CN112052663B - Customer service statement quality inspection method and related equipment - Google Patents
Customer service statement quality inspection method and related equipment Download PDFInfo
- Publication number
- CN112052663B CN112052663B CN202010898930.9A CN202010898930A CN112052663B CN 112052663 B CN112052663 B CN 112052663B CN 202010898930 A CN202010898930 A CN 202010898930A CN 112052663 B CN112052663 B CN 112052663B
- Authority
- CN
- China
- Prior art keywords
- customer service
- statement
- service statement
- sample
- memory
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- 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
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of artificial intelligence, and provides a customer service statement quality inspection method and related equipment. The customer service statement quality inspection method utilizes an input unit of a dynamic memory network model to encode a context statement sample of a customer service statement sample to obtain vector representation; coding a customer service statement sample by using a problem unit to obtain a first feature vector; calculating the vector representation and the first feature vector through a memory unit to obtain memory features; inputting the memory characteristics into the answering unit for recognition to obtain a quality inspection result of the customer service statement sample; training a dynamic memory network model according to the quality inspection result of the customer service statement sample; and outputting a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by using the dynamic memory network model. The invention improves the accuracy of detection. The method can be applied to the fields of intelligent government affairs and the like, so that the construction of an intelligent city can be promoted. Meanwhile, the invention also relates to a block chain.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a customer service statement quality inspection method, a customer service statement quality inspection device, computer equipment and a computer readable storage medium.
Background
In recent years, natural language processing is an important branch of the field of machine learning, and many problems of natural language processing can be handled by converting into text classification, and the term quality control belongs to the category of text classification.
In the contact exchange process of seat and customer, the condition that the words of saying of the seat can have a polite and a dirty word, leads to hindering or can't going on with customer's exchange, influences customer's experience, so need detect the words of seat art, traditional quality control mainly relies on artifical quality control, and the coverage is narrower, and the condition of lou examining is comparatively serious.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a method, an apparatus, a computer device, and a computer readable storage medium for quality inspection of customer service statements, which can perform quality inspection on customer service statements, reduce the occurrence of missed inspection, and improve the accuracy of inspection.
A first aspect of the present application provides a customer service statement quality inspection method, including:
acquiring a customer service statement sample, a quality label of the customer service statement sample and a context statement sample of the customer service statement sample;
coding a context statement sample of the customer service statement sample by using an input unit of a dynamic memory network model to obtain a vector representation of the context statement of the customer service statement sample, wherein the dynamic memory network model comprises the input unit, a question unit, a memory unit and an answer unit;
coding the customer service statement sample by using the problem unit to obtain a first feature vector of the customer service statement sample;
calculating the vector representation and the first feature vector through the memory unit to obtain memory features;
inputting the memory characteristics into the answer unit for identification to obtain a quality inspection result of the customer service statement sample;
training the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement;
acquiring a customer service statement to be detected and a context statement of the customer service statement to be detected;
and outputting a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by using the dynamic memory network model.
In another possible implementation manner, the input unit is a GRU neural network, which is denoted as a first GRU neural network, and the input of the first GRU neural network at a time step t is denoted as x t The hidden state at time step t-1 immediately preceding time step t is recorded as h t-1 Hidden state h of time step t in the first GRU neural network t ＝GRU′(x t ,h t-1 ) The calculation of (a) includes:
r t ＝σ(W r ·x t +U r ·h t-1 +b r )，
z t ＝σ(W z ·x t +U z ·h t-1 +b z )，
h t ＝(1-z t )⊙h t-1 +z t ⊙h′ t ，
wherein, W r 、W z 、W h And U r 、U z 、U h Representing a weight matrix, b r 、b z 、b h Indicating an offset, an, indicating the corresponding element multiply operator, a and
representing a non-linear activation function.
In another possible implementation manner, the calculating, by the storage unit, the vector representation and the first feature vector to obtain the memory feature includes:
sequentially taking 2, … and I-1 for the iteration times I, wherein I is a preset positive integer;
at time step t, based on the intermediate memory characteristics M of the i-1 th iteration t,i-1 The vector representation and the first feature vector calculation control gate g t,i ；
According to control of the door g t,i Computing an intermediate memory feature M for the ith iteration t,i ；
Intermediate memory characteristic M based on ith iteration t,i The vector representation and the first feature vector calculation control gate g t,i+1 ；
According to control of the door g t,i+1 Computing the intermediate memory feature M of the (i + 1) th iteration t,i+1 。
In another possible implementation, the memory unit is used to calculate the memory characteristics according to the vector representation and the second characteristic vector:
sequentially taking 2, … and I-1 for the iteration times I, wherein I is a preset positive integer;
acquiring an ith approximate sample of the customer service statement sample according to a preset customer service knowledge base;
coding the ith approximate sample by using the problem unit to obtain a second feature vector J of the customer service statement sample i ；
At time step t, based on the intermediate memory characteristics M of the i-1 th iteration t,i-1 The vector representation and a second feature vector J i Calculation control gate g t,i ；
According to control of the door g t,i Computing an intermediate memory feature M for the ith iteration t,i ；
Intermediate memory characteristic M based on ith iteration t,i The vector representation and the second feature vector calculation control gate g t,i+1 ；
According to control of the door g t,i+1 Computing the intermediate memory feature M of the (i + 1) th iteration t,i+1 。
In another possible implementation manner, the intermediate memory feature M based on the i-1 st iteration t,i-1 The vector representation and a second feature vector J i Calculation control gate g t,i The method comprises the following steps:
the t-th vector in the vector representation and a second feature vector J are combined i Performing element multiplication to obtain a first intermediate vector;
the t-th vector in the vector representation and the intermediate memory feature M are combined t,i-1 Performing element multiplication to obtain a second intermediate vector;
connecting the first intermediate vector and the second intermediate vector to obtain a third intermediate vector x t,i ；
Calculating a third intermediate vector x t,i Weight value X of t,i ；
According to the weight X t,i Calculation control gate g t,i 。
In another possible implementation, the control gate g is controlled according to the control signal t,i Computing memory characteristics M of the ith iteration t,i The method comprises the following steps:
obtaining input h of time step t t ；
Compute reset gate R t,i ，R t,i ＝σ(W R ·h t +U R ·M t,i-1 +b R )；
Calculating candidate memory feature M' t，i ，
Computing a context vector c from a control gate t ，c t ＝(1-g t,i )⊙M t,i-1 +g t,i ⊙M′ t,i ；
Calculating memory characteristics M t,i ＝GUR″(c t ,M t,i-1 )，W R 、W M And U R 、U M Representing a weight matrix, b R 、b M Indicating the bias.
