Method and apparatus for detecting quality of service

Abstract

The embodiment of the application discloses a service quality detection method and device, the method comprises the following steps: obtaining service data to be detected of a customer service providing a service for a customer, and determining a service quality evaluation value of the service according to the service data to be detected; obtaining service data of historical qualified services of the customer service, and determining a service difference degree value of the service and the service data of the historical qualified services of the customer service according to the service data to be detected and the service data of the historical qualified services; and determining whether the service provided by the customer service is a qualified service according to the service quality evaluation value and the service difference degree value. By adopting the embodiment of the application, the accuracy of service quality detection can be improved, the efficiency of service quality detection can be improved, and the applicability is high.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and in particular to a service quality testing method and apparatus. Background Technology

[0002] With the continuous growth of business volume and user base, service ticket voice messages, service ticket text messages, and user feedback on customer service from other channels have accumulated into a massive amount of heterogeneous service data from multiple channels. To improve customer service levels and quality, it is typically necessary to perform quality checks on this service data to identify employees who do not meet service standards. Currently, a common quality inspection method involves randomly sampling a certain percentage of service tickets to check employee service quality. Quality inspectors listen to the entire audio recording during these checks and make quality judgments based on the rules corresponding to each archive item. However, this method of quality judgment through random sampling is inaccurate and inefficient. Summary of the Invention

[0003] This application provides a service quality testing method and apparatus, which can improve the accuracy and efficiency of service quality testing and has high applicability.

[0004] In a first aspect, embodiments of this application provide a service quality detection method, the method comprising:

[0005] Obtain the service data to be inspected for the customer service to provide services, and determine the service quality evaluation value of the above services based on the above service data;

[0006] Obtain the service data of the above customer service's historical qualified services, and determine the service difference value between the above service and the above customer service's historical qualified services based on the above service data to be inspected and the above service data of historical qualified services.

[0007] Based on the above service quality evaluation value and the above service difference value, determine whether the above service provided by the above customer service is a qualified service.

[0008] In conjunction with the first aspect, in one possible implementation, determining the service quality evaluation value of the service based on the service data to be inspected includes:

[0009] Obtain the first feature vector corresponding to the above-mentioned service data to be inspected;

[0010] The first feature vector is input into the service quality prediction model, and the service quality evaluation value of the service corresponding to the first feature vector is obtained through the service quality prediction model. The service quality prediction model is trained by the service data corresponding to the historical services of at least one customer service representative and the preset service quality evaluation value. The historical services include at least one historical qualified service and at least one historical unqualified service.

[0011] In conjunction with the first aspect, in one possible implementation, determining the service difference value between the service to be inspected and the historical qualified service data based on the service data to be inspected and the historical qualified service data of the customer service includes:

[0012] Obtain the first feature vector corresponding to the above-mentioned service data to be inspected, and obtain the second feature vector corresponding to the service data of the above-mentioned historical qualified services;

[0013] The first feature vector is encoded based on the second feature vector to obtain the third feature vector corresponding to the first feature vector.

[0014] The service difference value between the above service and the above historical qualified service is determined based on the first feature vector and the third feature vector.

[0015] In conjunction with the first aspect, in one possible implementation, determining the service difference value between the service and the historically qualified service based on the first feature vector and the third feature vector includes:

[0016] Calculate the vector similarity value between the first feature vector and the third feature vector, and determine the vector similarity value as the service difference value between the service and the historical qualified service.

[0017] In conjunction with the first aspect, in one possible implementation, determining whether the service provided by the customer service representative is a qualified service based on the service quality evaluation value and the service difference value includes:

[0018] Obtain a first weight and a second weight. The first weight is used to mark the weight of the service quality evaluation value, and the second weight is used to mark the weight of the service difference value. The sum of the first weight and the second weight is equal to 1.

[0019] A first weighted service quality evaluation value is determined based on the first weight and the service quality evaluation value, and a second weighted service quality evaluation value is determined based on the second weight and the service difference value.

[0020] The service quality assessment value is determined based on the first weighted service quality evaluation value and the second weighted service quality evaluation value mentioned above.

[0021] If the above service quality assessment value is less than the service testing threshold, the above service is determined to be a qualified service; if the above service quality assessment value is not less than the above service testing threshold, the above service is determined to be a unqualified service.

[0022] In conjunction with the first aspect, in one possible implementation, the above-mentioned acquisition of pending service data provided by customer service includes:

[0023] The service voice recordings provided by the customer service representatives are obtained, and speech recognition is performed on these service voice recordings to obtain their text data.

[0024] Based on the aforementioned service voice and text data, generate the service data to be inspected corresponding to the aforementioned service.

[0025] In conjunction with the first aspect, in one possible implementation, acquiring pending service data provided by customer service includes:

[0026] The above-mentioned customer service voice recordings are obtained, and speech recognition is performed on the above-mentioned voice recordings to obtain the corresponding text data.

[0027] Obtain the historical service data of the aforementioned customer service representatives and / or the basic employee information of the aforementioned customer service representatives, and generate the service data to be inspected for the aforementioned services based on the aforementioned service voice, the aforementioned text data, the aforementioned historical service data of the aforementioned customer service representatives and / or the aforementioned basic employee information of the aforementioned customer service representatives.

[0028] In conjunction with the first aspect, in one possible implementation, the service data to be inspected includes the customer service voice messages provided by the customer service representative; obtaining the first feature vector corresponding to the service data to be inspected includes:

[0029] A first preset sampling frequency is obtained, the service voice is sampled based on the first preset sampling frequency to obtain a voice sampling signal, and the voice sampling signal is subjected to frame-by-frame windowing processing to obtain at least one frame signal that constitutes the voice sampling signal.

[0030] Short-time energy and / or short-time zero-crossing rate are extracted from each of the at least one framed signal to serve as audio feature parameters, and a first feature vector is determined based on the audio feature parameters.

[0031] In conjunction with the first aspect, in one possible implementation, obtaining the first feature vector corresponding to the service data to be inspected includes:

[0032] Determine the audio feature parameters of the service voice in the service data to be inspected, including short-time energy and / or short-time zero-crossing rate;

[0033] Based on the above text data, a first statistical feature parameter of the above service is determined. The first statistical feature parameter includes at least one of the following: the service type of the above service, the number of times the customer service representative speaks, the repetition rate of the customer service response, and the degree of customer service representative impatience.

[0034] The first feature vector corresponding to the service data to be inspected is determined based on the above-mentioned audio feature parameters and the above-mentioned first statistical feature parameters.

[0035] In conjunction with the first aspect, in one possible implementation, obtaining the first feature vector corresponding to the service data to be inspected includes:

[0036] Determine the audio feature parameters of the service voice in the service data to be inspected, including short-time energy and / or short-time zero-crossing rate;

[0037] The first statistical feature parameter is determined based on the text data in the service data to be inspected. The first statistical feature parameter includes at least one of the service type of the service, the number of times the customer service representative speaks, the repetition rate of the customer service reply, and the customer service representative's impatience.

