Customer complaint escalation prediction method, device and electronic equipment

Abstract

The present application provides a customer complaint escalation prediction method, device and electronic device, which are applied to the field of voice processing, wherein the method obtains the voice call information between the user and the customer service, performs voice recognition on the voice call information to obtain the conversation content between the user and the customer service, and determines the user's complaint problem based on the conversation content, and determines the user's static complaint escalation probability based on the complaint problem, extracts the user's voiceprint features based on the voice call information, and determines the user's dynamic complaint escalation probability based on the voiceprint features, and determines whether the user will escalate the customer complaint based on the static complaint escalation probability and the dynamic complaint escalation probability. That is, the automatic identification of customer complaint escalation is realized. In addition, determining whether the user will escalate the customer complaint based on the static complaint escalation probability and the dynamic complaint escalation probability can improve the accuracy of customer complaint escalation prediction.

Description

Technical Field

[0001] The present application relates to the field of speech processing technology, and more specifically, to a method, device and electronic device for predicting customer complaint upgrades. Background Art

[0002] Credit card customer complaint escalation usually refers to the subjective belief of credit card users that the credit card center cannot properly solve the problems they encounter, thus bypassing the service channels provided by the card center and making complaints to corporate superiors, regulatory authorities, media and other channels. The reasons for credit card customer complaint escalation generally include the following points: first, customers have high requirements for complaint handling, and want the problem to be solved immediately, obtain high economic compensation, mental damage compensation and other requirements that are difficult for the card center to accept; second, the problem complained by the customer is difficult to solve, the card center cannot solve it or cannot reach an agreement with the customer, and the customer hopes to put pressure on the card center by escalating the complaint; finally, the customer itself is difficult to deal with, and the customer who escalates the complaint is often good at catching the flaws of the card center's products, which makes it difficult for the card center to deal with. As a service-oriented enterprise, if the card center handles the escalation of complaints improperly, it will often cause serious impacts such as reputation damage and media exposure due to these escalated complaints. Therefore, the customer service department of the card center needs to focus on the prevention and management of customer escalation complaints. The traditional method of preventing customer complaint escalation is purely based on the judgment of customer service personnel, but such human judgment has inconsistent standards, relies heavily on the experience of customer service personnel, and has a very low accuracy rate. Summary of the invention

[0003] The present application provides a customer complaint escalation prediction method, device and electronic device for realizing automatic identification of customer complaint escalation and improving the accuracy of customer complaint escalation identification. The technical solution adopted by the present application is as follows:

[0004] In a first aspect, a method for predicting customer complaint escalation is provided, the method comprising:

[0005] Get the voice call information between the user and the customer service;

[0006] Perform voice recognition on the voice call information to obtain the conversation content between the user and the customer service, and determine the user's complaint issue based on the conversation content, and determine the user's static complaint escalation probability based on the complaint issue;

[0007] Extract the user's voiceprint features based on voice call information, and determine the user's dynamic complaint upgrade probability based on the voiceprint features;

[0008] Determine whether the user should escalate the customer complaint based on the static complaint escalation probability and the dynamic complaint escalation probability.

[0009] Optionally, determining the user's complaint issue based on the conversation content includes:

[0010] The user's complaint issue is determined based on the conversation content using the BERT algorithm.

[0011] Optionally, extracting the voiceprint feature of the user based on the voice call information, and determining the dynamic complaint escalation probability of the user based on the voiceprint feature includes:

[0012] Determining the basic emotional voiceprint characteristics of the user;

[0013] The dynamic upgrade probability of the user is determined based on the basic emotional voiceprint feature of the user and the voiceprint feature of the user.

[0014] Optionally, the method further comprises:

[0015] Determining whether the user has called before;

[0016] If the user has called before, searching the user emotion voiceprint feature library to determine the basic emotion voiceprint feature of the user;

[0017] If the user calls for the first time, the initial speech voiceprint feature of the user for a predetermined time length in the call voice information is used as the basic emotional voiceprint feature.

[0018] Optionally, extracting the voiceprint feature of the user based on the voice call information includes:

[0019] Performing voice separation processing on the voice call information to obtain the voice information of the user;

[0020] The voiceprint feature of the user is obtained based on the voice information of the user.

[0021] Optionally, extracting the voiceprint feature of the user based on the voice information of the user includes:

[0022] The voiceprint features of the user are extracted based on the voice information of the user through an entropy weighted method.