In another possible implementation manner, the obtaining a to-be-detected customer service statement and a context statement of the to-be-detected customer service statement includes:
obtaining a plurality of given customer service statements;
storing the given customer service statements into a preset message queue according to a time sequence;
selecting the customer service statement to be detected from the preset message queue;
in the preset message queue, determining the given customer service statements of the preset number in front of the customer service statement to be detected and the given customer service statements of the preset number behind the customer service statement to be detected as context statements of the customer service statement to be detected.
A second aspect of the present application provides a customer service statement quality inspection device, including:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring a customer service statement sample, a quality label of the customer service statement sample and a context statement sample of the customer service statement sample;
the first coding module is used for coding the context statement samples of the customer service statement samples by using an input unit of a dynamic memory network model to obtain vector representation of the context statements of the customer service statement samples, and the dynamic memory network model comprises the input unit, a question unit, a memory unit and an answer unit;
the second coding module is used for coding the customer service statement sample by using the problem unit to obtain a first feature vector of the customer service statement sample;
the calculation module is used for calculating the vector representation and the first characteristic vector through the memory unit to obtain memory characteristics;
the output module is used for inputting the memory characteristics into the answer unit for identification to obtain a quality inspection result of the customer service statement sample;
the training module is used for training the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement;
the second acquisition module is used for acquiring a customer service statement to be detected and a context statement of the customer service statement to be detected;
and the detection module is used for outputting a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by utilizing the dynamic memory network model.
A fourth aspect of the present application provides a computer-readable storage medium having stored thereon computer-readable instructions which, when executed by a processor, implement the customer service statement quality inspection method.
Training the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement; acquiring a customer service statement to be detected and a context statement of the customer service statement to be detected; and outputting a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by utilizing the dynamic memory network model. The dynamic memory network model can update the memory characteristics through multiple iterative computations of the memory unit, and extract semantic information which is beneficial to quality inspection in the context sentence of the customer service sentence to be detected. The invention performs quality detection on the customer service statements, reduces the occurrence of missing detection and can improve the detection accuracy.
Drawings
Fig. 1 is a flowchart of a quality inspection method for a customer service statement according to an embodiment of the present invention.
Fig. 2 is a block diagram of a customer service statement quality inspection apparatus according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of a computer device provided by an embodiment of the present invention.
Detailed Description
In order that the above objects, features and advantages of the present invention can be more clearly understood, a detailed description of the present invention will be given below with reference to the accompanying drawings and specific embodiments. It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflict.
In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention, and the described embodiments are merely a subset of the embodiments of the present invention, rather than a complete embodiment.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.
Preferably, the customer service statement quality inspection method is applied to one or more computer devices. The computer device is a device capable of automatically performing numerical calculation and/or information processing according to a preset or stored instruction, and the hardware includes, but is not limited to, a microprocessor, an Application Specific Integrated Circuit (ASIC), a Programmable Gate Array (FPGA), a Digital Signal Processor (DSP), an embedded device, and the like.
The computer device can be a desktop computer, a notebook, a palm computer, a cloud server and other computing devices. The computer equipment can carry out man-machine interaction with a user through a keyboard, a mouse, a remote controller, a touch panel or voice control equipment and the like.
Example one
Fig. 1 is a flowchart of a quality inspection method for a customer service statement according to an embodiment of the present invention. The customer service statement quality inspection method is applied to computer equipment and used for carrying out quality inspection on customer service statements, so that the occurrence of missed inspection is reduced, and the accuracy of inspection can be improved.
As shown in fig. 1, the quality inspection method for customer service statements includes:
101, obtaining a customer service statement sample, a quality label of the customer service statement sample and a context statement sample of the customer service statement sample.
Semantic feature vectors of the customer service statements can be extracted from the customer service statements in the WeChat chat records, and customer service statement samples are obtained.
The quality label of the customer service statement sample can be a positive identification or a negative identification; if the positive mark "1" represents that the quality of the customer service statement sample is high, and the negative mark "0" represents that the quality of the customer service statement sample is low. The quality label can be a preset value or can be marked by an expert.
And (3) obtaining the context sentences (such as the upper 5 sentences and the lower 5 sentences) of the customer service sentences from the WeChat chat records, and extracting semantic feature vectors of the context sentences to obtain context sentence samples.
102, encoding a context statement sample of the customer service statement sample by using an input unit of a dynamic memory network model to obtain a vector representation of the context statement of the customer service statement sample, wherein the dynamic memory network model comprises the input unit, a question unit, a memory unit and an answer unit.
The input unit is a GRU neural network and is marked as a first GRU neural network. Recording the input of the first GRU neural network at the time step t as x t The hidden state at time step t-1 immediately preceding time step t is recorded as h t-1 。
In an alternative embodiment, the hidden state h at time step t t ＝GRU′(x t ,h t-1 ) The calculation of (a) includes:
r t ＝σ(W r ·x t +U r ·h t-1 +b r )；
z t ＝σ(W z ·x t +U z ·h t-1 +b z )；
h t ＝(1-z t )⊙h t-1 +z t ⊙h′ t ；
wherein, W r 、W z 、W h And U r 、U z 、U h Representing a weight matrix, b r 、b z 、b h Indicating an offset, an indication of a corresponding element multiply operator, σ and
representing a non-linear activation function.
Sigma may be a sigmoid function that may,
may be a hyperbolic tangent activation function.
The last layer of hidden states may be determined as a vector representation of a context statement of the customer service statement sample.
In another embodiment, the last hidden state of the last time step may be determined as a vector representation of the context statement of the customer service statement sample.
In another embodiment, the input unit is an LSTM (Long Short Term Memory) neural network.