[0038] A second statistical characteristic parameter is determined based on the historical service data in the aforementioned service data to be inspected. This second statistical characteristic parameter includes the number of repeat calls from the aforementioned customer and / or the cumulative number of customer service errors; and / or

[0039] The third statistical feature parameter is determined based on the basic employee information of the customer service staff in the above-mentioned service data to be inspected. The third statistical feature parameter includes at least one of the customer service staff’s employment time, customer service staff’s gender, and customer service staff’s age.

[0040] The first feature vector is determined based on the aforementioned audio feature parameters, the aforementioned first statistical feature parameters, the aforementioned second statistical feature parameters, and / or the aforementioned third statistical feature parameters.

[0041] Secondly, embodiments of this application provide a service quality detection device, the device comprising:

[0042] The service quality evaluation value determination module is used to obtain the service data to be inspected provided by customer service to customers, and to determine the service quality evaluation value of the service based on the service data to be inspected.

[0043] The service difference value determination module is used to obtain the service data of the above customer service's historical qualified services, and determine the service difference value between the above service and the above customer service's historical qualified services based on the above service data to be inspected and the above service data of historical qualified services.

[0044] The service determination module is used to determine whether the service provided by the customer service representative is a qualified service based on the service quality evaluation value and the service difference value.

[0045] In conjunction with the second aspect, in one possible implementation, the service quality evaluation value determination module includes:

[0046] The first feature vector acquisition unit is used to acquire the first feature vector corresponding to the above-mentioned service data to be inspected.

[0047] The service quality evaluation value determination unit is used to input the first feature vector into the service quality prediction model and obtain the service quality evaluation value of the service corresponding to the first feature vector through the service quality prediction model. The service quality prediction model is trained by service data corresponding to the historical services of at least one customer service representative and preset service quality evaluation values. The historical services include at least one historical qualified service and at least one historical unqualified service.

[0048] In conjunction with the second aspect, in one possible implementation, the aforementioned service difference value determination module includes:

[0049] The feature vector acquisition unit is used to acquire the first feature vector corresponding to the service data to be inspected, and to acquire the second feature vector corresponding to the service data of the historical qualified services.

[0050] The encoding processing unit is used to encode the first feature vector based on the second feature vector to obtain the third feature vector corresponding to the first feature vector.

[0051] The service difference value determination unit is used to determine the service difference value between the service and the historical qualified service based on the first feature vector and the third feature vector.

[0052] In conjunction with the second aspect, in one possible implementation, the aforementioned service difference value determination unit is specifically used for:

[0053] Calculate the vector similarity value between the first feature vector and the third feature vector, and determine the vector similarity value as the service difference value between the service and the historical qualified service.

[0054] In conjunction with the second aspect, in one possible implementation, the above-mentioned service determination module includes:

[0055] The weight acquisition unit is used to acquire a first weight and a second weight. The first weight is used to mark the weight of the service quality evaluation value, and the second weight is used to mark the weight of the service difference value. The sum of the first weight and the second weight is equal to 1.

[0056] The weighted calculation unit is used to determine a first weighted service quality evaluation value based on the first weight and the service quality evaluation value, and to determine a second weighted service quality evaluation value based on the second weight and the service difference value.

[0057] The evaluation value determination unit is used to determine the service quality evaluation value based on the first weighted service quality evaluation value and the second weighted service quality evaluation value.

[0058] The service determination unit is used to determine that the service is a qualified service if the service quality assessment value is less than the service detection threshold, and to determine that the service is an unqualified service if the service quality assessment value is not less than the service detection threshold.

[0059] In conjunction with the second aspect, in one possible implementation, the aforementioned service quality evaluation value determination module or service difference value determination module is further used for:

[0060] The service voice recordings provided by the customer service representatives are obtained, and speech recognition is performed on these service voice recordings to obtain their text data.

[0061] Based on the aforementioned service voice and text data, generate the service data to be inspected corresponding to the aforementioned service.

[0062] In conjunction with the second aspect, in one possible implementation, the aforementioned service quality evaluation value determination module or service difference value determination module is further used for:

[0063] The above-mentioned customer service voice recordings are obtained, and speech recognition is performed on the above-mentioned voice recordings to obtain the corresponding text data.

[0064] Obtain the historical service data of the aforementioned customer service representatives and / or the basic employee information of the aforementioned customer service representatives, and generate the service data to be inspected for the aforementioned services based on the aforementioned service voice, the aforementioned text data, the aforementioned historical service data of the aforementioned customer service representatives and / or the aforementioned basic employee information of the aforementioned customer service representatives.

[0065] In conjunction with the second aspect, in one possible implementation, the service data to be inspected includes the customer service voice provided by the customer service representative; the first feature vector acquisition unit or feature vector acquisition unit is specifically used for:

[0066] A first preset sampling frequency is obtained, the service voice is sampled based on the first preset sampling frequency to obtain a voice sampling signal, and the voice sampling signal is subjected to frame-by-frame windowing processing to obtain at least one frame signal that constitutes the voice sampling signal.

[0067] Short-time energy and / or short-time zero-crossing rate are extracted from each of the at least one framed signal to serve as audio feature parameters, and a first feature vector is determined based on the audio feature parameters.

[0068] In conjunction with the second aspect, in one possible implementation, the aforementioned first feature vector acquisition unit or feature vector acquisition unit is specifically used for:

[0069] Determine the audio feature parameters of the service voice in the service data to be inspected, including short-time energy and / or short-time zero-crossing rate;

[0070] Based on the above text data, a first statistical feature parameter of the above service is determined. The first statistical feature parameter includes at least one of the following: the service type of the above service, the number of times the customer service representative speaks, the repetition rate of the customer service response, and the degree of customer service representative impatience.

[0071] The first feature vector corresponding to the service data to be inspected is determined based on the above-mentioned audio feature parameters and the above-mentioned first statistical feature parameters.

[0072] In conjunction with the second aspect, in one possible implementation, the aforementioned first feature vector acquisition unit or feature vector acquisition unit is specifically used for:

[0073] Determine the audio feature parameters of the service voice in the service data to be inspected, including short-time energy and / or short-time zero-crossing rate;

[0074] The first statistical feature parameter is determined based on the text data in the service data to be inspected. The first statistical feature parameter includes at least one of the service type of the service, the number of times the customer service representative speaks, the repetition rate of the customer service reply, and the customer service representative's impatience.

[0075] A second statistical characteristic parameter is determined based on the historical service data in the aforementioned service data to be inspected. This second statistical characteristic parameter includes the number of repeat calls from the aforementioned customer and / or the cumulative number of customer service errors; and / or

[0076] The third statistical feature parameter is determined based on the basic employee information of the customer service staff in the above-mentioned service data to be inspected. The third statistical feature parameter includes at least one of the customer service staff’s employment time, customer service staff’s gender, and customer service staff’s age.

[0077] The first feature vector is determined based on the aforementioned audio feature parameters, the aforementioned first statistical feature parameters, the aforementioned second statistical feature parameters, and / or the aforementioned third statistical feature parameters.

[0078] Thirdly, embodiments of this application provide a terminal device, which includes a processor, a memory, and a transceiver, all interconnected. The memory stores a computer program that supports the terminal device in executing the methods provided in the first aspect and / or any possible implementation of the first aspect. The computer program includes program instructions, and the processor and transceiver are configured to invoke the program instructions to execute the methods provided in the first aspect and / or any possible implementation of the first aspect.

