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention. In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
According to the quality inspection result data acquisition method, device and equipment provided by the embodiment of the invention, the voice file to be quality inspected is converted into the text to be quality inspected through an automatic voice recognition technology according to the acquired voice file to be quality inspected; then determining the files to be inspected corresponding to the selected inspection dimension for analysis and calculation, thereby obtaining the data of the inspection result, realizing the inspection process by utilizing the preset automatic speech recognition technology, NLP technology and deep learning technology, replacing manual quality inspection, improving the efficiency of quality inspection as a whole, reducing the labor cost, particularly, whether the customer service covers the key process or not can be quickly detected through the introduced keyword detection mechanism, the quality inspection coverage rate and the automation rate are improved, a large number of forbidden words can be rapidly detected through forbidden word detection, the detection quality is improved, and the emotion detection mechanism adopts a corresponding emotion recognition algorithm, the dimensionality reduction operation keeps more semantic information, the algorithm accuracy is improved, and the quality detection quality and the detection efficiency are further improved through a mute detection mechanism, a speech speed detection mechanism and a voice overlapping detection mechanism. Therefore, the quality inspection result data acquisition method, the quality inspection result data acquisition device and the quality inspection result data acquisition equipment are widely applied to various customer service quality inspection scenes in the fields of financial collection-promoting quality inspection, network commercial after-sales service and the like.
The following describes in detail a method, an apparatus, and a device for acquiring quality inspection result data according to an embodiment of the present invention with reference to specific embodiments and drawings.
Example 1
Fig. 1 is a flowchart of a quality inspection result data acquisition method according to embodiment 1 of the present invention, and as shown in fig. 1, the quality inspection result data acquisition method according to this embodiment includes the following steps:
101. and acquiring a voice file to be quality tested.
The voice file to be inspected can be a recording file or other forms of voice files collected by the customer service process.
102. And converting the voice file to be inspected into a text to be inspected by an automatic voice recognition technology.
Preferably, the audio file to be tested is converted into the text to be tested with a time stamp by an Automatic Speech Recognition (ASR) technology, and any other technology possible in the prior art may also be used to implement the conversion process between the above-mentioned speech texts, and the embodiment of the present invention is not particularly limited thereto.
103. And determining the files to be tested corresponding to the selected detection dimension, analyzing and calculating to obtain quality test result data, wherein the files to be tested comprise voice files to be tested and texts to be tested.
Preferably, as shown in fig. 2, step 103 includes a sub-step 1031 of obtaining quality inspection result data from a keyword detection dimension, that is, extracting corresponding keywords from the text to be inspected by a TextRank algorithm or a TF-IDF algorithm, and determining a keyword detection result according to a preset keyword comparison rule.
Further preferably, the 1031 substep is implemented as the following process:
1031a, and segmenting the text to be quality-tested to obtain a plurality of sentences. Illustratively, converting the recording of the person in chargeThe text T is segmented according to the complete sentence, i.e. T ═ S1,S2,…,Sm]Wherein S represents a sentence, and m is an integer greater than 1.
1031b, performing word segmentation and part-of-speech tagging processing and stop word filtering on each sentence, and reserving only words with specified part-of-speech, such as nouns, verbs, adjectives and the like, to obtain candidate words.
1031c, constructing a candidate keyword graph according to the node set formed by the candidate words and the relationship among the nodes. Exemplarily, constructing a candidate keyword graph G ═ V, E, where V is a node set and is composed of the candidate keywords generated in (2), and E is a relationship between nodes, and constructing an edge between any two points using a co-occurrence relationship (co-occurrence), and an edge exists between two nodes only when their corresponding words are co-occurring in a window with a length of K, where K denotes a window size, that is, at most K words co-occur.
1031d, iteratively propagating the weight of each node through a corresponding calculation formula until convergence. Illustratively, the weights of the nodes are propagated iteratively until convergence, according to the following formula:
S(vi) Is node viI.e. the weight of the word, out (v)j) Is a node vjIs connected to the successor node of wjiD is a damping coefficient, and i and j are integers greater than or equal to 1.
1031e, sorting the weights of the nodes in a reverse order, and obtaining a predetermined number of candidate keywords according to a preset screening rule. The preset screening rules and the preset quantity can be correspondingly set according to actual needs. Illustratively, the node weights are sorted in reverse order, so as to obtain the most important t words as candidate keywords.