And 103, encoding the customer service statement sample by using the problem unit to obtain a first feature vector of the customer service statement sample.
The problem unit is a GRU neural network, and is marked as a second GRU neural network.
And the second GRU neural network takes the customer service statement sample as an input item, and takes the last layer hidden layer state output by the second GRU neural network as a first feature vector of the customer service statement sample.
And 104, calculating the vector representation and the first feature vector by the memory unit to obtain memory features.
In a specific embodiment, the calculating, by the memory unit, the vector representation and the first feature vector to obtain the memory feature includes:
sequentially taking 2, … and I-1 for the iteration times I, wherein I is a preset positive integer;
at time step t, based on the intermediate memory characteristics M of the i-1 th iteration t,i-1 The vector representation and the first feature vector calculation control gate g t,i ；
According to control of the door g t,i Computing an intermediate memory feature M for the ith iteration t,i ；
Intermediate memory characteristic M based on ith iteration t,i The vector representation and the first feature vector calculation control gate g t,i+1 ；
According to control of the door g t,i+1 Computing the intermediate memory feature M of the (i + 1) th iteration t,i+1 。
And finally, obtaining the intermediate memory characteristic as the memory characteristic through iteration.
In an alternative embodiment, the memory feature is calculated from the vector representation and a second feature vector using the memory unit:
sequentially taking 2, … and I-1 for the iteration times I, wherein I is a preset positive integer;
acquiring an ith approximate sample of the customer service statement sample according to a preset customer service knowledge base;
coding the ith approximate sample by using the problem unit to obtain a second feature vector J of the customer service statement sample i ；
At time step t, based on the intermediate memory characteristics M of the i-1 th iteration t,i-1 The vector representation and a second feature vector J i Calculation control gate g t,i ；
According to control of the door g t,i Computing an intermediate memory feature M for the ith iteration t,i ；
Based on the ith iterationIntermediate memory characteristics of generations M t,i The vector representation and the second feature vector calculation control gate g t,i+1 ；
According to control of the door g t,i+1 Computing the intermediate memory feature M of the (i + 1) th iteration t,i+1 。
And coding the ith approximate sample of the customer service statement sample to obtain a second feature vector of the customer service statement sample. The method can increase the similar expression of the customer service statement sample, increase the similar expression of the same intention in the memory characteristic, reduce the condition of missing detection and improve the accuracy rate of quality detection on the customer service statement.
In one embodiment, the intermediate memory feature based on the (i-1) th iteration t,i-1 The vector representation and a second feature vector J i Calculation control gate g t,i The method comprises the following steps:
the t-th vector in the vector representation is compared with a second feature vector J i Carrying out element multiplication to obtain a first intermediate vector;
the t-th vector in the vector representation and the intermediate memory feature M are combined t,i-1 Performing element multiplication to obtain a second intermediate vector;
connecting the first intermediate vector and the second intermediate vector to obtain a third intermediate vector x t,i ；
Calculating a third intermediate vector x t,i Weight value X of t,i ；
According to the weight X t,i Calculation control gate g t,i 。
In particular, said calculation of the third intermediate vector x t,i Weight value X of t,i The method comprises the following steps:
X t,i ＝W X2 tanh(W X1 x t,i +b X1 )+b X2 wherein W is X1 、W X2 Representing a weight matrix, b X1 、b X2 Indicating the bias.
Specifically, the weight value X t，i Calculation control gate g t，i The method comprises the following steps:
In one embodiment, the control gate g t,i Computing memory characteristics M of the ith iteration t,i The method comprises the following steps:
obtaining input h of time step t t ；
Compute reset gate R t,i ，R t,i ＝σ(W R ·h t +U R ·M t,i-1 +b R )；
Calculating candidate memory characteristic M' t，i ，
Computing a context vector c from a control gate t ，c t ＝(1-g t,i )⊙M t,i-1 +g t,i ⊙M′ t,i ；
Calculating memory characteristics M t,i ＝GUR″(c t ,M t,i-1 ) Wherein W is R 、W M And U R 、U M Representing a weight matrix, b R 、b M Indicating the bias.
The original scheme lacks the information of the second feature vector and uses a control gate g ti Substitution of z in the original scheme t And the information of the second characteristic vector can be added in the memory process, so that the accuracy of quality detection on the customer service statement is improved. Computing a context vector c using a control gate t The control gate is a scalar generated using softmax activation, rather than a vector generated using sigmoid activation.
And 105, inputting the memory characteristics into the answer unit for recognition to obtain a quality inspection result of the customer service statement sample.
For example, the customer service statement is "we solve you immediately, ask you to wait slightly", and the quality control result of the customer service statement sample is output as "0.1" by the answering unit according to the obtained memory characteristics (indicating that the quality of the customer service statement is low).
106, training the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement.
As in the above example, the quality label of the customer service statement is "1" (indicating that the quality of the customer service statement is high and the quality label is a correct answer), the difference between the quality inspection result and the quality label is calculated according to the quality inspection result "0.1" of the customer service statement sample and the quality label "1" of the customer service statement, and the parameters in the dynamic memory network model are optimized through back propagation according to the difference. So that the dynamic memory network model can output the correct quality inspection result of the customer service statement.
It is emphasized that, in order to further ensure the privacy and security of the parameters of the dynamic memory network model, the parameters of the dynamic memory network model may also be stored in the nodes of a block chain.
And 107, acquiring the customer service statement to be detected and the context statement of the customer service statement to be detected.
In a specific embodiment, the obtaining the customer service statement to be detected and the context statement of the customer service statement to be detected includes:
obtaining a plurality of given customer service statements;
storing the given customer service statements into a preset message queue according to a time sequence;
selecting the customer service statement to be detected from the preset message queue;
in the preset message queue, determining the given customer service statements of the preset number in front of the customer service statement to be detected and the given customer service statements of the preset number behind the customer service statement to be detected as context statements of the customer service statement to be detected.
For example, the received customer service statements may be sequentially stored in a preset message queue in time order, one customer service statement to be detected may be selected from the preset message queue, the first 5 statements and the last 5 statements of the customer service statement to be detected may be obtained, and the previous 5 statements and the last 5 statements of the customer service statement to be detected may be determined as context statements of the customer service statement to be detected.