[0023] In a second aspect, a customer complaint escalation prediction device is provided, the device comprising:

[0024] The acquisition module is used to obtain the voice call information between the user and the customer service;

[0025] A first determination module is used to perform voice recognition on the voice call information to obtain the conversation content between the user and the customer service, and determine the user's complaint problem based on the conversation content, and determine the user's static complaint escalation probability based on the complaint problem;

[0026] A second determination module is used to extract the voiceprint feature of the user based on the voice call information, and determine the dynamic complaint escalation probability of the user based on the voiceprint feature;

[0027] The third determination module is used to determine whether the user upgrades the customer complaint based on the static complaint upgrade probability and the dynamic complaint upgrade probability.

[0028] Optionally, the first determination module is specifically used to determine the user's complaint issue through a BERT algorithm based on the conversation content.

[0029] Optionally, the second determining module includes:

[0030] A first determining unit, configured to determine a basic emotional voiceprint feature of the user;

[0031] The second determining unit is used to determine the dynamic upgrade probability of the user based on the basic emotional voiceprint feature of the user and the voiceprint feature of the user.

[0032] Optionally, the device further comprises:

[0033] The fourth determination module is used to determine whether the user has called before; and if the user has called before, search the user emotion voiceprint feature library to determine the user's basic emotion voiceprint feature; and if the user calls for the first time, use the initial speech voiceprint feature of the user for a predetermined time length in the call voice information as the basic emotion voiceprint feature.

[0034] Optionally, the second determining module includes:

[0035] A voice separation unit, used to perform voice separation processing on the voice call information to obtain the voice information of the user;

[0036] An extraction unit is used to extract the voiceprint features of the user based on the voice information of the user.

[0037] Optionally, the extraction unit is specifically configured to extract the voiceprint features of the user by using an entropy weighted method.

[0038] In a third aspect, an electronic device is provided, the electronic device comprising:

[0039] one or more processors;

[0040] Memory;

[0041] One or more applications, wherein the one or more applications are stored in a memory and configured to be executed by one or more processors, and the one or more programs are configured to: execute the customer complaint escalation prediction method shown in the first aspect.

[0042] In a fourth aspect, a computer-readable storage medium is provided, which is used to store computer instructions. When the computer-readable storage medium is executed on a computer, the computer can execute the customer complaint escalation prediction method shown in the first aspect.

[0043] The present application provides a customer complaint escalation prediction method, device and electronic device. Compared with the prior art that manually identifies customer complaint escalation, the present application obtains the voice call information between the user and the customer service, performs voice recognition on the voice call information to obtain the conversation content between the user and the customer service, and determines the user's complaint problem based on the conversation content, and determines the user's static complaint escalation probability based on the complaint problem, extracts the user's voiceprint features based on the voice call information, and determines the user's dynamic complaint escalation probability based on the voiceprint features, and determines whether the user upgrades the customer complaint based on the static complaint escalation probability and the dynamic complaint escalation probability. That is, the automatic identification of customer complaint escalation is realized. In addition, determining whether the user upgrades the customer complaint based on the static complaint escalation probability and the dynamic complaint escalation probability can improve the accuracy of customer complaint escalation prediction.

[0044] Additional aspects and advantages of the present application will be partially given in the following description, which will become apparent from the following description, or will be understood through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The above and additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0046] Figure 1 A flowchart of a customer complaint escalation prediction method according to an embodiment of the present application;

[0047] Figure 2 A schematic diagram of the structure of a customer complaint escalation prediction device according to an embodiment of the present application;

[0048] Figure 3 A schematic diagram of the structure of an electronic device according to an embodiment of the present application;

[0049] Figure 4 This is a schematic diagram of the BERT model structure for obtaining user questions in an embodiment of the present application;

[0050] Figure 5 Schematic diagram of the voiceprint feature extraction process according to an embodiment of the present application. DETAILED DESCRIPTION

[0051] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be interpreted as limiting the present application.

[0052] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of features, integers, steps, operations, elements and components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and groups thereof. The term "and" used herein includes all or any unit and all combinations of one or more associated listed items.