1031f, marking the text to be quality-checked according to the candidate keywords. Illustratively, according to the obtained most important t words, marking is carried out in the original text, and if adjacent phrases are formed, the words are combined into a multi-word keyword.
It should be noted that, the processes of step 103 and sub-step 1031 may be implemented in other ways besides the ways described in the above steps, and the specific ways are not limited in the embodiment of the present invention.
Preferably, as shown in fig. 2, the step 103 further includes a substep 1032 of obtaining quality inspection result data from the forbidden word detection dimension, that is, constructing a forbidden word tree by determining a finite automata algorithm, and matching the text to be quality inspected with the forbidden word tree to determine a forbidden word detection result.
Further preferably, as shown in fig. 3, the above sub-step 1032 is implemented as the following process:
1032a, building tree nodes, wherein the nodes are null;
1032b, searching predetermined forbidden words from each level of nodes, and if the searching result is present, continuing searching; and if the search result is not existed, constructing a new sub-tree until the search of the last forbidden word is finished. Exemplarily, searching the ith word Wi of the forbidden word from a root node of the tree, if the ith word Wi does not exist, proving that the forbidden word starting with Wi does not exist, directly constructing the subtree, and jumping to a judgment step of searching end: and judging whether the word is the last word in the word, if so, setting a flag bit isEnd to be 1, otherwise, setting a flag bit isEnd to be 0, if the word is found in the tree, indicating that the forbidden word starting with Wi exists, repeating the step 1032b, and then jumping to the judgment step of finding the end until one forbidden word is finished.
The following describes an exemplary process of matching the text to be quality-checked with the forbidden word tree in a financial collection quality-checking scenario. If the recorded text is fool, firstly, matching is started from a node 0 in fig. 4, the first character is matched to be fool, the node 1 is jumped to, then the second character is matched to be fool, the node 3 is jumped to, the fool in the recorded text can be matched to the forbidden word tree, and the word is judged to be the forbidden word.
It should be noted that, the processes of step 103 and sub-step 1032 may be implemented in other ways besides the ways described in the above steps, and the specific ways are not limited in the embodiment of the present invention.
Preferably, as shown in fig. 2, the step 103 further includes a substep 1033 of obtaining quality inspection result data from the emotion detection dimension, that is, converting the text to be quality inspected into vector data to be quality inspected, and inputting the vector data to be quality inspected into a pre-trained deep learning classification model to obtain an emotion detection result.
Further preferably, as shown in fig. 5, the above-mentioned 1033 sub-step is implemented as the following process:
1033a, preparing a corpus marked with emotion classification, preprocessing the corpus, and converting a sentence corpus into a vocabulary set. Illustratively, corpus preprocessing includes word segmentation and word decommissioning, such that a sentence corpus S is converted into sets of words [ W [ ]1,W2,…,Wm]。
1033b, converting the vocabulary set into a word vector set in a matrix form. Illustratively, the vocabulary set is subjected to vector conversion and converted into a word vector set [ V ]1,V2,…,Vm]Where V is a word vector of length n [ Vw1,Vw2,…,Vwm]In practice, the set of word vectors is an n × m matrix.
1033c, inputting the word vector set into a preset LTSM network model to obtain sentence vectors. Illustratively, a word vector set is fed into a preset LTSM network model, and an input in the form of an n × m matrix is encoded into a lower-dimensional one-dimensional sentence vector Vs。
1033d, inputting the sentence vectors into a preset machine learning classification model for supervised training until the model converges. Illustratively, a sentence vector VsAnd sending the model into a preset machine learning classification model, and carrying out supervised training until the model converges. The preset machine learning classification model may adopt decision trees, random forests, logistic regression, SVM, XGBoost, deep learning models (CNN, RNN, etc.), and the like.
1033e, re-executing the following steps on the text to be quality-checked: preparing linguistic data marked with emotion classification, preprocessing the linguistic data, converting a sentence of the linguistic data into a vocabulary set, and repeating the steps to obtain an emotion detection result. Illustratively, the step 5 is repeated on the text to be quality checked of the customer service staff to obtain the emotion classification result.