And 108, outputting a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by using the dynamic memory network model.
And carrying out iterative training on the dynamic memory network model by using the sample data until iterative convergence, wherein if the training times reach a preset time threshold value or the convergence condition of the loss function is met, the dynamic memory network model is suitable for an application scene for detecting the quality of the customer service statement and can be accurately classified. Namely, the quality inspection result of the customer service statement to be detected can be output by using the dynamic memory network model according to the customer service statement to be detected and the context statement of the customer service statement to be detected.
According to the quality inspection method for the customer service statements, the dynamic memory network model is trained through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement; acquiring a customer service statement to be detected and a context statement of the customer service statement to be detected; and outputting a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by using the dynamic memory network model. The dynamic memory network model can update the memory characteristics through multiple iterative computations of the memory unit, and extract semantic information which is beneficial to quality inspection in the context sentence of the customer service sentence to be detected. The embodiment performs quality detection on the customer service statements, reduces the occurrence of missed detection, and can improve the detection accuracy.
Example two
Fig. 2 is a structural diagram of a customer service statement quality inspection apparatus according to a second embodiment of the present invention. The customer service statement quality inspection device 20 is applied to computer equipment. The customer service statement quality inspection device 20 is used for performing quality inspection on customer service statements, so that the occurrence of missed inspection is reduced, and the accuracy of detection can be improved.
As shown in fig. 2, the customer service statement quality inspection apparatus 20 may include a first obtaining module 201, a first encoding module 202, a second encoding module 203, a calculating module 204, an output module 205, a training module 206, a second obtaining module 207, and a detecting module 208.
The first obtaining module 201 is configured to obtain a customer service statement sample, a quality label of the customer service statement sample, and a context statement sample of the customer service statement sample.
Semantic feature vectors of the customer service statements can be extracted from the customer service statements in the WeChat chat records, and customer service statement samples are obtained.
The quality label of the customer service statement sample can be a positive identification or a negative identification; if the positive sign "1" indicates that the quality of the customer service statement sample is high, and the negative sign "0" indicates that the quality of the customer service statement sample is low. The quality label can be a preset value or can be marked by an expert.
And (3) obtaining the context sentences (such as the upper 5 sentences and the lower 5 sentences) of the customer service sentences from the WeChat chat records, and extracting semantic feature vectors of the context sentences to obtain context sentence samples.
A first encoding module 202, configured to encode a context statement sample of the customer service statement sample by using an input unit of a dynamic memory network model, to obtain a vector representation of a context statement of the customer service statement sample, where the dynamic memory network model includes the input unit, a question unit, a memory unit, and an answer unit.
The input unit is a GRU neural network and is marked as a first GRU neural network. Recording the input of the first GRU neural network at the time step t as x t The hidden state at time step t-1 immediately preceding time step t is recorded as h t-1 。
In an alternative embodiment, the hidden state h at time step t t ＝GRU′(x t ,h t-1 ) The calculation of (a) includes:
r t ＝σ(W r ·x t +U r ·h t-1 +b r )；
z t ＝σ(W z ·x t +U z ·h t-1 +b z )；
h t ＝(1-z t )⊙h t-1 +z t ⊙h′ t ；
wherein, W r 、W z 、W h And U r 、U z 、U h Representing a weight matrix, b r 、b z 、b h Indicating an offset, an indication of a corresponding element multiply operator, σ and
representing a non-linear activation function.
Sigma may be a sigmoid function that may,
may be a hyperbolic tangent activation function.
The last layer of hidden states may be determined as a vector representation of a context statement of the customer service statement sample.
In another embodiment, the last hidden state of the last time step may be determined as a vector representation of the context statement of the customer service statement sample.
In another embodiment, the input unit is an LSTM (Long Short Term Memory) neural network.
The second encoding module 203 is configured to encode the customer service statement sample by using the problem unit to obtain a first feature vector of the customer service statement sample.
The problem unit is a GRU neural network, and is marked as a second GRU neural network.
And the second GRU neural network takes the customer service statement sample as an input item, and takes the last layer hidden layer state output by the second GRU neural network as a first feature vector of the customer service statement sample.
A calculating module 204, configured to calculate the vector representation and the first feature vector through the memory unit to obtain a memory feature.
In a specific embodiment, the calculating, by the storage unit, the vector representation and the first feature vector to obtain the memory feature includes:
sequentially taking 2, … and I-1 for the iteration times I, wherein I is a preset positive integer;
at time step t, based on the intermediate memory characteristics M of the i-1 th iteration t,i-1 The vector representation and the first feature vector calculation control gate g t,i ；
According to control of the door g t,i Computing an intermediate memory feature M for the ith iteration t,i ；
Intermediate memory characteristic M based on ith iteration t,i The vector representation and the first feature vector calculation control gate g t,i+1 ；
According to control of the door g t,i+1 Computing the intermediate memory feature M of the (i + 1) th iteration t,i+1 。
And finally, obtaining the intermediate memory characteristic as the memory characteristic through iteration.
In an alternative embodiment, the memory feature is calculated from the vector representation and a second feature vector using the memory unit:
sequentially taking 2, … and I-1 for the iteration times I, wherein I is a preset positive integer;
acquiring an ith approximate sample of the customer service statement sample according to a preset customer service knowledge base;
coding the ith approximate sample by using the problem unit to obtain a second feature vector J of the customer service statement sample i ；
At time step t, based on the intermediate memory characteristics M of the i-1 th iteration t,i-1 The vector representation and a second feature vector J i Calculation control gate g t,i ；
According to control of the door g t,i Computing an intermediate memory feature M for the ith iteration t,i ；
Intermediate memory characteristic M based on ith iteration t,i The vector representation and the second feature vector calculation control gate g t,i+1 ；
According to control of the door g t,i+1 Computing the intermediate memory feature M of the (i + 1) th iteration t,i+1 。
And coding the ith approximate sample of the customer service statement sample to obtain a second feature vector of the customer service statement sample. The method can increase the similar expression of the customer service statement sample, increase the similar expression of the same intention in the memory characteristic, reduce the condition of missing detection and improve the accuracy rate of quality detection on the customer service statement.