[0053] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0054] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0055] Implementation

[0056] The present application embodiment provides a method for predicting customer complaint escalation. Figure 1 As shown, the method may include the following steps:

[0057] Step S101, obtaining the voice call information between the user and the customer service;

[0058] Specifically, the voice information of the conversation between the user and the customer service is obtained. Specifically, the voice information of the conversation can be the voice information of the conversation between the user and the customer service after the user enters the customer service channel. Specifically, before the user enters the customer service channel, it is possible to initially identify whether the user is likely to complain (for example, based on the user's basic information and the user's previous complaint behavior, obtain the problems that the user may complain about, and the probability of complaining about the problem through the DNN algorithm). If the probability of complaint is high, it is assigned to experienced customer service personnel.

[0059] Step S102, performing voice recognition on the voice call information to obtain the conversation content between the user and the customer service, and determining the user's complaint problem based on the conversation content, and determining the user's static complaint escalation probability based on the complaint problem;

[0060] Specifically, the voice call information is subjected to voice recognition by a voice recognition method to obtain the conversation content between the user and the customer service, and the user's complaint problem is determined based on the conversation content, and the user's static complaint escalation probability is determined based on the complaint problem; wherein, the voice recognition method can be implemented by a pre-trained neural network model; wherein, the static complaint escalation probability corresponding to each complaint problem is pre-determined based on big data analysis technology.

[0061] For example, the relationship between the probability of user problems and complaints in two years is statistically analyzed, using Q = (q1, q2, ..., q m ) represents the set of user complaints, N = (n q1 ,n q2 ,…,n qm ) represents the number of complaints for each complaint issue, n qi Represents question q i The number of complaints is, then, for problem q i The probability of complaint escalation is:

[0062]

[0063] Step S103, extracting the voiceprint feature of the user based on the voice call information, and determining the dynamic complaint escalation probability of the user based on the voiceprint feature;

[0064] Specifically, the voiceprint features of the user are extracted by a corresponding voiceprint feature extraction method, and the dynamic complaint upgrade probability of the user is determined based on the voiceprint features.

[0065] Step S104: determining whether the user performs a customer complaint upgrade based on the static complaint upgrade probability and the dynamic complaint upgrade probability.

[0066] Specifically, a weighted calculation is performed on the static complaint escalation probability and the dynamic complaint escalation probability to determine whether the customer service is upgraded.

[0067] For example, using P S P represents the static complaint escalation probability of the user. A represents the dynamic escalation probability of the user's complaint. According to data experience, the user's final complaint escalation probability P can be expressed as:

[0068] P=0.7*P S +0.3*P A Formula 2

[0069] The embodiment of the present application provides a method for predicting customer complaint escalation. Compared with the prior art that manually identifies customer complaint escalation, the present application obtains the voice call information between the user and the customer service, performs voice recognition on the voice call information to obtain the conversation content between the user and the customer service, determines the user's complaint problem based on the conversation content, and determines the user's static complaint escalation probability based on the complaint problem, extracts the user's voiceprint features based on the voice call information, determines the user's dynamic complaint escalation probability based on the voiceprint features, and determines whether the user will escalate the customer complaint based on the static complaint escalation probability and the dynamic complaint escalation probability. That is, the automatic identification of customer complaint escalation is realized. In addition, determining whether the user will escalate the customer complaint based on the static complaint escalation probability and the dynamic complaint escalation probability can improve the accuracy of customer complaint escalation prediction.

[0070] The embodiment of the present application also provides a possible implementation method, wherein the determining the user's complaint problem based on the conversation content in step S102 includes:

[0071] Step S1021 (not shown in the figure), determining the user's complaint issue based on the conversation content through the BERT algorithm.

[0072] For example, after the user enters the customer service channel, the prediction of the user's question will be based on the BERT (Bidirectional Encoder Representations from Transformers) model, which reflects the different meanings of words through context. The BERT model is a feature extraction model composed of bidirectional transformers. The core idea of ​​the attention mechanism used by the transformer is to calculate the relationship between each word in a sentence and all the words in the sentence, and then believe that the relationship between words reflects the relevance and importance of different words in the sentence to a certain extent. The BERT model uses a masked language model during training, randomly masks some tokens in the input, and performs pre-training. At the same time, it adds a sentence-level task, next sentence prediction, randomly replaces some sentences, and uses the previous sentence to predict isNext / notNext. Through these two tasks, large-scale unlabeled corpus is used for optimization, and finally a pre-trained BERT model is obtained. The model in this paper is based on the Google Bert Multilingual pre-trained model, including 12-layers, 768-hidden size, 12-heads, and about 100 million parameters. The question-answering model uses the user's answers to perform multiple classification tasks on the pre-trained model, conducts supervised training, picks out incorrect labels in the corpus with low prediction confidence, re-labels them, and performs fine-tuning. When the average error rate of training predictions no longer decreases significantly (after about 20 to 30 epochs), the model stops fine-tuning and obtains a question recognition model.