For example, if the recorded text is "please pay at tomorrow". Firstly, remove stop words and participles, change the text into word set "[ tomorrow, on time, repayment]", respectively convert 3 words into a word vector [ V ]11,V12,…,V1m][V21,V22,…,V2m][V31,V32,…,V3m]In practice, it is a 3 × m matrix. Sending the matrix into LTSM network, converting into one-dimensional sentence vector Vs1,Vs2,…,Vsm]. And predicting the emotion of the recorded text by adopting a machine learning classification model and taking the sentence vector as input.
The recorded text is in the form:
……
SentenceN,role:”RoleN”,startTime:”startTimeN”,endTime:”endTimeN”
It should be noted that, the processes of step 103 and sub-step 1033 may be implemented in other ways besides the ways described in the above steps, and the specific ways are not limited in the embodiment of the present invention.
Preferably, as shown in fig. 2, the step 103 further includes 1034 substep of obtaining quality test result data from the silence detection dimension, 1035 substep of obtaining quality test result data from the speech rate detection dimension, and 1036 substep of obtaining quality test result data from the speech overlap detection dimension.
Specifically, sub-step 1034 is implemented as the following process: and judging whether the time difference between the customer service response time and the time when the customer speaks to finish in the voice file to be subjected to quality inspection exceeds a preset threshold value or not, and acquiring a mute detection result. Illustratively, the silence detection calculates the time difference between the response time of the receiver and the speaking finish time of the client, and if the time difference exceeds a preset threshold value, the receiver does not respond in time. In the above recorded text, the time when the client finishes speaking in the sequence 1 is 10, the response time of the person who asks for the speech in the sequence 2 is 15, the time difference is 5, and if the predetermined threshold is 3, the silence detection determines that the person who asks for the speech does not respond in time.
1035 substep is implemented as the following process: and (4) calculating the customer service average speech speed and the single sentence speech speed according to the speech file to be quality tested, and obtaining a speech speed detection result. Illustratively, the speech rate detection counts the average speech rate and the single sentence speech rate of the learner. The single sentence speech rate is calculated as follows:
wherein, Count is the number of Chinese characters in a single sentence, and T is the time spent in the single sentence. The average speech rate is calculated as follows:
using the recorded text above as an example, the prompt says that the Senentce 2 single Sentence speed is
The 1036 substep is implemented as the following procedure: and judging whether time overlap exists between the customer service single sentence time period and the customer single sentence time period according to the voice file to be quality checked, and acquiring a voice overlap detection result. Illustratively, the voice overlap detection is to detect whether a single sentence time period between an acquirer and a client has time overlap, and if so, the voice overlap detection indicates that a call snatching phenomenon exists. For the above recorded text, the speaking time of the client in the sequence 3 is 20-30 seconds, and the speech of the driver in the sequence 4 is started in 25 seconds, so that the problem of speech snatching obviously exists.
It should be noted that the processes of step 103 and sub-steps 1033, 1034 and 1035 may be implemented in other ways besides the ways described in the above steps, and the specific ways are not limited by the embodiments of the present invention.
Example 2
Fig. 6 is a schematic structural diagram of a quality inspection result data acquisition apparatus according to embodiment 2 of the present invention. As shown in fig. 6, the quality inspection result data acquiring apparatus provided in this embodiment includes a data acquiring module 21, a file converting module 22, and an analyzing and calculating module 23. Specifically, the data obtaining module 21 is configured to: acquiring a voice file to be quality tested; a file conversion module 22, configured to: converting the voice file to be inspected into a text to be inspected by an automatic voice recognition technology; an analysis calculation module 23 configured to: and determining the files to be tested corresponding to the selected detection dimension, analyzing and calculating to obtain quality test result data, wherein the files to be tested comprise voice files to be tested and texts to be tested.
Preferably, the analysis and calculation module 23 further includes a keyword detection module 231, a forbidden word detection module 232, an emotion detection module 233, a silence detection module 234, a speech rate detection module 235, and a voice overlap detection module 236.