In one embodiment, the intermediate memory feature based on the (i-1) th iteration t,i-1 The vector representation and a second feature vector J i Calculation control gate g t,i The method comprises the following steps:
the t-th vector in the vector representation is compared with a second feature vector J i Carrying out element multiplication to obtain a first intermediate vector;
the t-th vector in the vector representation and the intermediate memory feature M are combined t,i-1 Performing element multiplication to obtain a second intermediate vector;
connecting the first intermediate vector and the second intermediate vector to obtain a third intermediate vector x t,i ；
Calculating a third intermediate vector x t,i Weight value X of t,i ；
According to the weight X t,i Calculation control gate g t,i 。
In particular, said calculation of the third intermediate vector x t,i Weight value X of t,i The method comprises the following steps:
X t,i ＝W X2 tanh(W X1 x t,i +b X1 )+b X2 wherein W is X1 、W X2 Representing a weight matrix, b X1 、b X2 Indicating the bias.
Specifically, the weight value X t,i Calculation control gate g t,i The method comprises the following steps:
In one embodiment, the control gate g t,i Computing memory characteristics M of the ith iteration t,i The method comprises the following steps:
obtaining input h of time step t t ；
Compute reset gate R t,i ，R t,i ＝σ(W R ·h t +U R ·M t,i-1 +b R )；
Calculating candidate memory feature M' t，i ，
Computing a context vector c from a control gate t ，c t ＝(1-g t,i )⊙M t,i-1 +g t,i ⊙M′ t,i ；
Calculating memory characteristics M t,i ＝GUR″(c t ,M t,i-1 ) Wherein W is R 、W M And U R 、U M Representing a weight matrix, b R 、b M Indicating the bias.
The original scheme lacks the information of the second feature vector and uses a control gate g ti Replacing z in the original scheme t And the information of the second characteristic vector can be added in the memory process, so that the accuracy of quality detection on the customer service statement is improved. Computing a context vector c using a control gate t The control gate is a scalar generated using softmax activation, rather than a vector generated using sigmoid activation.
And the output module 205 is configured to input the memory characteristics into the answer unit for identification to obtain a quality inspection result of the customer service statement sample.
For example, the customer service statement is "we solve you immediately, ask you to wait slightly", and the quality control result of the customer service statement sample is output as "0.1" by the answering unit according to the obtained memory characteristics (indicating that the quality of the customer service statement is low).
And the training module 206 is configured to train the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement.
As in the above example, the quality label of the customer service statement is "1" (indicating that the quality of the customer service statement is high and the quality label is a correct answer), the difference between the quality inspection result and the quality label is calculated according to the quality inspection result "0.1" of the customer service statement sample and the quality label "1" of the customer service statement, and the parameters in the dynamic memory network model are optimized through back propagation according to the difference. So that the dynamic memory network model can output the correct quality inspection result of the customer service statement.
It is emphasized that, in order to further ensure the privacy and security of the parameters of the dynamic memory network model, the parameters of the dynamic memory network model may also be stored in the nodes of a block chain.
The second obtaining module 207 is configured to obtain the customer service statement to be detected and the context statement of the customer service statement to be detected.
In a specific embodiment, the obtaining the customer service statement to be detected and the context statement of the customer service statement to be detected includes:
obtaining a plurality of given customer service statements;
storing the given customer service statements into a preset message queue according to a time sequence;
selecting the customer service statement to be detected from the preset message queue;
in the preset message queue, determining the given customer service statements of the preset number in front of the customer service statement to be detected and the given customer service statements of the preset number behind the customer service statement to be detected as context statements of the customer service statement to be detected.
For example, the received customer service statements may be sequentially stored in a preset message queue in time order, one customer service statement to be detected may be selected from the preset message queue, the first 5 statements and the last 5 statements of the customer service statement to be detected may be obtained, and the previous 5 statements and the last 5 statements of the customer service statement to be detected may be determined as context statements of the customer service statement to be detected.
And the detection module 208 is configured to output a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by using the dynamic memory network model.
And carrying out iterative training on the dynamic memory network model by using the sample data until iterative convergence, wherein if the training times reach a preset time threshold value or the convergence condition of the loss function is met, the dynamic memory network model is suitable for an application scene for detecting the quality of the customer service statement and can be accurately classified. Namely, the quality inspection result of the customer service statement to be detected can be output by using the dynamic memory network model according to the customer service statement to be detected and the context statement of the customer service statement to be detected.
The customer service statement quality inspection device 20 of the second embodiment trains the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement; acquiring a customer service statement to be detected and a context statement of the customer service statement to be detected; and outputting a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by using the dynamic memory network model. The dynamic memory network model can update the memory characteristics through multiple iterative computations of the memory unit, and extract semantic information which is beneficial to quality inspection in the context sentence of the customer service sentence to be detected. The embodiment performs quality detection on the two pairs of customer service statements, reduces the occurrence of missed detection, and can improve the detection accuracy.
EXAMPLE III
The present embodiment provides a computer-readable storage medium, where computer-readable instructions are stored on the computer-readable storage medium, and when the computer-readable instructions are executed by a processor, the steps in the quality inspection method for customer service statements described above are implemented, for example, the steps 101 and 108 shown in fig. 1:
101, obtaining a customer service statement sample, a quality label of the customer service statement sample and a context statement sample of the customer service statement sample;
102, encoding a context statement sample of the customer service statement sample by using an input unit of a dynamic memory network model to obtain a vector representation of a context statement of the customer service statement sample, wherein the dynamic memory network model comprises the input unit, a question unit, a memory unit and an answer unit;
103, encoding the customer service statement sample by using the problem unit to obtain a first feature vector of the customer service statement sample;
104, calculating the vector representation and the first feature vector by the memory unit to obtain memory features;
105, inputting the memory characteristics into the answer unit for recognition to obtain a quality inspection result of the customer service statement sample;
106, training the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement;
107, acquiring a customer service statement to be detected and a context statement of the customer service statement to be detected;
and 108, outputting a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by using the dynamic memory network model.