[0073] The model structure used in the present invention is as follows Figure 4 As shown in the figure, the text after speech-to-text conversion is preprocessed into a text vector through one-hot encoding. The text vector first passes through a 12-layer attention network, then a fully connected layer, and finally passes through a softmax layer to obtain the user's question. The training data comes from business annotations, and the summarized user questions also come from business experience.

[0074] For the embodiment of the present application, the problem of determining user complaint issues is solved.

[0075] This embodiment of the application provides a possible implementation method, step S103 includes:

[0076] Step S1031 (not shown in the figure), determining the basic emotional voiceprint features of the user;

[0077] Step S1032 (not shown in the figure), determining the dynamic upgrade probability of the user based on the basic emotional voiceprint features of the user and the voiceprint features of the user.

[0078] Specifically, the dynamic upgrade probability of the user may be determined by calculating the inner product of the basic emotional voiceprint feature and the voiceprint feature of the user.

[0079] For example, WMFCC=(M1, M2, M3, ..., M i ,…,M N ) represents the WMFCC feature of a sentence of the user, and w i represents the voiceprint feature of the user's i-th sentence, then:

[0080]

[0081] Use w pr Represents the user's standard voiceprint feature, and the user's dynamic complaint probability is:

[0082]

[0083] For the embodiment of the present application, the specific determination problem of the user's dynamic complaint probability is solved.

[0084] This embodiment of the application provides a possible implementation method, and the method further includes:

[0085] Step S105 (not shown in the figure), determining whether the user has called before;

[0086] Step S106 (not shown in the figure), if the user has called before, search the user emotion voiceprint feature library to determine the basic emotion voiceprint feature of the user;

[0087] Step S107 (not shown in the figure), if the user calls for the first time, the initial speech voiceprint feature of the user for a predetermined time length in the call voice information is used as the basic emotional voiceprint feature.

[0088] For example, it can be determined whether the user has received a call before based on the user's mobile phone number information. If there is an incoming call, the standard emotional features of the user are extracted from the voiceprint database; if there is no incoming call, the voiceprint emotional features of the user's first three sentences are extracted as the user's basic emotional features.

[0089] For the embodiment of the present application, the problem of determining the basic emotional characteristics of the user is solved.

[0090] This embodiment of the present application provides a possible implementation method, wherein the extracting the voiceprint feature of the user based on the voice call information in step S103 includes:

[0091] Step S1031 (not shown in the figure), performing voice separation processing on the voice call information to obtain the voice information of the user;

[0092] Step S1032 (not shown in the figure), extracting the voiceprint features of the user based on the voice information of the user.

[0093] Specifically, the voice call information can be voice separated by a corresponding voice separation algorithm to obtain the user's voice information.

[0094] Specifically, the user's voiceprint features may be extracted through WMFCC, and the user's voiceprint features may be extracted through an entropy weighted method based on the user's voice information.

[0095] Among them, the extraction of speech feature parameters is crucial for voiceprint recognition. In the field of voiceprint recognition, the most commonly used method is the MFCC method, but the voiceprint features extracted by this method have low-order cepstral coefficients and characteristic components, which have poor expression ability for audio files. In order to balance the sensitive parts of the audio features, the present invention calculates the entropy value to analyze the contribution of each dimensional feature parameter to the voiceprint representation, and extracts the voiceprint features through the entropy weighting method, thereby improving the recognition accuracy of the system. For example, the recognition flow chart is as follows: Figure 5 As shown:

[0096] Pre-emphasis and framing

[0097] In order to eliminate the influence of lips and vocal cords during phonation, compensate for the high-frequency part of the speech signal suppressed by the vocal system and highlight the high-frequency formant. Therefore, a first-order differential equation is applied to the speech sample to increase the amplitude of the high-frequency formant, and the speech signal is passed through a high-pass filter:

[0098] H(z)=1-kz -1 Formula 5

[0099] Wherein, k is the pre-emphasis coefficient, whose value is between [0, 1], and 0.97 is usually selected.

[0100] In the framing step, the speech signal is divided into N sampling frames. In order to avoid excessive changes between two adjacent frames, there is an overlapping area between adjacent frames, and the overlapping area contains M sampling frames, and M < N. Generally, the audio signal is framed, assuming that each frame is stable, and generally 20-30ms is used as a frame, with an overlap rate of 25%, 50%, and 75%.

[0101] Window

[0102] The purpose of windowing is to reduce signal discontinuity and make the end smooth enough to connect with the starting point. The commonly used window function is the Hamming window. Assume that the framed signal is S n , where n = 1, ..., N, the windowed signal can be expressed as:

[0103]

[0104] Fast Fourier Transform

[0105] The Fast Fourier Transform (FFT) converts N samples from the time domain to the frequency domain. FFT is used because it is a fast algorithm for implementing the Discrete Fourier Transform (DFT). The formula is as follows:

[0106]

[0107] Among them, s k is the input speech signal, and N is the sampling point of Fourier transform.

[0108] Mel filter analysis

[0109] There are multiple redundant signals in the frequency domain. The Mel filter bank can simplify the amplitude of the frequency domain according to the hearing range of the human ear. The human ear's perception of sound is not linear, and the nonlinear relationship of log can better describe it. The relationship between the Mel frequency and the speech signal is as follows:

[0110]

[0111] Wherein, Mel(f) is the Mel frequency, and f is the frequency of the signal in Hertz.

[0112] DCT Transform

[0113] MFCC is calculated by discrete cosine transform, the formula is as follows:

[0114]

[0115] Among them, N is the filter channel, m j is the strength of the jth Mel filter.

[0116] Parameter weighting

[0117] The main advantage of using cepstral coefficients is that they are uncorrelated with each other. The cepstral coefficients of each order can be easily analyzed. However, the high-frequency cepstral coefficients related to emotional information are very small and cannot be effectively distinguished. Therefore, the present invention uses the entropy method to improve the ability of features to describe features. The entropy weight method is an objective weighting method for calculating the weights between independent variables. The weight of the component is determined based on the information entropy of the calculated component. The larger the entropy value, the less information it carries and the smaller the weight of the component. On the contrary, the smaller the entropy value, the more information it carries and the larger the weight of the component.

[0118] For the voiceprint features shown below:

[0119] MFCC=(M1,M2,M3,…,M i ,…,M N ) Formula 10

[0120] Among them, M i =(mel i(1) ,…,mel i(j) ,…,mel i(D) ) is the i-th feature in MFCC. D is the dimension of the feature, N is the number of frames, mel i(j) The jth feature vector representing the voiceprint feature of the i-th frame.

[0121] First, normalize the features:

[0122]

[0123] Entropy is defined as follows:

[0124]

[0125] in The entropy of the eigenvector is as follows:

[0126]

[0127] Calculate the weight of each component of MFCC, and the new parameters obtained are as follows:

[0128] wxya i =(w1·mel i(1) ,…,w D ·mel i(D) ) Formula 14

[0129] That is, the output of WMFCC is

[0130] wmel=(w1·mel i(1) ,…,w D ·mel i(D) ) Formula 15

[0131] Embodiment 2

[0132] Figure 2 A customer complaint escalation prediction device is provided in an embodiment of the present application. The device 20 includes: an acquisition module 201, a first determination module 202, a second determination module 203, and a third determination module 204, wherein:

[0133] Acquisition module 201, used to obtain voice call information between the user and the customer service;

[0134] A first determination module 202 is configured to perform voice recognition on the voice call information to obtain the conversation content between the user and the customer service, determine the user's complaint problem based on the conversation content, and determine the user's static complaint escalation probability based on the complaint problem;

[0135] A second determination module 203 is configured to extract the voiceprint feature of the user based on the voice call information, and determine the dynamic complaint upgrade probability of the user based on the voiceprint feature;

[0136] The third determination module 204 is used to determine whether the user performs a customer complaint upgrade based on the static complaint upgrade probability and the dynamic complaint upgrade probability.

[0137] The embodiment of the present application provides a customer complaint escalation prediction device. Compared with the prior art that manually identifies customer complaint escalation, the present application obtains the voice call information between the user and the customer service, performs voice recognition on the voice call information to obtain the conversation content between the user and the customer service, and determines the user's complaint problem based on the conversation content, and determines the user's static complaint escalation probability based on the complaint problem, extracts the user's voiceprint features based on the voice call information, and determines the user's dynamic complaint escalation probability based on the voiceprint features, and determines whether the user will escalate the customer complaint based on the static complaint escalation probability and the dynamic complaint escalation probability. That is, the automatic identification of customer complaint escalation is realized. In addition, determining whether the user will escalate the customer complaint based on the static complaint escalation probability and the dynamic complaint escalation probability can improve the accuracy of customer complaint escalation prediction.

[0138] The customer complaint escalation prediction device of this embodiment can execute a customer complaint escalation prediction method provided in the above embodiments of the present application. The implementation principle is similar and will not be repeated here.

[0139] An embodiment of the present application provides a possible implementation method, wherein the first determination module is specifically used to determine the user's complaint issue through the BERT algorithm based on the conversation content.

[0140] The embodiment of the present application provides a possible implementation manner. Specifically, the second determining module includes:

[0141] A first determining unit, configured to determine a basic emotional voiceprint feature of the user;

[0142] The second determining unit is used to determine the dynamic upgrade probability of the user based on the basic emotional voiceprint feature of the user and the voiceprint feature of the user.

[0143] The embodiment of the present application provides a possible implementation method. Further, the device also includes:

[0144] The fourth determination module is used to determine whether the user has called before; and if the user has called before, search the user emotion voiceprint feature library to determine the user's basic emotion voiceprint feature; and if the user calls for the first time, use the initial speech voiceprint feature of the user for a predetermined time length in the call voice information as the basic emotion voiceprint feature.

[0145] This embodiment of the present application provides a possible implementation manner, where the second determining module includes:

[0146] A voice separation unit, used to perform voice separation processing on the voice call information to obtain the voice information of the user;

[0147] An extraction unit is used to extract the voiceprint features of the user based on the voice information of the user.

[0148] Specifically, the extraction unit is used to extract the voiceprint features of the user through an entropy weighted method.

[0149] The embodiment of the present application provides a customer complaint escalation prediction device, which is applicable to the method shown in the above embodiment and will not be described in detail here.

[0150] Embodiment 3

[0151] The present application embodiment provides an electronic device, such as Figure 3 As shown, Figure 3 The electronic device 30 shown includes: a processor 3001 and a memory 3003. The processor 3001 and the memory 3003 are connected, such as through a bus 3002. Further, the electronic device 30 may also include a transceiver 3003. It should be noted that in actual applications, the transceiver 3004 is not limited to one, and the structure of the electronic device 30 does not constitute a limitation on the embodiments of the present application. The processor 3001 is used in the embodiments of the present application to implement Figure 2 The functions of the modules shown are as follows: The transceiver 3003 includes a receiver and a transmitter.

[0152] Processor 3001 may be a CPU, a general purpose processor, a DSP, an ASIC, an FPGA or other programmable logic device, a transistor logic device, a hardware component or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 3001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0153] The bus 3002 may include a path to transmit information between the above components. The bus 3002 may be a PCI bus or an EISA bus, etc. The bus 3002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0154] The memory 3003 may be a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, or an EEPROM, a CD-ROM or other optical disk storage, an optical disk storage (including a compressed optical disk, a laser disk, an optical disk, a digital versatile disk, a Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0155] The memory 3003 is used to store the application code for executing the solution of the present application, and the execution is controlled by the processor 3001. The processor 3001 is used to execute the application code stored in the memory 3003 to implement Figure 2 The illustrated embodiment provides the functions of a customer complaint escalation prediction device.

[0156] The embodiment of the present application provides an electronic device. Compared with the prior art that manually identifies customer complaint upgrades, the present application obtains the voice call information between the user and the customer service, performs voice recognition on the voice call information to obtain the conversation content between the user and the customer service, and determines the user's complaint problem based on the conversation content, and determines the user's static complaint upgrade probability based on the complaint problem, extracts the user's voiceprint features based on the voice call information, and determines the user's dynamic complaint upgrade probability based on the voiceprint features, and determines whether the user upgrades the customer complaint based on the static complaint upgrade probability and the dynamic complaint upgrade probability. That is, the automatic identification of customer complaint upgrades is realized. In addition, determining whether the user upgrades the customer complaint based on the static complaint upgrade probability and the dynamic complaint upgrade probability can improve the accuracy of customer complaint upgrade prediction.

[0157] The present application embodiment provides an electronic device applicable to the above method embodiment, which will not be described in detail here.

[0158] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the method shown in the above embodiment is implemented.

[0159] The embodiment of the present application provides a computer-readable storage medium. Compared with the prior art that manually identifies customer complaint escalation, the present application obtains the voice call information between the user and the customer service, performs voice recognition on the voice call information to obtain the conversation content between the user and the customer service, and determines the user's complaint problem based on the conversation content, and determines the user's static complaint escalation probability based on the complaint problem, extracts the user's voiceprint features based on the voice call information, and determines the user's dynamic complaint escalation probability based on the voiceprint features, and determines whether the user upgrades the customer complaint based on the static complaint escalation probability and the dynamic complaint escalation probability. That is, the automatic identification of customer complaint escalation is realized. In addition, determining whether the user upgrades the customer complaint based on the static complaint escalation probability and the dynamic complaint escalation probability can improve the accuracy of customer complaint escalation prediction.

[0160] The present application embodiment provides a computer-readable storage medium applicable to the above method embodiment, which will not be described in detail here.

[0161] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0162] The above are only some implementation methods of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for predicting customer complaint escalation, characterized in that: include: Get the voice call information between the user and the customer service; Based on the user's basic information and previous complaints, the DNN algorithm is used to obtain the issues that the user may complain about, as well as the probability of complaining about the issue. If the complaint probability is high, it will be assigned to experienced customer service personnel. Performing voice recognition on the voice call information to obtain the conversation content between the user and the customer service, and determining the user's complaint problem through the BERT algorithm based on the conversation content, and determining the user's static complaint escalation probability based on the complaint problem; Extracting the voiceprint feature of the user based on the voice call information, and determining the dynamic complaint escalation probability of the user based on the voiceprint feature; A weighted calculation is performed based on the static complaint escalation probability and the dynamic complaint escalation probability to determine whether the user escalates the customer complaint.

2. The method according to claim 1, characterized in that The extracting the voiceprint feature of the user based on the voice call information, and determining the dynamic complaint upgrade probability of the user based on the voiceprint feature, includes: Determining the basic emotional voiceprint characteristics of the user; The dynamic upgrade probability of the user is determined based on the basic emotional voiceprint feature of the user and the voiceprint feature of the user.

3. The method according to claim 1, characterized in that The method further includes: Determining whether the user has called before; If the user has called before, searching the user emotion voiceprint feature library to determine the basic emotion voiceprint feature of the user; If the user calls for the first time, the initial speech voiceprint feature of the user for a predetermined time length in the voice call information is used as the basic emotional voiceprint feature.

4. The method according to any one of claims 1 to 3, characterized in that: The extracting the voiceprint feature of the user based on the voice call information includes: Performing voice separation processing on the voice call information to obtain the voice information of the user; The voiceprint feature of the user is obtained based on the voice information of the user.

5. The method according to claim 4, characterized in that The extracting the voiceprint feature of the user based on the voice information of the user includes: The voiceprint features of the user are extracted based on the voice information of the user through the entropy weighted method.

6. A customer complaint escalation prediction device, characterized in that: The device also includes: The acquisition module is used to obtain the voice call information between the user and the customer service. Based on the user's basic information and the user's previous complaint behavior, the DNN algorithm is used to obtain the issues that the user may complain about, as well as the probability of complaining about the issue. If the complaint probability is high, it will be assigned to experienced customer service personnel. A first determination module is used to perform voice recognition on the voice call information to obtain the conversation content between the user and the customer service, and determine the user's complaint problem based on the conversation content through the BERT algorithm, and determine the user's static complaint escalation probability based on the complaint problem; A second determination module is used to extract the voiceprint feature of the user based on the voice call information, and determine the dynamic complaint escalation probability of the user based on the voiceprint feature; The third determination module is used to determine whether the user upgrades the customer complaint based on a weighted calculation of the static complaint upgrade probability and the dynamic complaint upgrade probability.

7. An electronic device, characterized in that: include: one or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to: execute the customer complaint escalation prediction method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer storage medium is used to store computer instructions, and when the computer storage medium is executed on a computer, the computer can execute the customer complaint escalation prediction method according to any one of claims 1 to 5.

Images