Specifically, the keyword detection module 231 is configured to obtain quality inspection result data from the keyword detection dimension, that is, extract corresponding keywords from the text to be quality inspected through a TextRank algorithm or a TF-IDF algorithm, and determine a keyword detection result according to a preset keyword comparison rule. The forbidden word detection module 232 is configured to obtain quality inspection result data from the forbidden word detection dimension, that is, a forbidden word tree is constructed by determining a finite automaton algorithm, and a text to be quality inspected is matched with the forbidden word tree to determine a forbidden word detection result. The emotion detection module 233 is configured to obtain quality detection result data from the emotion detection dimension, that is, convert the text to be quality detected into vector data to be quality detected, and input the vector data to be quality detected into the trained deep learning classification model to obtain an emotion detection result. The silence detection module 234 is configured to obtain quality inspection result data from a silence detection dimension, that is, determine whether a time difference between a customer service response time and a time when a customer speaks to be completed in a voice file to be quality inspected exceeds a preset threshold, and obtain a silence detection result. The speech rate detection module 235 is configured to obtain the quality inspection result data from the speech rate detection dimension, that is, statistically calculate the customer service average speech rate and the single sentence speech rate according to the speech file to be quality inspected, and obtain the speech rate detection result. The voice overlap detection module 236 is configured to obtain quality inspection result data from the voice overlap detection dimension, that is, determine whether there is time overlap between the customer service single sentence time period and the customer single sentence time period according to the voice file to be quality inspected, and obtain a voice overlap detection result.
Example 3
Fig. 7 is a schematic structural diagram of quality inspection result data acquisition equipment provided in embodiment 3 of the present invention. As shown in fig. 7, the quality inspection result data acquisition apparatus according to this embodiment includes: a processor 31; a memory 32 for storing executable instructions for the processor 31; wherein, the processor 31 is configured to execute the steps of the quality inspection result data acquisition method according to any one of the embodiments 1 via executable instructions (i.e. program). The processor 31 and the memory 32 realize communication processes through a communication bus.
It should be noted that: the quality inspection result data obtaining device and the quality inspection result data obtaining apparatus for the quality inspection result data obtaining service provided in the above embodiments are exemplified by only the division of the above functional modules when triggering the intelligent network service, and in practical applications, the function distribution may be completed by different functional modules as needed, that is, the internal structure of the device or the apparatus may be divided into different functional modules to complete all or part of the above described functions. In addition, the quality inspection result data acquisition device, and the quality inspection result data acquisition method provided by the embodiments belong to the same concept, and specific implementation processes thereof are described in the method embodiments and are not described herein again.
All the above-mentioned optional technical solutions can be combined arbitrarily to form the optional embodiments of the present invention, and are not described herein again.
In summary, the quality inspection result data acquisition method, apparatus and device provided in the embodiments of the present invention have the following beneficial effects, compared with the prior art:
1. converting the voice file to be inspected into a text to be inspected by an automatic voice recognition technology; then, analyzing and calculating the voice file to be quality-checked or the text to be quality-checked from at least one detection dimension of a keyword detection dimension, a forbidden word detection dimension, an emotion detection dimension, a silence detection dimension, a speech speed detection dimension and a speech overlapping detection dimension so as to obtain quality-check result data, and realizing a quality-check process by using a preset automatic speech recognition technology, an NLP technology and a deep learning technology to replace manual quality check, so that the quality-check efficiency is improved on the whole, and the labor cost is reduced;
2. by introducing a keyword detection mechanism, whether the customer service covers a key process or not can be detected quickly, the quality inspection coverage rate and the automation rate are improved, a large number of forbidden words can be detected quickly by detecting the forbidden words, the detection quality is improved, the emotion detection mechanism adopts a corresponding emotion recognition algorithm, more semantic information is kept by dimension reduction operation, the algorithm accuracy is improved, and the quality inspection detection quality and the detection efficiency are further improved by a silence detection mechanism, a speech speed detection mechanism and a speech overlapping detection mechanism.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
Embodiments of the present application are described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including the preferred embodiment and all changes and modifications that fall within the true scope of the embodiments of the present application.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (10)
1. A method for acquiring quality inspection result data, the method comprising:
acquiring a voice file to be quality tested;
converting the voice file to be subjected to quality inspection into a text to be subjected to quality inspection by an automatic voice recognition technology;
and determining a file to be quality-tested corresponding to the selected detection dimension, analyzing and calculating to obtain quality-test result data, wherein the file to be quality-tested comprises the voice file to be quality-tested and the text to be quality-tested.
2. The method according to claim 1, wherein determining the file to be inspected corresponding to the selected inspection dimension for analysis and calculation to obtain the data of the inspection result comprises:
extracting corresponding keywords from the text to be tested through a TextRank algorithm or a TF-IDF algorithm, and determining a keyword detection result according to a preset keyword comparison rule.
3. The method of claim 2, wherein extracting corresponding keywords from the text to be quality-checked through a preset TextRank algorithm or a preset TF-IDF algorithm comprises:
segmenting the text to be quality tested to obtain a plurality of sentences;
performing word segmentation and part-of-speech tagging and stop word filtering on each sentence, and only retaining words with specified part-of-speech so as to obtain candidate words;
constructing a candidate keyword graph according to the node set formed by the candidate words and the relationship between the nodes;
iteratively propagating the weight of each node through a corresponding calculation formula until convergence;
the weights of all the nodes are sorted in a reverse order, and a preset number of candidate keywords are obtained according to a preset screening rule;
and marking the text to be subjected to quality inspection according to the candidate keywords.
4. The method according to claim 1, wherein determining the file to be inspected corresponding to the selected inspection dimension for analysis and calculation to obtain the data of the inspection result comprises:
and constructing a forbidden word tree by determining a finite automaton algorithm, and matching the text to be quality tested with the forbidden word tree to determine a forbidden word detection result.
5. The method of claim 4, wherein constructing the forbidden word tree by determining a finite automaton algorithm comprises:
constructing tree nodes, wherein the nodes are null;
searching predetermined forbidden words from each level of nodes, and if the searching result is existence, continuing searching; and if the search result is not existed, constructing a new sub-tree until the search of the last forbidden word is finished.
6. The method according to claim 1, wherein determining the file to be inspected corresponding to the selected inspection dimension for analysis and calculation to obtain the data of the inspection result comprises:
and converting the text to be tested into vector data to be tested, inputting the vector data to be tested into a pre-trained deep learning classification model, and obtaining an emotion detection result.
7. The method according to claim 6, wherein converting the text to be tested into vector data to be tested, inputting the vector data to be tested into a trained deep learning classification model, and obtaining emotion detection results, comprises:
preparing a corpus marked with emotion classification, preprocessing the corpus, and converting a sentence of corpus into a vocabulary set;
converting the vocabulary set into a word vector set in a matrix form;
inputting the word vector set into a preset LTSM network model to obtain a sentence vector;
inputting the sentence vectors into a preset machine learning classification model for supervised training until the model converges;
and re-executing the following steps on the text to be quality checked: preparing linguistic data marked with emotion classification, preprocessing the linguistic data, converting a sentence of the linguistic data into a vocabulary set, and repeating the steps to obtain an emotion detection result.
8. The method according to claim 1, wherein determining the file to be inspected corresponding to the selected inspection dimension for analysis and calculation to obtain the data of the inspection result comprises:
judging whether the time difference between the customer service response time and the time when the customer speaks to finish in the voice file to be quality tested exceeds a preset threshold value or not, and obtaining a mute detection result; and/or the presence of a gas in the gas,
calculating the average speech speed of customer service and the speech speed of a single sentence according to the voice file to be quality checked in a statistical manner, and obtaining a speech speed detection result; and/or the presence of a gas in the gas,
and judging whether time overlap exists between the customer service single sentence time period and the customer single sentence time period according to the voice file to be quality tested, and obtaining a voice overlap detection result.
9. A quality inspection result data acquisition apparatus, comprising:
a data acquisition module to: acquiring a voice file to be quality tested;
a file conversion module to: converting the voice file to be subjected to quality inspection into a text to be subjected to quality inspection by an automatic voice recognition technology;
an analysis computation module to: and determining a file to be quality-tested corresponding to the selected detection dimension, analyzing and calculating to obtain quality-test result data, wherein the file to be quality-tested comprises the voice file to be quality-tested and the text to be quality-tested.
10. A quality inspection result data acquisition apparatus characterized by comprising:
a processor;
a memory for storing executable instructions of the processor;
wherein the processor is configured to perform the steps of the quality inspection result data acquisition method of any one of claims 1 to 8 via the executable instructions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010779770.6A CN112036705A (en) | 2020-08-05 | 2020-08-05 | Quality inspection result data acquisition method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010779770.6A CN112036705A (en) | 2020-08-05 | 2020-08-05 | Quality inspection result data acquisition method, device and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112036705A true CN112036705A (en) | 2020-12-04 |
Family
ID=73582391
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010779770.6A Pending CN112036705A (en) | 2020-08-05 | 2020-08-05 | Quality inspection result data acquisition method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112036705A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112885332A (en) * | 2021-01-08 | 2021-06-01 | 天讯瑞达通信技术有限公司 | Voice quality inspection method, system and storage medium |
| CN114157765A (en) * | 2022-01-19 | 2022-03-08 | 阳光财产保险股份有限公司 | Voice quality inspection method and device, electronic equipment and storage medium |
| WO2022126969A1 (en) * | 2020-12-15 | 2022-06-23 | 平安科技（深圳）有限公司 | Service voice quality inspection method, apparatus and device, and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106919673A (en) * | 2017-02-21 | 2017-07-04 | 浙江工商大学 | Text mood analysis system based on deep learning |
| CN111489765A (en) * | 2019-01-28 | 2020-08-04 | 国家电网有限公司客户服务中心 | Telephone traffic service quality inspection method based on intelligent voice technology |
-
2020
- 2020-08-05 CN CN202010779770.6A patent/CN112036705A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106919673A (en) * | 2017-02-21 | 2017-07-04 | 浙江工商大学 | Text mood analysis system based on deep learning |
| CN111489765A (en) * | 2019-01-28 | 2020-08-04 | 国家电网有限公司客户服务中心 | Telephone traffic service quality inspection method based on intelligent voice technology |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022126969A1 (en) * | 2020-12-15 | 2022-06-23 | 平安科技（深圳）有限公司 | Service voice quality inspection method, apparatus and device, and storage medium |
| CN112885332A (en) * | 2021-01-08 | 2021-06-01 | 天讯瑞达通信技术有限公司 | Voice quality inspection method, system and storage medium |
| CN114157765A (en) * | 2022-01-19 | 2022-03-08 | 阳光财产保险股份有限公司 | Voice quality inspection method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110096570B (en) | Intention identification method and device applied to intelligent customer service robot | |
| CN110110062B (en) | Machine intelligent question and answer method and device and electronic equipment | |
| CN108073568B (en) | Keyword extraction method and device | |
| CN112069811B (en) | Electronic text event extraction method with multi-task interaction enhancement | |
| CN110321563B (en) | Text emotion analysis method based on hybrid supervision model | |
| US8639517B2 (en) | Relevance recognition for a human machine dialog system contextual question answering based on a normalization of the length of the user input | |
| CN111143576A (en) | Event-oriented dynamic knowledge graph construction method and device | |
| CN112036705A (en) | Quality inspection result data acquisition method, device and equipment | |
| CN111858878B (en) | Method, system and storage medium for automatically extracting answer from natural language text | |
| CN111709630A (en) | Voice quality inspection method, device, equipment and storage medium | |
| Probierz et al. | Rapid detection of fake news based on machine learning methods | |
| CN112100377B (en) | Text classification method, apparatus, computer device and storage medium | |
| CN113672931B (en) | Software vulnerability automatic detection method and device based on pre-training | |
| RU2730449C2 (en) | Method of creating model for analysing dialogues based on artificial intelligence for processing user requests and system using such model | |
| Balli et al. | Sentimental analysis of Twitter users from Turkish content with natural language processing | |
| CN112287090A (en) | Financial question asking back method and system based on knowledge graph | |
| CN113326374B (en) | Short text emotion classification method and system based on feature enhancement | |
| CN112632272B (en) | Microblog emotion classification method and system based on syntactic analysis | |
| CN113486143A (en) | User portrait generation method based on multi-level text representation and model fusion | |
| CN112989001B (en) | Question and answer processing method and device, medium and electronic equipment | |
| CN115906824A (en) | Text fine-grained emotion analysis method, system, medium and computing equipment | |
| CN114841143A (en) | Voice room quality evaluation method and device, equipment, medium and product thereof | |
| CN114970467A (en) | Composition initial draft generation method, device, equipment and medium based on artificial intelligence | |
| CN114239555A (en) | Training method of keyword extraction model and related device | |
| CN114116967A (en) | Data cleaning method and device, electronic equipment and computer readable medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201204 |