Alternatively, the computer readable instructions, when executed by the processor, implement the functions of the modules in the above device embodiments, for example, the module 201 and 208 in fig. 2:
a first obtaining module 201, configured to obtain a customer service statement sample, a quality label of the customer service statement sample, and a context statement sample of the customer service statement sample;
a first encoding module 202, configured to encode a context statement sample of the customer service statement sample by using an input unit of a dynamic memory network model, to obtain a vector representation of a context statement of the customer service statement sample, where the dynamic memory network model includes the input unit, a question unit, a memory unit, and an answer unit;
the second coding module 203 is configured to code the customer service statement sample by using the problem unit to obtain a first feature vector of the customer service statement sample;
a calculating module 204, configured to calculate the vector representation and the first feature vector through the memory unit to obtain a memory feature;
the output module 205 is configured to input the memory characteristics into the answer unit for identification to obtain a quality inspection result of the customer service statement sample;
a training module 206, configured to train the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement;
a second obtaining module 207, configured to obtain a customer service statement to be detected and a context statement of the customer service statement to be detected;
and the detection module 208 is configured to output a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by using the dynamic memory network model.
Example four
Fig. 3 is a schematic diagram of a computer device according to a third embodiment of the present invention. The computer device 30 includes a memory 301, a processor 302, and computer readable instructions 303, such as a customer service statement quality check program, stored in the memory 301 and executable on the processor 302. The processor 302, when executing the computer readable instructions 303, implements the steps in the above-mentioned quality inspection method embodiment of the customer service statement, for example, 101-:
101, obtaining a customer service statement sample, a quality label of the customer service statement sample and a context statement sample of the customer service statement sample;
102, encoding a context statement sample of the customer service statement sample by using an input unit of a dynamic memory network model to obtain a vector representation of a context statement of the customer service statement sample, wherein the dynamic memory network model comprises the input unit, a question unit, a memory unit and an answer unit;
103, encoding the customer service statement sample by using the problem unit to obtain a first feature vector of the customer service statement sample;
104, calculating the vector representation and the first feature vector by the memory unit to obtain memory features;
105, inputting the memory characteristics into the answer unit for recognition to obtain a quality inspection result of the customer service statement sample;
106, training the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement;
107, acquiring a customer service statement to be detected and a context statement of the customer service statement to be detected;
and 108, outputting a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by using the dynamic memory network model.
Alternatively, the computer readable instructions, when executed by the processor, implement the functions of the modules in the above device embodiments, for example, the module 201 and 208 in fig. 2:
a first obtaining module 201, configured to obtain a customer service statement sample, a quality label of the customer service statement sample, and a context statement sample of the customer service statement sample;
a first encoding module 202, configured to encode a context statement sample of the customer service statement sample by using an input unit of a dynamic memory network model, to obtain a vector representation of a context statement of the customer service statement sample, where the dynamic memory network model includes the input unit, a question unit, a memory unit, and an answer unit;
the second coding module 203 is configured to code the customer service statement sample by using the problem unit to obtain a first feature vector of the customer service statement sample;
a calculating module 204, configured to calculate the vector representation and the first feature vector through the memory unit to obtain a memory feature;
the output module 205 is configured to input the memory characteristics into the answer unit for identification to obtain a quality inspection result of the customer service statement sample;
a training module 206, configured to train the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement;
a second obtaining module 207, configured to obtain a customer service statement to be detected and a context statement of the customer service statement to be detected;
and the detection module 208 is configured to output a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by using the dynamic memory network model.
Illustratively, the computer readable instructions 303 may be partitioned into one or more modules that are stored in the memory 301 and executed by the processor 302 to perform the present method. The one or more modules may be a series of computer-readable instructions capable of performing certain functions and describing the execution of the computer-readable instructions 303 in the computer device 30. For example, the computer-readable instructions 303 may be divided into a first obtaining module 201, a first encoding module 202, a second encoding module 203, a calculating module 204, an output module 205, a training module 206, a second obtaining module 207, and a detecting module 208 in fig. 2, where specific functions of the modules are described in embodiment two.
Those skilled in the art will appreciate that the schematic diagram 3 is merely an example of the computer device 30 and does not constitute a limitation of the computer device 30, and may include more or less components than those shown, or combine certain components, or different components, for example, the computer device 30 may also include input and output devices, network access devices, buses, etc.
The Processor 302 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor 302 may be any conventional processor or the like, the processor 302 being the control center for the computer device 30 and connecting the various parts of the overall computer device 30 using various interfaces and lines.
The memory 301 may be used to store the computer-readable instructions 303, and the processor 302 may implement the various functions of the computer device 30 by executing or executing the computer-readable instructions or modules stored in the memory 301 and invoking data stored in the memory 301. The memory 301 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data created according to the use of the computer device 30, and the like. In addition, the Memory 301 may include a hard disk, a Memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash Memory Card (Flash Card), at least one disk storage device, a Flash Memory device, a Read-Only Memory (ROM), a Random Access Memory (RAM), or other non-volatile/volatile storage devices.
The modules integrated by the computer device 30 may be stored in a computer-readable storage medium if they are implemented in the form of software functional modules and sold or used as separate products. Based on such understanding, all or part of the flow of the method according to the above embodiments may be implemented by hardware that is configured to be instructed by computer readable instructions, which may be stored in a computer readable storage medium, and when the computer readable instructions are executed by a processor, the steps of the method embodiments may be implemented. Wherein the computer readable instructions comprise computer readable instruction code which may be in source code form, object code form, an executable file or some intermediate form, and the like. The computer-readable medium may include: any entity or device capable of carrying the computer readable instruction code, recording medium, U disk, removable hard disk, magnetic disk, optical disk, computer memory, Read Only Memory (ROM), Random Access Memory (RAM), etc.
Further, the computer usable storage medium may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function, and the like; the storage data area may store data created according to the use of the blockchain node, and the like.
The block chain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism, an encryption algorithm and the like. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.
In the embodiments provided in the present invention, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is only one logical functional division, and other divisions may be realized in practice.
The modules described as separate parts may or may not be physically separate, and parts displayed as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.
In addition, functional modules in the embodiments of the present invention may be integrated into one processing module, or each of the modules may exist alone physically, or two or more modules are integrated into one module. The integrated module can be realized in a hardware form, and can also be realized in a form of hardware and a software functional module.