[0079] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the methods provided in the first aspect and / or any possible implementation of the first aspect.

[0080] In this embodiment, service quality evaluation value is determined based on the service data provided by customer service representatives. Historical qualified service data is also obtained, and a service difference value between the current service and the historical qualified service data is determined. Finally, the service quality evaluation value and the service difference value are used to determine whether the service provided by the customer service representative is qualified. In this embodiment, service quality detection based on the service quality evaluation value and the service difference value improves the accuracy and efficiency of service quality detection, and has high applicability. Attached Figure Description

[0081] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0082] Figure 1 This is a schematic diagram of a network architecture provided in an embodiment of this application;

[0083] Figure 2 A flowchart illustrating the service quality testing method provided in this application embodiment;

[0084] Figure 3 This is a schematic diagram illustrating the calculation process of customer service response repetition rate provided in an embodiment of this application;

[0085] Figure 4 This is a schematic diagram of the structure of a self-encoder;

[0086] Figure 5 This is another flowchart illustrating the service quality testing method provided in the embodiments of this application;

[0087] Figure 6 This is a schematic diagram of the service quality detection device provided in an embodiment of this application;

[0088] Figure 7 This is another structural schematic diagram of the service quality testing device provided in the embodiments of this application;

[0089] Figure 8This is a schematic diagram of the structure of the terminal device provided in the embodiments of this application. Detailed Implementation

[0090] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0091] Please see Figure 1 , Figure 1 This is a schematic diagram of a network architecture provided in an embodiment of this application. Figure 1 As shown, the network architecture may include server 10 and a terminal device cluster. The terminal device cluster may include one or more terminal devices; the number of terminal devices will not be limited here. Figure 1 As shown, the multiple terminal devices may specifically include terminal device 100a, terminal device 101a, terminal device 102a, etc.; for example Figure 1 As shown, terminal devices such as terminal device 100a, terminal device 101a, and terminal device 102a can all connect to server 10 via the network, so that each terminal device can interact with server 10 through the network connection.

[0092] like Figure 1 The server 10 shown can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, and big data and artificial intelligence platforms. The terminal device can be a smartphone, tablet, laptop, desktop computer, smart TV, or other smart terminal. The following description uses the communication between terminal device 100a and server 10 as an example to illustrate the specific implementation of this application.

[0093] The service quality detection method provided in this application involves a server acquiring customer service voice messages from a terminal device. Based on these voice messages, the server can determine the corresponding service data to be inspected. A service quality evaluation value is determined based on the service data, and a service difference value is determined between the service data and the customer service representative's historical qualified services. Therefore, the service quality evaluation value and the service difference value can be used to determine whether the service provided by the customer service representative is qualified. Using this application's embodiment improves the accuracy and efficiency of service quality detection and has high applicability.

[0094] The following will combine Figures 2 to 8 The methods and related apparatuses provided in the embodiments of this application will be described in detail respectively.

[0095] Please see Figure 2 , Figure 2 This is a flowchart illustrating the service quality detection method provided in this application embodiment. The method provided in this application embodiment may include the following steps S201 to S203:

[0096] S201. Obtain the service data to be inspected for the customer service representatives, and determine the service quality evaluation value based on the service data to be inspected.

[0097] In some feasible implementations, with the development of online businesses, such as the rise of online shopping malls and online medical platforms, it is usually necessary to configure human customer service representatives to provide after-sales service, consultation service, and other services to users in order to provide users with more satisfactory services. In order to improve the service level and quality of human customer service representatives, it is usually necessary to conduct service quality testing to screen out employees who do not meet service standards. In this application, the server obtains the service data to be tested from the terminal device, and can determine the service quality evaluation value based on the service data.

[0098] Specifically, the server obtains the service voice recordings from the terminal device when customer service representatives provide services to customers, and determines the service data to be inspected corresponding to this service based on the service voice recordings. The services provided by customer service representatives to customers can be complaints, inquiries, suggestions, etc. Service voice recordings can be understood as recordings of the communication process between customer service representatives and customers. Typically, the terminal device can collect the voice recordings of customer service representatives providing services to customers through a microphone or microphone array (i.e., multiple microphones arranged in a row). After the terminal device collects the service voice recordings between customer service representatives and customers, it can send the service voice recordings to the server for processing via wired or wireless communication. Alternatively, the terminal device can directly process the service voice recordings to achieve service quality inspection of the customer service representative providing a particular service; no limitation is made here. For ease of description, the following embodiments of this application use a server as an example.

[0099] In some feasible implementations, once the server obtains the service voice, it can identify the service voice as the service data to be inspected corresponding to the service, and then perform subsequent processing based on the service voice.

[0100] Optionally, in some feasible implementations, after the server obtains the service voice, it can also perform speech recognition on the service voice using the automated speech recognition (ASR) system mounted on the server to convert the service voice into text data. Then, based on the service voice and text data, it generates service data to be inspected corresponding to this service, and performs subsequent processing on the service voice and text data in the service data to be inspected.

[0101] Optionally, in some feasible implementations, in addition to acquiring the service voice and the corresponding text data, the customer service representative's historical service data and / or the customer service representative's basic employee information can also be acquired. Then, service data to be inspected can be generated based on the service voice, text data, historical service data and / or employee basic information, so as to perform subsequent processing on the various data included in the service data to be inspected.

[0102] Specifically, after obtaining the service data for a specific service provided by a customer service representative (hereinafter referred to as the service to be tested for ease of description), the service quality evaluation value of the service to be tested can be determined based on the service data. Determining the service quality evaluation value based on the service data can be understood as follows: obtaining the first feature vector corresponding to the service data, inputting the first feature vector into the service quality prediction model, and obtaining the service quality evaluation value corresponding to the service based on the first feature vector through the service quality prediction model. The service quality prediction model is trained using service data corresponding to historical services of at least one customer service representative and preset service quality evaluation values. Historical services include at least one historical qualified service and at least one historical unqualified service. In other words, the service quality prediction model is trained using supervised learning based on service data corresponding to historical qualified services of different customer service representatives, service data corresponding to historical unqualified services of different customer service representatives, and preset service quality evaluation values ​​corresponding to each service. Understandably, in this embodiment, a larger service quality evaluation value determined by the service quality prediction model indicates a lower service quality, and a smaller service quality evaluation value indicates a higher service quality.

[0103] In some feasible implementations, if the service data to be inspected includes service voice, then the above-mentioned acquisition of the first feature vector corresponding to the service data to be inspected can be understood as follows: acquiring a first preset sampling frequency, sampling the service voice based on the first preset sampling frequency to obtain a voice sampling signal, and performing frame-by-frame windowing processing on the voice sampling signal to obtain at least one frame signal constituting the voice sampling signal. Short-time energy and / or short-time zero-crossing rate are extracted from each frame signal in the at least one frame signal as audio feature parameters, and the first feature vector is determined based on the audio feature parameters. It is understood that the window function used in the above-mentioned frame-by-frame windowing processing can be a Hanning window or a Hamming window, etc., and is not limited here.

[0104] Optionally, in some feasible implementations, if the service data to be inspected includes not only service voice but also text data corresponding to the service voice, then the aforementioned acquisition of the first feature vector corresponding to the service data to be inspected can be understood as: determining the audio feature parameters of the service voice in the service data to be inspected, and determining the first statistical feature parameters of the service based on the text data, and then determining the first feature vector based on the audio feature parameters and the first statistical feature parameters. The audio feature parameters include short-time energy and / or short-time zero-crossing rate, and the first statistical feature parameters include at least one of the following: service type, number of times the customer service representative speaks, customer service response repetition rate, and customer service impatience level. Generally, the service type can include risk control type, complaint type, consultation type, and suggestion type, etc. Customer service response repetition rate is used to detect whether the customer service representative gives mechanical replies, i.e., whether the customer service representative always repeats the same words in the service call text (i.e., the text data corresponding to the service voice). Customer service impatience level is used to measure the patience of the customer service representative when providing service to the customer.

[0105] For ease of understanding, this application uses the calculation of customer service response repetition as an example. The specific calculation process for customer service response repetition is as follows: First, for the customer service responses included in the text data corresponding to the service voice (e.g., each answer from the customer service representative to a customer's question), each response can be vectorized using the Term Frequency (TF)-Inverse Document Frequency (IDF) method to obtain a vector corresponding to each response. Second, the similarity between each pair of responses is calculated to obtain an n*n similarity matrix, where n represents the number of responses from the customer service representative in the service voice, and n is an integer greater than 1. Next, based on the similarity matrix, the sum of the similarities between each response and other responses is calculated as the total similarity value between this response and other responses. Finally, the maximum total similarity value among the multiple total similarity values ​​is determined as the customer service response repetition rate.

[0106] For example, suppose customer service representative A provides a service message to customer B, and replies to customer B's question three times. These three replies are referred to as Answer 1 to Answer 3. By vectorizing Answer 1 to Answer 3, we can obtain vector 1 for Answer 1, vector 2 for Answer 2, and vector 3 for Answer 3. For each answer, the similarity value between that answer and every other answer can be calculated. For example, taking cosine similarity as an example, for vector 1, the cosine similarity between vector 1 and each of vectors 1 to 3 can be calculated to obtain the cosine similarity 1-1 between vector 1 and vector 1, the cosine similarity 1-2 between vector 1 and vector 2, and the cosine similarity 1-3 between vector 1 and vector 3. For vector 2, the cosine similarity between vector 2 and each of vectors 1 to 3 can be calculated to obtain the cosine similarity 2-1 between vector 2 and vector 1, the cosine similarity 2-2 between vector 2 and vector 2, and the cosine similarity 2-3 between vector 2 and vector 3. Similarly, the cosine similarity between vector 3 and each of vectors 1 to 3 can be calculated to obtain the cosine similarity 3-1 between vector 3 and vector 1, the cosine similarity 3-2 between vector 3 and vector 2, and the cosine similarity 3-3 between vector 3 and vector 3. Furthermore, for each answer, the total similarity value between that answer and other answers can be obtained by calculating the sum of the similarities between each answer and other answers. For example, for answer 1, the total similarity value 1 between answer 1 and other answers (i.e., answers 1 to 3) is 1 = cosine similarity 1 - 1 + cosine similarity 1 - 2 + cosine similarity 1 - 3; for answer 2, the total similarity value 2 between answer 2 and other answers is 2 = cosine similarity 2 - 1 + cosine similarity 2 - 2 + cosine similarity 2 - 3; and for answer 3, the total similarity value 3 between answer 3 and other answers is 3 = cosine similarity 3 - 1 + cosine similarity 3 - 2 + cosine similarity 3 - 3. Therefore, the maximum value among the total similarity values ​​1, 2, and 3 can be determined as the repetition rate of the customer service response.

[0107] Optionally, in some feasible implementations, to make the calculation of customer service response repetition more accurate, for each customer service response included in the text data corresponding to the service voice, each response can be vectorized using at least two vectorization methods to obtain different vectorized representations of each response after quantization using different vectorization methods. Then, for each vector corresponding to each response obtained based on each vectorization method, the similarity between each pair of responses is calculated to obtain an n*n similarity matrix, where n represents the number of customer service responses in the service voice. Next, based on the similarity matrix, the sum of the similarities between each response and other responses is calculated as the total similarity value between this response and other responses. Finally, the multiple total similarity values ​​corresponding to various vectorization methods are summed to obtain multiple fused total similarity values. Then, the maximum fused total similarity value among the multiple fused total similarity values ​​is determined as the customer service response repetition rate of that customer service representative.

[0108] For ease of understanding, this application uses at least two vectorization methods, namely TF-IDF and word2vec, as examples for illustration. Please refer to... Figure 3 , Figure 3 This is a schematic diagram illustrating the calculation process of customer service response repetition rate according to an embodiment of this application. Assume that in the text data corresponding to the service voice of a certain service to be tested, the customer service representative answered the customer's question n times, i.e., the customer service responses are answer 1 to answer n. Each response is vectorized using both TF-IDF and word2vec methods, resulting in n vectors obtained by vectorizing the n responses using each vectorization method. For example... Figure 3 As shown, after vectorizing each of the n answers using the TF-IDF method, we obtain vector F1 for answer 1, vector F2 for answer 2, vector F3 for answer 3, ..., vector Fn for answer n. Similarly, after vectorizing each of the n answers using the word2vec method, we obtain vector F1' for answer 1, vector F2' for answer 2, vector F3' for answer 3, ..., vector Fn' for answer n. Then, for each of the n answers, we can calculate the similarity between each pair of answers based on the vector corresponding to each answer, resulting in an n*n similarity matrix, where n represents the number of times customer service responded in the service voice message. Figure 3As shown, by calculating the cosine similarity between vector F1 and each vector from F1 to Fn, vector F2 and each vector from F1 to Fn, vector F3 and each vector from F1 to Fn, ..., vector Fn and each vector from F1 to Fn, the similarity matrix S can be obtained. Similarly, by calculating the cosine similarity between each vector from F1' to Fn', vector F2' to F1' to Fn', vector F3' to F1' to Fn', ..., vector Fn' to F1' to Fn', the similarity matrix S' can be obtained. Furthermore, for the similarity matrix S, by summing the columns of the similarity matrix S and normalizing, the total similarity value between each answer and other answers can be obtained (e.g., ...). Figure 3 The matrix SUM shown is used to calculate the total similarity value between each answer and other answers by summing the columns of the similarity matrix S' and then normalizing it. Figure 3 The matrix SUM' is shown. Further, matrices SUM and SUM' can be added together to obtain the total fusion similarity value between each answer and other answers. Finally, the multiple total fusion similarity values ​​are sorted in descending or ascending order to determine the maximum total fusion similarity value as the customer service response repetition rate of the customer service representative.

[0109] Optionally, in some feasible implementations, if the service data to be tested includes not only service voice and corresponding text data, but also historical service data from customer service, then the aforementioned acquisition of the first feature vector corresponding to the service data to be tested can be understood as: determining audio feature parameters based on the service voice, determining a first statistical feature parameter based on the corresponding text data, and determining a second statistical feature parameter based on the historical service data in the service data to be tested. Then, the first feature vector is determined based on the audio feature parameters, the first statistical feature parameter, and the second statistical feature parameter. The audio feature parameters include short-time energy and / or short-time zero-crossing rate; the first statistical feature parameter includes at least one of the service type, the number of times customer service representatives speak, the repetition rate of customer service replies, and the level of customer service impatience; and the second statistical feature parameter includes the number of repeat calls from customers and / or the cumulative number of service errors by customer service representatives.

[0110] Optionally, if the service data to be tested includes not only service voice and corresponding text data, but also basic employee information of the customer service representative, then the above-mentioned acquisition of the first feature vector corresponding to the service data to be tested can be understood as: determining audio feature parameters based on the service voice, determining first statistical feature parameters based on the corresponding text data of the service voice, and determining third statistical feature parameters based on the basic employee information of the customer service representative in the service data to be tested. Then, the first feature vector is determined based on the audio feature parameters, the first statistical feature parameters, and the third statistical feature parameters. The audio feature parameters include short-time energy and / or short-time zero-crossing rate; the first statistical feature parameters include at least one of the following: service type, number of times the customer service representative speaks, repetition rate of customer service replies, and customer service representative impatience level; and the third statistical feature parameters include at least one of the following: customer service representative's start date, customer service representative's gender, and customer service representative's age.

[0111] Optionally, if the service data to be tested includes service voice, corresponding text data, historical service data of customer service representatives, and basic employee information of customer service representatives, then obtaining the first feature vector corresponding to the service data to be tested can be understood as: determining audio feature parameters based on the service voice, determining a first statistical feature parameter based on the text data corresponding to the service voice, determining a second statistical feature parameter based on the historical service data, and determining a third statistical feature parameter based on the basic employee information of customer service representatives in the service data to be tested. Then, the first feature vector is determined based on the audio feature parameter, the first statistical feature parameter, the second statistical feature parameter, and the third statistical feature parameter. Optionally, the service data to be tested can also be other combinations of data such as service voice, corresponding text data, historical service data of customer service representatives, and basic employee information of customer service representatives, which will not be listed here.

[0112] S202. Obtain the service data of the customer service's historical qualified services, and determine the service difference value between the service to be inspected and the service data of the historical qualified services based on the service data of the service to be inspected and the service data of the historical qualified services.

[0113] In some feasible implementations, after obtaining the service data to be tested (i.e., the service to be tested) from the customer service representative, the service data of the customer service representative's historical qualified services can be further obtained. The service difference value between the service to be tested and the historical qualified services can be determined based on the service data to be tested and the historical qualified services. Determining the service difference value between the service to be tested and the historical qualified services can be understood as follows: obtaining a first feature vector corresponding to the service data to be tested, and obtaining a second feature vector corresponding to the service data of the historical qualified services. Encoding the first feature vector based on the second feature vector to obtain a third feature vector corresponding to the first feature vector. Then, determining the service difference value between the service and the historical qualified services based on the first and third feature vectors. The above-mentioned encoding the first feature vector based on the second feature vector to obtain the third feature vector corresponding to the first feature vector can be understood as follows: converting the historical qualified services data of the customer service representative into a second feature vector as a training sample, and training the autoencoder (AE) using unsupervised learning, so that the autoencoder trained based on the above training samples learns the patterns of the customer service representative's historical qualified services. Therefore, when predicting the service data to be inspected, the first feature vector corresponding to the service data is input into a pre-trained autoencoder, and a third feature vector corresponding to the first feature vector can be obtained based on the autoencoder. Finally, the service difference value between the service and the historical qualified services provided by the customer service provider is determined based on the first and third feature vectors.

[0114] The process of determining the service difference between the current service and the historical qualified services provided by the customer service provider, based on the first and third feature vectors, can be understood as follows: calculating the vector similarity value between the first and third feature vectors, and determining this vector similarity value as the service difference between the current service and the customer service provider's historical qualified services. In other words, by comparing the similarity between the input layer vector (i.e., the first feature vector) and the output layer vector (i.e., the third feature vector) of the autoencoder, the similarity value between the first and third feature vectors can be obtained. Furthermore, the similarity value between the first and third feature vectors is determined as the service difference between the service to be detected and the customer service provider's historical qualified services. Generally, a larger service difference value indicates a lower service quality, while a smaller service difference value indicates a higher service quality. Understandably, the method used to determine the similarity value between the first and third feature vectors can be any of the following: cosine similarity, Minkowski distance, Euclidean distance, Pearson correlation coefficient, etc., depending on the specific application scenario, and is not limited here.

[0115] Please see Figure 4 , Figure 4 This is a schematic diagram of an autoencoder. An autoencoder is a neural network where the input and learning target are the same, and its structure consists of two parts: an encoder and a decoder. Figure 4 As shown, an autoencoder includes an encoder and a decoder. The encoder maps an input sample x to a feature space, obtaining abstract features z. The decoder then maps these abstract features z back to the original space, obtaining a reconstructed sample x'. The autoencoder optimizes without using sample labels, significantly improving the model's versatility. In other words, through unsupervised learning, the trained autoencoder can perform feature transformation on the input data, encoding it into another form, and then perform a series of learning steps based on this. In this embodiment, by inputting a first feature vector into the autoencoder trained based on a second feature vector, a third feature vector corresponding to the first feature vector output by the autoencoder can be obtained. Therefore, by comparing the similarity between the input layer vector (i.e., the first feature vector) and the output layer vector (i.e., the third feature vector), a similarity value between the first and third feature vectors can be obtained, which serves as the service difference value between the service and historically qualified services.

[0116] S203. Determine whether the service provided by customer service is qualified based on the service quality evaluation value and service difference value.

[0117] In some feasible implementations, after determining the service quality evaluation value of the service to be tested based on the service data to be tested, and determining the service difference value between the service and the customer service's historical qualified services based on the service data to be tested, it can be further determined whether the service provided by the customer service is a qualified service based on the service quality evaluation value and the service difference value.

[0118] Specifically, a first weight and a second weight can be obtained first, where the sum of the first weight and the second weight equals 1. Then, the sum of the product of the first weight and the service quality evaluation value, and the product of the second weight and the service difference value, is determined as the service quality assessment value. If the service quality assessment value is less than the service detection threshold, the service is determined to be a qualified service; if the service quality assessment value is not less than (i.e., greater than or equal to) the service detection threshold, the service is determined to be a unqualified service. Alternatively, if the service quality assessment value is not greater than (i.e., less than or equal to) the service detection threshold, the service is determined to be a qualified service; if the service quality assessment value is greater than the service detection threshold, the service is determined to be a unqualified service. The specific determination depends on the actual application scenario and is not limited here.

[0119] For example, suppose customer service representative A provides a service quality rating of 0.55 for a service server1 provided to user B, with a service difference score of 0.7. The first weight is 0.4, and the second weight is 0.6 (meaning the sum of the first and second weights equals 1). Then, the service quality assessment value C for this service provided by customer service representative A to user B is C = 0.55 × 0.4 + 0.7 × 0.6 = 0.64. Assuming the service detection threshold C0 = 0.5, since the service quality assessment value C = 0.64 is greater than the service detection threshold C0 = 0.5, the service server1 can be determined to be an unqualified service.

[0120] Optionally, in some feasible implementations, if a customer service representative's service is detected as substandard, a service behavior inspection report can be generated and sent to the representative via SMS / email. If the representative has doubts about the report, they can appeal to their superior or quality inspector via their user device. The quality inspector will then manually review the appeal and save the results. Finally, the successfully appealed service is used as a training sample for model optimization in service quality prediction models, autoencoders, etc.

[0121] In this embodiment, the service voice recordings provided by customer service representatives are acquired, and the corresponding service data to be inspected is determined based on these recordings. A service quality evaluation value is determined based on the service data to be inspected, and a service difference value between the service and the customer service representative's historical qualified services is also determined based on the service data. Finally, the service quality evaluation value and the service difference value are used to determine whether the service provided by the customer service representative is a qualified service. Using this embodiment improves the accuracy and efficiency of service quality detection and has high applicability.

[0122] For example, please see Figure 5 , Figure 5 This is another flowchart illustrating the service quality detection method provided in this application embodiment. For example... Figure 5 As shown, firstly, the service data to be inspected can be constructed based on data such as the service voice recording of a specific service provided by customer service, the text data obtained from the service voice recording through speech recognition, the customer service's historical service data, and the customer service's basic employee information. This service data is then processed through the aforementioned... Figure 2 After processing through the steps described herein, a service quality inspection result (whether the service is qualified) can be obtained. The specific implementation process of the service quality inspection can be found in [link to relevant documentation]. Figure 2The descriptions of steps S201 to S203 in the illustrated embodiment will not be repeated here. Understandably, if the service inspection result indicates that the service is unqualified, the service inspection result can be sent to the corresponding customer service representative. When the customer service representative receives the service inspection result, if they disagree with it, they can appeal to the quality inspector or their immediate supervisor. The quality inspector will then conduct a service quality review. Finally, the services for which the appeal was successfully filed will be used as training samples for model optimization in service quality prediction models, autoencoders, etc.

[0123] Understandably, in this embodiment of the application, the service data to be inspected for service quality detection includes not only service voice and text data corresponding to the service voice, but also historical service data of customer service and basic employee information of customer service, making the feature parameters for service quality evaluation more comprehensive and improving the accuracy of the model.

[0124] Please see Figure 6 , Figure 6 This is a schematic diagram of the service quality detection device provided in an embodiment of this application. The service quality detection device provided in an embodiment of this application includes:

[0125] The service quality evaluation value determination module 31 is used to obtain the service data to be inspected provided by the customer service to the customer, and to determine the service quality evaluation value of the service based on the service data to be inspected.

[0126] The service difference value determination module 32 is used to obtain the service data of the customer service's historical qualified services, and determine the service difference value between the service to be inspected and the historical qualified services of the customer service based on the service data to be inspected and the service data of the historical qualified services.

[0127] The service determination module 33 is used to determine whether the service provided by the customer service is a qualified service based on the service quality evaluation value and the service difference value.

[0128] Please see Figure 7 , Figure 7 This is another structural schematic diagram of the service quality detection device provided in the embodiments of this application. Wherein:

[0129] In some possible implementations, the service quality evaluation value determination module 31 includes:

[0130] The first feature vector acquisition unit 311 is used to acquire the first feature vector corresponding to the above-mentioned service data to be inspected.

[0131] The service quality evaluation value determination unit 312 is used to input the first feature vector into the service quality prediction model and obtain the service quality evaluation value of the service corresponding to the first feature vector through the service quality prediction model. The service quality prediction model is trained by service data corresponding to the historical services of at least one customer service representative and preset service quality evaluation values. The historical services include at least one historical qualified service and at least one historical unqualified service.

[0132] In some possible implementations, the service difference value determination module 32 described above includes:

[0133] The feature vector acquisition unit 321 is used to acquire the first feature vector corresponding to the service data to be inspected and to acquire the second feature vector corresponding to the service data of the historical qualified services.

[0134] The encoding processing unit 322 is used to encode the first feature vector based on the second feature vector to obtain the third feature vector corresponding to the first feature vector.

[0135] The service difference value determination unit 323 is used to determine the service difference value between the service and the historical qualified service based on the first feature vector and the third feature vector.

[0136] In some possible implementations, the service difference value determination unit 323 described above is specifically used for:

[0137] Calculate the vector similarity value between the first feature vector and the third feature vector, and determine the vector similarity value as the service difference value between the service and the historical qualified service.

[0138] In some possible implementations, the service determination module 33 includes:

[0139] The weight acquisition unit 331 is used to acquire a first weight and a second weight. The first weight is used to mark the weight of the service quality evaluation value, and the second weight is used to mark the weight of the service difference value. The sum of the first weight and the second weight is equal to 1.

[0140] The weighted calculation unit 332 is used to determine a first weighted service quality evaluation value based on the first weight and the service quality evaluation value, and to determine a second weighted service quality evaluation value based on the second weight and the service difference value.

[0141] The evaluation value determination unit 333 is used to determine the service quality evaluation value based on the first weighted service quality evaluation value and the second weighted service quality evaluation value.

[0142] The service determination unit 334 is used to determine that the service is a qualified service if the service quality assessment value is less than the service detection threshold, and to determine that the service is an unqualified service if the service quality assessment value is not less than the service detection threshold.

[0143] In some possible implementations, the service quality evaluation value determination module 31 or the service difference value determination module 32 described above is further used for:

[0144] The service voice recordings provided by the customer service representatives are obtained, and speech recognition is performed on these service voice recordings to obtain their text data.

[0145] Based on the aforementioned service voice and text data, generate the service data to be inspected corresponding to the aforementioned service.

[0146] In some possible implementations, the service quality evaluation value determination module 31 or the service difference value determination module 32 described above is further used for:

[0147] The above-mentioned customer service voice recordings are obtained, and speech recognition is performed on the above-mentioned voice recordings to obtain the corresponding text data.

[0148] Obtain the historical service data of the aforementioned customer service representatives and / or the basic employee information of the aforementioned customer service representatives, and generate the service data to be inspected for the aforementioned services based on the aforementioned service voice, the aforementioned text data, the aforementioned historical service data of the aforementioned customer service representatives and / or the aforementioned basic employee information of the aforementioned customer service representatives.

[0149] In some possible implementations, the service data to be inspected includes the customer service voice provided by the customer service representative; the first feature vector acquisition unit 311 or the feature vector acquisition unit 321 is specifically used for:

[0150] A first preset sampling frequency is obtained, and the service voice is sampled based on the first preset sampling frequency to obtain a voice sampling signal. The voice sampling signal is then subjected to frame-by-frame windowing processing to obtain at least one frame signal that constitutes the voice sampling signal.

[0151] Short-time energy and / or short-time zero-crossing rate are extracted from each of the at least one framed signal to serve as audio feature parameters, and a first feature vector is determined based on the audio feature parameters.

[0152] In some possible implementations, the first feature vector acquisition unit 311 or the feature vector acquisition unit 321 described above is specifically used for:

[0153] Determine the audio feature parameters of the service voice in the service data to be inspected, including short-time energy and / or short-time zero-crossing rate;

[0154] Based on the above text data, a first statistical feature parameter of the above service is determined. The first statistical feature parameter includes at least one of the following: the service type of the above service, the number of times the customer service representative speaks, the repetition rate of the customer service response, and the degree of customer service representative impatience.

[0155] The first feature vector corresponding to the service data to be inspected is determined based on the above-mentioned audio feature parameters and the above-mentioned first statistical feature parameters.

[0156] In some possible implementations, the first feature vector acquisition unit 311 or the feature vector acquisition unit 321 described above is specifically used for:

[0157] Determine the audio feature parameters of the service voice in the service data to be inspected, including short-time energy and / or short-time zero-crossing rate;

[0158] The first statistical feature parameter is determined based on the text data in the service data to be inspected. The first statistical feature parameter includes at least one of the service type of the service, the number of times the customer service representative speaks, the repetition rate of the customer service reply, and the customer service representative's impatience.

[0159] A second statistical characteristic parameter is determined based on the historical service data in the aforementioned service data to be inspected. This second statistical characteristic parameter includes the number of repeat calls from the aforementioned customer and / or the cumulative number of customer service errors; and / or

[0160] The third statistical feature parameter is determined based on the basic employee information of the customer service staff in the above-mentioned service data to be inspected. The third statistical feature parameter includes at least one of the customer service staff’s employment time, customer service staff’s gender, and customer service staff’s age.

[0161] The first feature vector is determined based on the aforementioned audio feature parameters, the aforementioned first statistical feature parameters, the aforementioned second statistical feature parameters, and / or the aforementioned third statistical feature parameters.

[0162] In this embodiment, the service quality detection device can acquire the customer service voice recordings provided by customer service representatives, determine the corresponding service data to be inspected based on the voice recordings, determine the service quality evaluation value of the service based on the service data to be inspected, and determine the service difference value between the service and the customer service representative's historical qualified services based on the service data to be inspected. Finally, it determines whether the service provided by the customer service representative is a qualified service based on the service quality evaluation value and the service difference value. Using this embodiment can improve the accuracy and efficiency of service quality detection, and has high applicability.

[0163] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of the terminal device provided in the embodiments of this application. For example... Figure 8As shown, the terminal device in this embodiment may include one or more processors 401, a memory 402, and a transceiver 403. The processors 401, memory 402, and transceiver 403 are connected via a bus 404. The memory 402 stores a computer program, which includes program instructions. The processors 401 and transceiver 403 execute the program instructions stored in the memory 402, performing the following operations:

[0164] Transceiver 403 is used to acquire pending service data provided by customer service to customers;

[0165] Processor 401 is used to determine the service quality evaluation value of the above-mentioned service based on the above-mentioned service data to be inspected.

[0166] Transceiver 403 is used to obtain service data of the above-mentioned customer service's historical qualified service.

[0167] Processor 401 is used to determine the service difference value between the above-mentioned service and the above-mentioned customer service's historical qualified service based on the above-mentioned service data to be inspected and the service data of the above-mentioned historical qualified service.

[0168] Processor 401 is used to determine whether the service provided by the customer service representative is a qualified service based on the service quality evaluation value and the service difference value.

[0169] In some feasible implementations, the processor 401 is specifically used for:

[0170] Obtain the first feature vector corresponding to the above-mentioned service data to be inspected;

[0171] The first feature vector is input into the service quality prediction model, and the service quality evaluation value of the service corresponding to the first feature vector is obtained through the service quality prediction model. The service quality prediction model is trained by the service data corresponding to the historical services of at least one customer service representative and the preset service quality evaluation value. The historical services include at least one historical qualified service and at least one historical unqualified service.

[0172] In some feasible implementations, the processor 401 is specifically used for:

[0173] Obtain the first feature vector corresponding to the above-mentioned service data to be inspected, and obtain the second feature vector corresponding to the service data of the above-mentioned historical qualified services;

[0174] The first feature vector is encoded based on the second feature vector to obtain the third feature vector corresponding to the first feature vector.

[0175] The service difference value between the above service and the above historical qualified service is determined based on the first feature vector and the third feature vector.

[0176] In some feasible implementations, the processor 401 is specifically used for:

[0177] Calculate the vector similarity value between the first feature vector and the third feature vector, and determine the vector similarity value as the service difference value between the service and the historical qualified service.

[0178] In some feasible implementations, the processor 401 is specifically used for:

[0179] Obtain a first weight and a second weight. The first weight is used to mark the weight of the service quality evaluation value, and the second weight is used to mark the weight of the service difference value. The sum of the first weight and the second weight is equal to 1.

[0180] A first weighted service quality evaluation value is determined based on the first weight and the service quality evaluation value, and a second weighted service quality evaluation value is determined based on the second weight and the service difference value.

[0181] The service quality assessment value is determined based on the first weighted service quality evaluation value and the second weighted service quality evaluation value mentioned above.

[0182] If the above service quality assessment value is less than the service testing threshold, the above service is determined to be a qualified service; if the above service quality assessment value is not less than the above service testing threshold, the above service is determined to be a unqualified service.

[0183] In some feasible implementations, the processor 401 is specifically used for:

[0184] The service voice recordings provided by the customer service representatives are obtained, and speech recognition is performed on these service voice recordings to obtain their text data.

[0185] Based on the aforementioned service voice and text data, generate the service data to be inspected corresponding to the aforementioned service.

[0186] In some feasible implementations, the processor 401 is specifically used for:

[0187] The above-mentioned customer service voice recordings are obtained, and speech recognition is performed on the above-mentioned voice recordings to obtain the corresponding text data.

[0188] Obtain the historical service data of the aforementioned customer service representatives and / or the basic employee information of the aforementioned customer service representatives, and generate the service data to be inspected for the aforementioned services based on the aforementioned service voice, the aforementioned text data, the aforementioned historical service data of the aforementioned customer service representatives and / or the aforementioned basic employee information of the aforementioned customer service representatives.

[0189] In some feasible implementations, the service data to be inspected includes the customer service voice provided by the customer service representative; the processor 401 is specifically used for:

[0190] A first preset sampling frequency is obtained, and the service voice is sampled based on the first preset sampling frequency to obtain a voice sampling signal. The voice sampling signal is then subjected to frame-by-frame windowing processing to obtain at least one frame signal that constitutes the voice sampling signal.

[0191] Short-time energy and / or short-time zero-crossing rate are extracted from each of the at least one framed signal to serve as audio feature parameters, and a first feature vector is determined based on the audio feature parameters.

[0192] In some feasible implementations, the processor 401 is specifically used for:

[0193] Determine the audio feature parameters of the service voice in the service data to be inspected, including short-time energy and / or short-time zero-crossing rate;

[0194] Based on the above text data, a first statistical feature parameter of the above service is determined. The first statistical feature parameter includes at least one of the following: the service type of the above service, the number of times the customer service representative speaks, the repetition rate of the customer service response, and the degree of customer service representative impatience.

[0195] The first feature vector corresponding to the service data to be inspected is determined based on the above-mentioned audio feature parameters and the above-mentioned first statistical feature parameters.

[0196] In some feasible implementations, the processor 401 is specifically used for:

[0197] Determine the audio feature parameters of the service voice in the service data to be inspected, including short-time energy and / or short-time zero-crossing rate;

[0198] The first statistical feature parameter is determined based on the text data in the service data to be inspected. The first statistical feature parameter includes at least one of the service type of the service, the number of times the customer service representative speaks, the repetition rate of the customer service reply, and the customer service representative's impatience.

[0199] A second statistical characteristic parameter is determined based on the historical service data in the aforementioned service data to be inspected. This second statistical characteristic parameter includes the number of repeat calls from the aforementioned customer and / or the cumulative number of customer service errors; and / or

[0200] The third statistical feature parameter is determined based on the basic employee information of the customer service staff in the above-mentioned service data to be inspected. The third statistical feature parameter includes at least one of the customer service staff’s employment time, customer service staff’s gender, and customer service staff’s age.

[0201] The first feature vector is determined based on the aforementioned audio feature parameters, the aforementioned first statistical feature parameters, the aforementioned second statistical feature parameters, and / or the aforementioned third statistical feature parameters.

[0202] It should be understood that in some feasible implementations, the processor 401 described above may be a central processing unit (CPU), which may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The memory 402 may include read-only memory and random access memory, and provides instructions and data to the processor 401. A portion of the memory 402 may also include non-volatile random access memory. For example, the memory 402 may also store device type information.

[0203] In specific implementation, the aforementioned terminal device can perform the above-described actions through its built-in functional modules. Figure 2 The implementation methods provided for each step are detailed in the above-mentioned implementation methods, and will not be repeated here.

[0204] In this embodiment, the terminal device can acquire the service voice recordings provided by customer service representatives, determine the corresponding service data to be inspected based on the service voice recordings, determine the service quality evaluation value of the service based on the service data to be inspected, and determine the service difference value between the service and the customer service representative's historical qualified services based on the service data to be inspected. Finally, it determines whether the service provided by the customer service representative is a qualified service based on the service quality evaluation value and the service difference value. Using this embodiment can improve the accuracy and efficiency of service quality detection, and has high applicability.

[0205] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which are implemented when executed by a processor. Figure 2 The service quality testing methods provided in each step are detailed in the implementation methods provided in the above steps, and will not be repeated here.

[0206] The aforementioned computer-readable storage medium can be the internal storage unit of the service quality detection device provided in any of the foregoing embodiments or the terminal device, such as a hard disk or memory of an electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store the computer program and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0207] The terms "first," "second," "third," "fourth," etc., in the claims, description, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0208] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The presentation of this phrase in various locations throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments. The term "and / or" as used in this specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0209] The methods and related apparatuses provided in this application are described with reference to the method flowcharts and / or structural diagrams provided in this application. Specifically, each block of the method flowchart and / or structural diagram, as well as combinations of blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to create a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, generate instructions for implementing the process. Figure 1 A schematic diagram of one or more processes and / or structures. Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 A schematic diagram of one or more processes and / or structures. Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 A process or multiple processes and / or structures illustrate the steps of the functions specified in one or more boxes.

Claims

1. A service quality testing method, characterized in that, The method includes: Obtain pending service data of customer service representatives providing services to customers, and determine the service quality evaluation value of the service based on the pending service data. The pending service data includes the service voice of the customer service representatives providing services to customers, or the pending service data includes the service voice of the customer service representatives providing services to customers and the text data of the service voice. Obtain the service data of the customer service representative’s historical qualified services, obtain the first feature vector corresponding to the service data to be inspected, and obtain the second feature vector corresponding to the service data of the historical qualified services. The first feature vector is encoded based on the second feature vector to obtain the third feature vector corresponding to the first feature vector. The service difference value between the service and the historical qualified service is determined based on the first feature vector and the third feature vector. The quality of service provided by the customer service representative is determined to be qualified based on the service quality evaluation value and the service difference value.

2. The method according to claim 1, characterized in that, Determining the service difference value between the service and the historical qualified service based on the first feature vector and the third feature vector includes: Calculate the vector similarity value between the first feature vector and the third feature vector, and determine the vector similarity value as the service difference value between the service and the historical qualified service.

3. The method according to any one of claims 1-2, characterized in that, The step of determining whether the service provided by the customer service representative is a qualified service based on the service quality evaluation value and the service difference value includes: Obtain a first weight and a second weight. The first weight is used to mark the weight of the service quality evaluation value, and the second weight is used to mark the weight of the service difference value. The sum of the first weight and the second weight is equal to 1. A first weighted service quality evaluation value is determined based on the first weight and the service quality evaluation value, and a second weighted service quality evaluation value is determined based on the second weight and the service difference value. The service quality assessment value is determined based on the first weighted service quality evaluation value and the second weighted service quality evaluation value; If the service quality assessment value is less than the service detection threshold, the service is determined to be a qualified service; if the service quality assessment value is not less than the service detection threshold, the service is determined to be a unqualified service.

4. The method according to claim 3, characterized in that, The service data to be inspected includes the customer service voice message and the text data of the voice message; obtaining the service data to be inspected from the customer service representative includes: The service voice provided by the customer service representative is obtained, and speech recognition is performed on the service voice to obtain the text data of the service voice. Generate the service data to be inspected corresponding to the service based on the service voice and the text data.

5. The method according to claim 1, characterized in that, The service data to be inspected includes the customer service voice messages provided by the customer service representative; obtaining the first feature vector corresponding to the service data to be inspected includes: A first preset sampling frequency is obtained, the service voice is sampled based on the first preset sampling frequency to obtain a voice sampling signal, and the voice sampling signal is subjected to frame-by-frame windowing processing to obtain at least one frame signal that constitutes the voice sampling signal. Short-time energy and / or short-time zero-crossing rate are extracted from each frame signal in the at least one frame signal as audio feature parameters, and a first feature vector is determined based on the audio feature parameters.

6. The method according to claim 4, characterized in that, The step of obtaining the first feature vector corresponding to the service data to be inspected includes: Determine the audio feature parameters of the service speech in the service data to be inspected, the audio feature parameters including short-time energy and / or short-time zero-crossing rate; The first statistical feature parameter of the service is determined based on the text data. The first statistical feature parameter includes at least one of the service type of the service, the number of times the customer service representative speaks, the repetition rate of the customer service response, and the customer service representative's impatience level. The first feature vector corresponding to the service data to be inspected is determined based on the audio feature parameters and the first statistical feature parameters.

7. A service quality testing device, characterized in that, The device includes: The service quality evaluation value determination module is used to acquire the service data to be inspected provided by the customer service representative, and to determine the service quality evaluation value of the service based on the service data to be inspected. The service data to be inspected includes the service voice provided by the customer service representative, or the service data to be inspected includes the service voice provided by the customer service representative and the text data of the service voice. The service difference value determination module is used to obtain service data of the customer service's historical qualified services, obtain a first feature vector corresponding to the service data to be inspected, and obtain a second feature vector corresponding to the service data of the historical qualified services; encode the first feature vector based on the second feature vector to obtain a third feature vector corresponding to the first feature vector, and determine the service difference value between the service and the historical qualified service based on the first feature vector and the third feature vector. The service determination module is used to determine whether the service provided by the customer service is a qualified service based on the service quality evaluation value and the service difference value.

8. A terminal device, characterized in that, It includes a processor, a memory, and a transceiver, wherein the processor, the memory, and the transceiver are interconnected. The memory is used to store a computer program, the computer program including program instructions, and the processor and the transceiver are configured to invoke the program instructions to perform the method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method as described in any one of claims 1-6.

Images