The integrated module implemented in the form of a software functional module may be stored in a computer-readable storage medium. The software functional module is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute some steps of the customer service statement quality inspection method according to various embodiments of the present invention.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference signs in the claims shall not be construed as limiting the claim concerned. Furthermore, it is to be understood that the word "comprising" does not exclude other modules or steps, and the singular does not exclude the plural. A plurality of modules or means recited in the system claims may also be implemented by one module or means in software or hardware. The terms first, second, etc. are used to denote names, but not any particular order.
Finally, it should be noted that the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting, and although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims (8)
1. A customer service statement quality inspection method is characterized by comprising the following steps:
acquiring a customer service statement sample, a quality label of the customer service statement sample and a context statement sample of the customer service statement sample;
coding a context statement sample of the customer service statement sample by using an input unit of a dynamic memory network model to obtain a vector representation of the context statement of the customer service statement sample, wherein the dynamic memory network model comprises the input unit, a question unit, a memory unit and an answer unit;
using the memory unit to represent and represent a second bit from the vectorComputing memory features from the eigenvector, the computing memory features from the vector representation and a second eigenvector using the memory unit comprising: sequentially taking 2, 1 and I-1 for the iteration times I, wherein I is a preset positive integer; acquiring an ith approximate sample of the customer service statement sample according to a preset customer service knowledge base; coding the ith approximate sample by using the problem unit to obtain a second feature vector J of the customer service statement sample i (ii) a At time step t, based on the intermediate memory characteristics M of the i-1 th iteration t，i-1 The vector representation and a second feature vector J i Calculation control gate g t，i (ii) a According to control of the door g t，i Computing an intermediate memory feature M for the ith iteration t，i (ii) a Intermediate memory characteristic M based on ith iteration t，i The vector representation and the second feature vector calculation control gate g t，i+1 (ii) a According to control of the door g t，i+1 Computing the intermediate memory feature M of the (i + 1) th iteration t，i+1 ；
Inputting the memory characteristics into the answer unit for recognition to obtain a quality inspection result of the customer service statement sample;
training the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement;
acquiring a customer service statement to be detected and a context statement of the customer service statement to be detected;
and outputting a quality inspection result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by using the dynamic memory network model.
2. The method of claim 1, wherein the input unit is a GRU neural network, and is denoted as a first GRU neural network, and the input of the first GRU neural network at time step t is denoted as x t The hidden state at time step t-1 immediately preceding time step t is recorded as h t-1 Hidden state h of time step t in the first GRU neural network t ＝GRU′(x t ，h t-1 ) The calculation of (a) includes:
r t ＝σ(W r ·x t +U r ·h t-1 +b r )，
z t ＝σ(W z ·x t +U z ·h t-1 +b z )，
h t ＝(1-z t )⊙h t-1 +z t ⊙h′ t ，
wherein, W r 、W z 、W h And U r 、U z 、U h Representing a weight matrix, b r 、b z 、b h Indicating an offset, an indication of a corresponding element multiply operator, σ and
representing a non-linear activation function.
3. The method for quality inspection of customer service statements according to claim 1, wherein the intermediate memory characteristics M based on the i-1 st iteration t，i-1 The vector representation and a second feature vector J i Calculation control gate g t，i The method comprises the following steps:
the t-th vector in the vector representation and a second feature vector J are combined i Carrying out element multiplication to obtain a first intermediate vector;
the t-th vector in the vector representation and the intermediate memory feature M are combined t，i-1 Performing element multiplication to obtain a second intermediate vector;
connecting the first intermediate vector and the second intermediate vector to obtain a third intermediate vector x t，i ；
Calculating a third intermediate vector x t，i Weight value X of t，i ；
According to the weight X t，i Calculation control gate g t，i 。
4. The method of claim 1, wherein the customer service statement quality inspection is based on a control gate g t，i Computing memory characteristics M of the ith iteration t，i The method comprises the following steps:
obtaining input h of time step t t ；
Compute reset gate R t，i ，R t，i ＝σ(W R ·h t +U R ·M t，i-1 +b R )；
Calculating candidate memory characteristic M' t，i ，
Computing a context vector c from a control gate t ，c t ＝(1-g t，i )⊙M t，i-1 +g t，i ⊙M′ t，i ；
Calculating memory characteristics M t，i ＝GUR″(c t ，M t，i-1 )，W R 、W M And U R 、U M Representing a weight matrix, b R 、b M Indicating the bias.
5. The method for quality inspection of customer service statements according to claim 1, wherein the obtaining of the customer service statement to be detected and the context statement of the customer service statement to be detected comprises:
obtaining a plurality of given customer service statements;
storing the given customer service statements into a preset message queue according to a time sequence;
selecting the customer service statement to be detected from the preset message queue;
in the preset message queue, determining the given customer service statements of the preset number in front of the customer service statement to be detected and the given customer service statements of the preset number behind the customer service statement to be detected as context statements of the customer service statement to be detected.
6. A customer service statement quality inspection device, characterized by comprising:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring a customer service statement sample, a quality label of the customer service statement sample and a context statement sample of the customer service statement sample;
the first coding module is used for coding the context statement samples of the customer service statement samples by using an input unit of a dynamic memory network model to obtain vector representation of the context statements of the customer service statement samples, and the dynamic memory network model comprises the input unit, a question unit, a memory unit and an answer unit;
a computing module, configured to compute memory characteristics from the vector representation and the second feature vector using the memory unit, where computing memory characteristics from the vector representation and the second feature vector using the memory unit includes: sequentially taking 2, 1 and I-1 for the iteration times I, wherein I is a preset positive integer; acquiring an ith approximate sample of the customer service statement sample according to a preset customer service knowledge base; coding the ith approximate sample by using the problem unit to obtain a second feature vector J of the customer service statement sample i (ii) a At time step t, based on the intermediate memory characteristics M of the i-1 th iteration t，i-1 The vector representation and a second feature vector J i Calculation control gate g t，i (ii) a According to control of the door g t，i Computing an intermediate memory feature M for the ith iteration t，i (ii) a Intermediate memory characteristic M based on ith iteration t，i The vector representation and the second feature vector calculation control gate g t，i+1 (ii) a According to control of the door g t，i+1 Computing the intermediate memory feature M of the (i + 1) th iteration t，i+1 ；
The output module is used for inputting the memory characteristics into the answer unit for identification to obtain a quality inspection result of the customer service statement sample;
the training module is used for training the dynamic memory network model through back propagation according to the quality inspection result of the customer service statement sample and the quality label of the customer service statement;
the second acquisition module is used for acquiring a customer service statement to be detected and a context statement of the customer service statement to be detected;
and the detection module is used for outputting a quality test result of the customer service statement to be detected according to the customer service statement to be detected and the context statement of the customer service statement to be detected by utilizing the dynamic memory network model.
7. A computer device comprising a processor for executing computer readable instructions stored in a memory to implement the customer service statement quality inspection method of any one of claims 1 to 5.
8. A computer readable storage medium having computer readable instructions stored thereon, wherein the computer readable instructions, when executed by a processor, implement the customer service statement quality inspection method according to any one of claims 1 to 5.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010898930.9A CN112052663B (en) | 2020-08-31 | 2020-08-31 | Customer service statement quality inspection method and related equipment |
| PCT/CN2020/122921 WO2021147405A1 (en) | 2020-08-31 | 2020-10-22 | Customer-service statement quality detection method and related device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010898930.9A CN112052663B (en) | 2020-08-31 | 2020-08-31 | Customer service statement quality inspection method and related equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112052663A CN112052663A (en) | 2020-12-08 |
| CN112052663B true CN112052663B (en) | 2022-08-02 |
Family
ID=73606536
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010898930.9A Active CN112052663B (en) | 2020-08-31 | 2020-08-31 | Customer service statement quality inspection method and related equipment |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN112052663B (en) |
| WO (1) | WO2021147405A1 (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115168568B (en) * | 2022-03-16 | 2024-04-05 | 腾讯科技（深圳）有限公司 | Data content identification method, device and storage medium |
| CN117151228B (en) * | 2023-10-31 | 2024-02-02 | 深圳大数信科技术有限公司 | Intelligent customer service system based on large model and knowledge base generation |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019037382A1 (en) * | 2017-08-24 | 2019-02-28 | 平安科技（深圳）有限公司 | Emotion recognition-based voice quality inspection method and device, equipment and storage medium |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11113598B2 (en) * | 2015-06-01 | 2021-09-07 | Salesforce.Com, Inc. | Dynamic memory network |
| WO2019000170A1 (en) * | 2017-06-26 | 2019-01-03 | Microsoft Technology Licensing, Llc | Generating responses in automated chatting |
| CN110232114A (en) * | 2019-05-06 | 2019-09-13 | 平安科技（深圳）有限公司 | Sentence intension recognizing method, device and computer readable storage medium |
| CN110309857A (en) * | 2019-06-03 | 2019-10-08 | 平安科技（深圳）有限公司 | Book classification device, method, equipment and storage medium based on artificial intelligence |
| CN111274375B (en) * | 2020-01-20 | 2022-06-14 | 福州大学 | Multi-turn dialogue method and system based on bidirectional GRU network |
| CN111508498B (en) * | 2020-04-09 | 2024-01-30 | 携程计算机技术（上海）有限公司 | Conversational speech recognition method, conversational speech recognition system, electronic device, and storage medium |
| CN111522916B (en) * | 2020-04-20 | 2021-03-09 | 马上消费金融股份有限公司 | Voice service quality detection method, model training method and device |
-
2020
- 2020-08-31 CN CN202010898930.9A patent/CN112052663B/en active Active
- 2020-10-22 WO PCT/CN2020/122921 patent/WO2021147405A1/en active Application Filing
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019037382A1 (en) * | 2017-08-24 | 2019-02-28 | 平安科技（深圳）有限公司 | Emotion recognition-based voice quality inspection method and device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2021147405A1 (en) | 2021-07-29 |
| CN112052663A (en) | 2020-12-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10628731B1 (en) | Deep convolutional neural networks for automated scoring of constructed responses | |
| WO2021114840A1 (en) | Scoring method and apparatus based on semantic analysis, terminal device, and storage medium | |
| CN107644011A (en) | System and method for the extraction of fine granularity medical bodies | |
| JP6772213B2 (en) | Question answering device, question answering method and program | |
| CN111461168A (en) | Training sample expansion method and device, electronic equipment and storage medium | |
| CN111538868A (en) | Knowledge tracking method and exercise recommendation method | |
| CN112434131B (en) | Text error detection method and device based on artificial intelligence and computer equipment | |
| Steele et al. | Algorithms for data science | |
| CN112052663B (en) | Customer service statement quality inspection method and related equipment | |
| Singer | Symmetry in mechanics: A gentle, modern introduction | |
| CN109948735A (en) | A kind of multi-tag classification method, system, device and storage medium | |
| Kandler et al. | Analysing cultural frequency data: Neutral theory and beyond | |
| KR20230013793A (en) | Method and Apparatus for Classifying Document Based on Attension Mechanism and Semantic Analysis | |
| CN111695335A (en) | Intelligent interviewing method and device and terminal equipment | |
| CN113704393A (en) | Keyword extraction method, device, equipment and medium | |
| El Fouki et al. | Multidimensional Approach Based on Deep Learning to Improve the Prediction Performance of DNN Models. | |
| CN115169252A (en) | Structured simulation data generation system and method | |
| CN115222443A (en) | Client group division method, device, equipment and storage medium | |
| CN111488460A (en) | Data processing method, device and computer readable storage medium | |
| CN112434862B (en) | Method and device for predicting financial dilemma of marketing enterprises | |
| CN114298299A (en) | Model training method, device, equipment and storage medium based on course learning | |
| CN116402166B (en) | Training method and device of prediction model, electronic equipment and storage medium | |
| CN112036439A (en) | Dependency relationship classification method and related equipment | |
| CN112102062A (en) | Risk assessment method and device based on weak supervised learning and electronic equipment | |
| CN117034916A (en) | Method, device and equipment for constructing word vector representation model and word vector representation |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |