Customer service call quality control method and system based on speech recognition

Abstract

Multiple embodiments of this specification relate to the field of information technology, and specifically to a method and system for customer service call quality control based on speech recognition. The method includes the steps of: identifying the voice as a customer voice and a customer service voice; identifying the topic of the call; obtaining a number of optional replies and calculating the relevance respectively; intercepting the customer service voice when the relevance is lower than a preset threshold; identifying the voice emotion of the customer service voice, comparing the voice emotion with a pre-configured screening emotion set, and comparing the customer service text with a pre-configured screening vocabulary; when the voice emotion does not belong to the screening emotion set and the customer service text does not include words in the screening vocabulary, identifying the topic of the customer service text; when the relevance between the customer service text and the call topic is higher than a preset threshold, or the relevance between the topic of the customer service text and the call topic is higher than a preset threshold, releasing the customer service voice.

Description

Technical Field

[0001] Multiple embodiments of this specification relate to the field of information technology, and specifically to a method and system for customer service call quality control based on speech recognition. Background Art

[0002] Customer service calls are used across a wide range of industries. Almost every industry that interacts directly with customers involves customer service via calls. This is particularly true in sectors that rely heavily on customer service to maintain customer satisfaction and loyalty, such as telecommunications, e-commerce, and financial services. While businesses can enhance their market competitiveness by providing professional and efficient customer support, they also need to monitor and supervise the quality of customer service calls. High-quality call service not only builds a positive image and promotes employee development, but also reduces operating costs and mitigates legal risks. However, due to the immediate nature of call services and the difficulty of rapid verification, quality management presents challenges. Currently, users often rate customer service calls afterward, but this approach is subject to time lags. Therefore, research is needed to improve customer service call quality control technologies. Summary of the Invention

[0003] Multiple embodiments of this specification describe a method and system for customer service call quality control based on speech recognition.

[0004] In a first aspect, an embodiment of this specification provides a method for controlling customer service call quality based on speech recognition, comprising the steps of:

[0005] Read the voice of customer service calls in real time and identify the voice as customer voice or customer service voice based on the timbre of the voice;

[0006] Recognize the text of the customer voice and the customer service voice, record them as customer text and customer service text respectively, and identify the conversation topic based on the customer text and customer service text;

[0007] According to the customer text and a pre-configured answer library, a number of optional answers are obtained, and the relevance between the customer service text and the plurality of optional answers is calculated respectively;

[0008] When the relevance between the customer service text and several of the optional replies is lower than a preset threshold, intercepting the customer service voice;

[0009] Identify the voice emotion of the customer service voice, compare the voice emotion with a pre-configured screening emotion set, and compare the customer service text with a pre-configured screening vocabulary;

[0010] When the voice emotion does not belong to the screening emotion set and the customer service text does not include words in the screening vocabulary, identifying the topic of the customer service text;

[0011] When the relevance between the customer service text and the conversation topic is higher than a preset threshold, or the relevance between the topic of the customer service text and the conversation topic is higher than a preset threshold, the customer service voice is released.

[0012] In a second aspect, the embodiments of this specification provide a customer service call quality control system based on speech recognition, including:

[0013] The reading module reads the voice of the customer service call in real time and identifies the voice as the customer's voice or the customer service voice based on the timbre of the voice;

[0014] a recognition module for recognizing the text of the customer voice and the customer service voice, recording them as customer text and customer service text respectively, and identifying the conversation topic based on the customer text and customer service text;

[0015] A reply module, which obtains a plurality of optional replies based on the customer text and a pre-configured reply library, and calculates the relevance between the customer service text and the plurality of optional replies;

[0016] an interception module, which intercepts the customer service voice when the relevance between the customer service text and the plurality of optional replies is lower than a preset threshold;

[0017] A comparison module, which identifies the voice emotion of the customer service voice, compares the voice emotion with a pre-configured screening emotion set, and compares the customer service text with a pre-configured screening vocabulary;

[0018] a topic module, which identifies the topic of the customer service text when the voice emotion does not belong to the screening emotion set and the customer service text does not include words in the screening vocabulary;

[0019] The management and control module releases the customer service voice when the relevance between the customer service text and the conversation topic is higher than a preset threshold, or the relevance between the topic of the customer service text and the conversation topic is higher than a preset threshold.

[0020] In a third aspect, embodiments of this specification provide an electronic device, including a processor and a memory;

[0021] The processor is connected to the memory;

[0022] The memory is used to store executable program code;

[0023] The processor reads the executable program code stored in the memory to run a program corresponding to the executable program code, so as to execute the method described in any one of the above aspects.

[0024] In a fourth aspect, an embodiment of this specification provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method described in any one of the above aspects is implemented.

[0025] In a fifth aspect, embodiments of this specification provide a computer program product, including a computer program, which implements the method described in any of the above aspects when executed by a processor.

[0026] The beneficial effects of the technical solutions provided by some embodiments of this specification include at least:

[0027] In multiple embodiments of this specification, the provided customer service call quality control method identifies voice text, distinguishes between customer text and customer service text, and performs a preliminary check on the customer service voice in combination with a response library. When the customer service voice exceeds the response library, the corresponding steps are initiated to check the call voice and intercept customer service voice that does not meet the corresponding conditions. This helps to improve the quality of customer service calls, reduce or avoid non-standard customer service call content, and ultimately improve service quality. By identifying the similarity and correlation between keywords in customer text and customer service text and topics in the topic library, the call topic is made more consistent with the context, helping to more reasonably control the call quality.

[0028] Other features and advantages of the various embodiments of this specification will be further disclosed in the following detailed description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of this specification, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0030] Figure 1 Schematic diagram of an application scenario of the customer service call quality control method provided in the embodiments of this specification.

[0031] Figure 2 This is a schematic diagram of the customer service call quality control method architecture provided in the embodiments of this specification.

[0032] Figure 3 This is a schematic diagram of the operation interface provided in the embodiments of this specification.

[0033] Figure 4 Schematic diagram of the customer service call quality control method provided in the embodiments of this specification.

[0034] Figure 5 This is a schematic diagram of a method for obtaining a comparison correlation degree provided in an embodiment of this specification.

[0035] Figure 6 A schematic diagram of a method for obtaining a response library provided in an embodiment of this specification.

[0036] Figure 7 Schematic diagram of the customer service call quality control system provided in the embodiments of this specification.

[0037] Figure 8 This is a schematic diagram of an electronic device provided in an embodiment of this specification. DETAILED DESCRIPTION

[0038] The following is an explanation and description of the technical solutions of the embodiments of this specification in conjunction with the drawings of the embodiments of this specification. However, the following embodiments are only preferred embodiments of this specification and are not exhaustive. Based on the embodiments in the implementation mode, other embodiments obtained by those skilled in the art without making any creative work are all within the scope of protection of this specification.

[0039] Throughout this specification, the claims, and the accompanying drawings, the terms "first," "second," "third," and the like are used to distinguish between different items, not to describe a particular order. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may include other steps or elements inherent to the process, method, product, or apparatus.

[0040] In the following description, terms such as "inside", "outside", "up", "down", "left", "right", etc. that indicate directions or positional relationships are only used to facilitate the description of the embodiments and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, they should not be understood as limitations on this specification.

[0041] The data involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection of relevant data complies with the relevant laws, regulations and standards of relevant countries and regions.

[0042] Before introducing the technical solution in this specification, the application scenarios and related technologies of the technical solution are introduced.

[0043] Call services refer to real-time voice communication and support provided between businesses and customers via the phone or other voice communication tools. This service is typically provided by a dedicated customer service team, aiming to answer customer questions, handle complaints, and provide technical support or guidance. Call services are the most direct and immediate form of customer service, enabling rapid responses to customer needs and resolving practical issues. The call service process includes answering the call, greeting and verification, needs analysis, problem resolution, call termination, recording, and feedback. Call services can directly answer questions, provide operational guidance, handle complaints, and record and forward issues to the relevant departments. They can soothe customers and resolve urgent or complex issues. Consequently, businesses across many industries have established call services of varying sizes. For example, call services can answer customer questions regarding product features, usage, and pricing. They can also provide technical support for equipment installation, software usage, and troubleshooting. They can also handle order inquiries, modifications, and cancellations. They can also resolve customer complaints and offer reasonable compensation or solutions. They can also assist customers with account registration, login, and password resets. And provide immediate help and support when customers encounter urgent problems.

[0044] Providing services through phone calls is characterized by immediacy, personalization, strong interactivity, emotional connection, and a wide range of applicability. Specifically, phone service can immediately respond to the needs of customers 10 and provide real-time support and assistance, which is faster than other methods such as email and online chat. Through direct dialogue, customer service 33 can better understand customer needs and provide personalized services based on non-verbal information such as customer 10's tone and emotions. Voice communication allows both parties to communicate in a two-way manner, which can more effectively clarify problems, confirm solutions, and improve the efficiency of problem solving. During the call, customer service 33 can convey care and warmth through tone and intonation, enhance the emotional connection with customers 10, and improve customer satisfaction. Phone service is suitable for problems of various complexities. Whether it is a simple consultation or complex troubleshooting, it can be effectively resolved over the phone.

[0045] As a direct and efficient way of customer service, call service plays an important role in many industries. Through professional customer service and standardized service processes, enterprises can better meet customer needs, improve customer satisfaction and enhance brand loyalty. Therefore, the quality control of call service is of great significance. At present, feedback on the quality of call service is often obtained through the customer's post-rating method. It not only has the disadvantage of lag, but also cannot accurately know which call content causes customer dissatisfaction. To this end, this manual provides a customer service call quality control method based on voice recognition, please refer to the attached Figure 1The customer service call quality control method provided in this specification runs on a server 31 and is connected to a customer service call system and a terminal device 34 for background management.

[0046] Customer service representative 33 connects to call service software 32 on a computer via a headset. Call service software 32 establishes data exchange with communication device 20, which is typically provided by a telecommunications provider. Customer 10 connects to communication device 20 via telephone to establish a voice call with customer service representative 33. Call service software 32 running on the computer is also connected to a server 31, exchanging data with server 31 and submitting relevant voice, data, and status information to server 31. Server 31 has the function of providing feedback and sending control instructions to call service software 32. Administrator 35 connects to server 31 via terminal device 34 to view relevant logs. Furthermore, a reporting system can be run on server 31, allowing administrator 35 to directly view reports and understand the quality of customer service representative 33's calls.

[0047] The method provided in this application is applied to Figure 2 The system architecture shown, Figure 2 This is a schematic diagram of the system architecture in the embodiment of the present application. Figure 2 As shown, the system architecture includes a server 31 and a terminal device 34, and the client (i.e., the interactive interface for viewing relevant quality control data) is deployed on the terminal device 34. The client can be run on the terminal device 34 in the form of a browser or in the form of an independent application (APP). The specific display form of the client is not limited here. Please refer to the attached Figure 3 The client or operation interface can be run on the terminal device 34 in the form of a browser or an independent application (APP). The specific display format of the client is not limited here. Relevant reports can be viewed on the operation interface. The production of reports can be achieved by using the technology disclosed in the art.

[0048] The manager 35 can use a terminal device 34 such as a PC, smartphone, or tablet computer to connect to the server 31. And obtain the log or report of the server 31 about the quality of customer service calls. The terminal device 34 can run a client, or at least run an interactive interface. The server 31 involved in this application can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. The terminal device 34 can be a smartphone, tablet computer, laptop computer, PDA, personal computer, smart speaker, smart TV, smart watch, car-mounted device, wearable device, etc., but is not limited to this. The terminal device 34 and the server 31 can be directly or indirectly connected via wired or wireless communication, and this application does not limit this. The number of servers 31 and terminal devices 34 is also not limited. The solution provided in this application can be completed independently by the terminal device 34, can be completed independently by the server 31, or can be completed in cooperation between the terminal device 34 and the server 31. This application does not make any specific limitations on this.

[0049] Since this application involves some professional terms, these professional terms will be introduced below.

[0050] timbre

[0051] Timbre refers to the characteristics or qualities of a voice, including pitch, timbre, volume, and variations in timbre. Timbre is determined by factors such as vocal cord vibration and oral and nasal cavity resonance. Individual timbre varies, and timbre can be used to quickly identify the speaker. Timbre recognition extracts the frequency domain features of speech and uses these features to distinguish different timbres. Voice volume can also be used to assist in distinguishing timbre. Since only the timbre of the customer and the customer service representative needs to be distinguished, the error tolerance for timbre recognition is high.

[0052] Call topic

[0053] A call topic refers to the specific content or subject discussed during a call. It can be a question, request, or complaint raised by a customer, or information, suggestions, or solutions provided by customer service. Topics can be created by creating natural language tasks based on customer and customer service text, submitting them to a large language model, and obtaining responses from the large language model. Alternatively, multiple call topics can be preset and provided simultaneously to the large language model, requiring it to select the most suitable topic from the provided topics.

[0054] Voice emotion

[0055] Speech emotion refers to the emotional state conveyed through speech, including happiness, anger, sadness, anxiety, and calmness. Speech emotion can be expressed through changes in pitch, speaking rate, volume, and other factors. Speech emotion recognition can be performed using publicly available techniques in the art. For example, this can be achieved using a trained machine learning model, specifically by extracting speech feature parameters, such as pitch, intensity, speaking rate, and pauses, manually annotating speech emotion to generate sample data, and then training the machine learning model with this sample data to obtain a machine learning model capable of recognizing speech emotion.

[0056] Large Language Model

[0057] Large language models are deep learning models with massive parameters designed to understand and generate natural language. These models are trained on vast amounts of text data to learn patterns and structures in language, enabling them to perform a variety of natural language tasks such as machine translation, text summarization, sentiment analysis, and question-answering. Large language models have millions or even billions of parameters, enabling them to capture highly complex patterns in language.

[0058] Natural language tasks

[0059] Natural language tasks refer to natural language processing. Natural language processing (NLP) is an interdisciplinary field in computer science, artificial intelligence, and linguistics that aims to enable computers to understand, interpret, and generate human language. Natural language tasks (NLP tasks) refer to a series of language-related tasks accomplished through algorithms and techniques.

[0060] This manual first provides a customer service call quality control method based on voice recognition. Figure 4 , including the steps of:

[0061] Step S102) Read the voice of the customer service call in real time, and identify the voice as the customer voice and the customer service voice according to the timbre of the voice. Identify the voices of different people by timbre, and then distinguish the customer service voice and the customer voice, and further divide the voice into the customer voice and the customer service voice. Since it is only necessary to distinguish the customer voice and the customer service voice, the sound volume can be used to assist in the distinction of timbre. Specifically, the sound volume is used to preliminarily distinguish the customer voice and the customer service voice, and then the timbre can be used for comparison. Compared with identifying the timbre, the execution efficiency can be improved. Since the customer service is fixed or the frequency of change is low on a specific call service software 32, it is easier to distinguish the customer service voice. The essence of timbre is the difference in the resonant frequency composition in the frequency domain. Therefore, after extracting the frequency domain features of the voice, based on the comparison of the frequency domain composition, it is possible to quickly complete the judgment of whether it is the timbre of the identified customer service. And it is possible to complete the judgment of whether the two voices have the same timbre.

[0062] Step S104) Recognize the text of the customer and customer service voice, recording them as customer text and customer service text, respectively. The conversation topic is identified based on the customer and customer service text. After recognizing the customer and customer service voice, further text recognition is performed based on the voice. Audio-to-text recognition can be performed using techniques already known in the art. Identifying the conversation topic based on the customer and customer service text is performed using a large language model.

[0063] For example, the customer text and customer service text are as follows:

[0064] Customer: Hello, I would like to know more about your air fryer, specifically model A123.

[0065] Customer Service: Hello, thank you for your call. Model A123 air fryer is one of our best-selling products.

[0066] Customer: What is the capacity of this air fryer? How many people in the family can it accommodate?

[0067] Customer Service: The capacity of this air fryer is 5 liters, suitable for families of 2-4 people.

[0068] Customer: How fast does it heat up? Does it have a preheat function?

[0069] Customer Service: It heats up very quickly, and the general preheating time only takes 3-5 minutes. It also has a preheating function, and the preheating time can be set in advance.

[0070] Customer: I also want to know whether it is easy to clean and maintain. Is there anything I need to pay attention to?

[0071] Customer Service: Regarding cleaning and maintenance, the inner pot of this air fryer is non-stick coated, which is very easy to clean. It is recommended to wipe it with a damp cloth after each use and clean the filter regularly.

[0072] Customer: How much does this air fryer cost? Are there any promotions?

[0073] Customer Service: Regarding the price, the model A123 air fryer is priced at 999 yuan. We are currently running a promotion where you can get a premium baking tray and a cookbook with your purchase.

[0074] The conversation topics of customer texts identified through large language models are, in order: "Inquiring about the model A123 air fryer," "Inquiring about specifications," "Inquiring about technical specifications," "Inquiring about usage recommendations," and "Inquiring about prices and promotions." The conversation topics of customer service texts identified through large language models are, in order: "Introducing the product," "Introducing specifications," "Introducing technical specifications," "Introducing usage and maintenance," and "Introducing prices and promotions."

[0075] Step S106) According to the customer text and the pre-configured answer library, a number of optional answers are obtained, and the correlation between the customer service text and the several optional answers is calculated respectively.

[0076] For example, for the customer text "How is its heating speed? Is there a preheating function?", the reply library contains optional reply one "The heating speed is XXX, and the preheating function is XXX", optional reply two "The product introduction page has a detailed introduction, and the content of the product introduction page is a true introduction, you can refer to it", and optional reply three "Please wait, I will check the relevant information and reply to you."

[0077] Step S108) When the correlation between the customer service text and several optional replies is lower than a preset threshold, intercepting subsequent customer service voice.

[0078] If the customer service representative's response exceeds the three aforementioned options, meaning its relevance to all three options is below a preset threshold, the customer service representative's response is deemed to be at risk of irrelevant to the question, and subsequent customer service calls are intercepted. Because the analyzed customer service call has already been delivered to the customer by the communication device 20 and cannot be intercepted in time, subsequent calls are intercepted to prevent potentially problematic responses from continuing.

[0079] Step S110) Identify the voice emotion of the subsequent customer service speech, compare the voice emotion with a pre-configured screening emotion set, identify the text of the subsequent customer service speech as a subsequent customer service text, and compare the subsequent customer service text with a pre-configured screening vocabulary.

[0080] For example, in response to the customer's text "How fast does it heat up? Is there a preheating function?", the customer service text is "Such a simple question, I don't need to answer it". After identification, the voice emotion is obtained as "impatience". For example, the screening emotion set is {"anger", "contempt", "dissatisfaction", "impatience", "disappointment", "confusion", "indifference", "anxiety"}. Therefore, the customer service's reply is "Such a simple question, I don't need to answer it". Because the identified voice emotion is "impatience", the customer service voice will remain intercepted. At this time, it is sufficient to use publicly available technologies in this field or non-technical means to process it. For example, technical means are used to remind customer service to pay attention to standardized replies, or management means are used to switch the customer to other customer service for processing after being prompted.

[0081] As another example, in response to a customer's text message, "How fast does it heat up? Does it have a preheating function?", the customer service text is, "Heating speed isn't the most important technical specification for an air fryer; temperature control stability and uniformity are more important. This model of air fryer uses the latest algorithm and optimized air ducts, resulting in more accurate temperature control and more uniform temperature distribution." After emotion recognition, the voice emotion is determined to be "stable," and no words in the screening vocabulary exist. Therefore, the customer service voice message is released and delivered to the customer via the communication device 20. The customer service voice message will not be intercepted again until the next time the customer service message's relevance to any optional answer falls below a preset threshold.

[0082] Step S112) When the voice emotion does not belong to the screening emotion set and the subsequent customer service text does not include a word in the screening vocabulary, the topic of the subsequent customer service text is identified. When the customer service text includes a word in the screening vocabulary, the corresponding customer service voice will be intercepted.

[0083] Step S114) If the relevance of the subsequent customer service text to the conversation topic exceeds a preset threshold, or if the relevance of the topic of the subsequent customer service text to the conversation topic exceeds a preset threshold, the subsequent customer service voice is released. Even if the customer service voice does not contain any emotions in the screening emotion set or words in the screening vocabulary, the relevance of the customer service text to the conversation topic must still exceed the preset threshold. For example, the customer text "How is its heating speed? Does it have a preheating function?" is related to the conversation topic "Introducing technical specifications." However, the customer service text "Heating speed is not the most important technical specification for an air fryer; temperature control stability and uniformity are more important. This model of air fryer uses the latest algorithm and optimized air ducts, achieving more accurate temperature control and more uniform temperature distribution" is related to "Introducing technical specifications." This means that the relevance exceeds the preset threshold (although it is not a direct response to the heating speed, it is a relevant response). Therefore, the subsequent customer service voice is released. All thresholds in this embodiment can be set using techniques known in the art or manually.

[0084] On the other hand, in another embodiment, the method for identifying the call topic based on the customer text and the customer service text includes:

[0085] Try to group the multiple customer texts and customer service texts together;

[0086] Identify several keywords in the customer text and customer service text within the group;

[0087] Comparing the keywords with topics in a preset topic library to obtain topic relevance and a maximum relevance value among the topic relevances;

[0088] Comparing the topic corresponding to the maximum correlation value with the previously identified conversation topic to obtain a comparison correlation degree;

[0089] When the maximum correlation value is greater than a preset threshold or the comparison correlation degree is greater than a preset threshold, the topic corresponding to the maximum correlation value is used as the conversation topic;

[0090] When the maximum correlation value is not greater than a preset threshold and the comparison correlation degree is not greater than a preset threshold, an attempt is made to group the plurality of customer texts and customer service texts again.

[0091] Dividing each customer voice and customer service voice into a separate group for call topic recognition is a special grouping method, that is, each group contains only one customer voice or customer service voice. As an improved method, try to group two or more customer voices and customer service voices into a group and recognize the call topic together. This can improve the accuracy of call topic recognition. For example, "Customer: Hello, I would like to know about your air fryer, especially model A123" and "Customer Service: Hello, thank you for your call. Model A123 air fryer is one of our best-selling products" are grouped together. For example, the topic library is {"Introduction to the product", "Price discount", "After-sales", "Instructions for use and maintenance", "Recipe introduction", "Logistics consultation"}.

[0092] The keywords obtained are "air fryer," "model A123," and "hot-selling product." First, using the large language model, we determine the correlation between "air fryer" and each topic in the topic database {"product introduction," "discounted price," "after-sales service," "instructions for use and maintenance," "recipe introduction," and "logistics consultation." The correlation between "air fryer" and "product introduction" is 0.5, "air fryer" and "discounted price" is 0.4, "air fryer" and "after-sales service" is 0.3, "air fryer" and "instructions for use and maintenance" is 0.4, "air fryer" and "recipe introduction" is 0.3, and "air fryer" and "logistics consultation" is 0.2. This means that the maximum correlation value among the topic associations is 0.5, corresponding to the topic "product introduction." Similarly, the topic corresponding to the maximum correlation between "model A123" and "hot-selling product" and the topic database is also "product introduction." The final topic of the conversation for this group is "product introduction."

[0093] As another example, “Customer: What is the capacity of this air fryer? How many people in the family is it suitable for?”, “Customer Service: The capacity of this air fryer is 5 liters, suitable for a family of 2-4 people”, “Customer: How is its heating speed? Is there a preheating function?”, “Customer Service: It heats up very quickly, and the general preheating time only takes 3-5 minutes. It also has a preheating function, and the preheating time can be set in advance” are grouped together. The keywords that can be obtained are “capacity”, “home use”, “heating speed”, “preheating”, and “function”. The correlation values ​​of these keywords with the topics in the topic library are all lower than the preset threshold. For example, the threshold is 0.5. When the correlation values ​​are all lower than 0.5, the topic corresponding to the maximum correlation value is “Instructions for Use and Maintenance”. The comparison correlation degree between the “Instructions for Use and Maintenance” and the previously identified call topic “Introduction to the Product” is 0.8.

[0094] It can be seen that although the maximum correlation value between the topic "Instructions for Use and Maintenance" and the keywords in the group is no higher than 0.5, the comparison correlation between the topic "Instructions for Use and Maintenance" and the previous conversation topic "Introduction of Products" is 0.8, which is sufficient to indicate that the topic "Instructions for Use and Maintenance" is suitable as the conversation topic for this group.

[0095] On the other hand, this specification provides a specific method for comparing the topic corresponding to the maximum correlation value with the previously identified call topic to obtain the comparison correlation degree, please refer to the attached Figure 5 ,include:

[0096] Step S202) Read the voice of the historical customer service call, recognize the text of the voice, and record it as the historical text;

[0097] Step S204) identifying a plurality of keywords of the historical text, and grouping the historical text according to the keywords;

[0098] Step S206) obtaining the topic of the grouped historical texts based on the comparison of the grouped keywords with a preset topic library, and recording it as the grouped topic;

[0099] Step S208) Arrange all grouped topics in chronological order to obtain a historical topic sequence;

[0100] Step S210) Repeat the above steps to obtain multiple historical topic sequences;

[0101] Step S212) Pairing two adjacent historical topics in the historical topic sequence and recording them as a topic pair;

[0102] Step S214) Obtaining the relevance of historical topics based on the frequency of appearance of the topic pair in the entire historical topic sequence;

[0103] Step S216) The topic corresponding to the maximum correlation value and two historical topics closest to the previously identified conversation topic are respectively obtained, and a comparison correlation is obtained based on the correlation between the two closest historical topics.

[0104] The comparison correlation can be directly obtained through a large language model. However, the large language model lacks domain specificity. The method provided in this embodiment can obtain a comparison correlation that is more suitable for the industry field based on the historical texts in the industry field. Among them, the method for obtaining the correlation between the topic of the customer service text and the topic of the call includes: respectively obtaining the two historical topics that are closest to the topic of the customer service text and the topic of the call, and obtaining the correlation between the topic of the customer service text and the topic of the call based on the correlation between the two closest historical topics.

[0105] On the other hand, this manual provides a method to configure the reply library, please refer to the attached Figure 6 ,include:

[0106] Step S302) Read the voice of the historical customer service call, recognize the text of the voice, and record it as the historical text;

[0107] Step S304) extracting the customer text and customer service text from the historical text;

[0108] Step S306) picking up keywords of the customer text and recording them as customer keywords, and classifying the customer text according to the customer keywords;

[0109] Step S308) Obtaining customer questions based on the customer texts of the same category;

[0110] Step S310) obtaining a plurality of customer service texts that are adjacent to the customer text of the same category, and obtaining a plurality of answers to the customer question based on the plurality of customer service texts;

[0111] Step S312) Obtain a response library based on the customer question and the corresponding multiple responses.

[0112] The response library configured using a large language model lacks specificity, and manually setting up a response library is inefficient. By classifying customer texts based on keywords, similar questions or similar topics asked by customers can be classified into one category. A response library is obtained based on all historical customer service texts under this category, which can be more in line with the situation in this industry field, improve the relevance between the responses in the response library and the customer texts, and help improve the quality of the responses. Among them, the method for obtaining customer questions based on the customer texts of the same category includes: generating customer questions based on the keywords of the customer texts of the same category and the pre-connected large language model.

[0113] On the other hand, please refer to the attached customer service call quality control system based on voice recognition. Figure 7 ,include:

[0114] The reading module 100 reads the voice of the customer service call in real time and identifies the voice as the customer voice or the customer service voice based on the timbre of the voice;

[0115] Recognition module 200, which recognizes the text of the customer voice and customer service voice, records them as customer text and customer service text respectively, and identifies the conversation topic based on the customer text and customer service text;

[0116] The reply module 300 obtains a plurality of optional replies based on the customer text and a pre-configured reply library, and calculates the relevance between the customer service text and the plurality of optional replies;

[0117] The interception module 400 intercepts the customer service voice when the relevance between the customer service text and the plurality of optional responses is lower than a preset threshold;

[0118] Comparison module 500, identifying the voice emotion of the customer service voice, comparing the voice emotion with a pre-configured screening emotion set, and comparing the customer service text with a pre-configured screening vocabulary;

[0119] Topic module 600, when the voice emotion does not belong to the screening emotion set and the customer service text does not include words in the screening vocabulary, identifies the topic of the customer service text;

[0120] The control module 700 releases the customer service voice when the relevance between the customer service text and the conversation topic is higher than a preset threshold, or the relevance between the topic of the customer service text and the conversation topic is higher than a preset threshold.

[0121] See also Figure 8 A schematic structural diagram of an electronic device provided in an embodiment of this specification is shown.

[0122] like Figure 8 As shown, the electronic device 1100 may include: at least one processor 1101, at least one network interface 1104, a user interface 1103, a memory 1105, and at least one communication bus 1102. The communication bus 1102 may be used to implement communication between the aforementioned components. The user interface 1103 may include buttons, and optionally may also include a standard wired interface or a wireless interface. The network interface 1104 may include, but is not limited to, a Bluetooth module, an NFC module, a Wi-Fi module, etc. The processor 1101 may include one or more processing cores. The processor 1101 utilizes various interfaces and circuits to connect the various components within the entire electronic device 1100. By running or executing instructions, programs, code sets, or instruction sets stored in the memory 1105, and accessing data stored in the memory 1105, the processor 1101 performs various functions of the routing device 1100 and processes data. Optionally, the processor 1101 may be implemented in hardware using at least one of a DSP, an FPGA, and a PLA. The processor 1101 may integrate one or a combination of a CPU, a GPU, and a modem. The CPU primarily processes the operating system, user interface, and applications; the GPU is responsible for rendering and drawing content displayed on the display; and the modem handles wireless communications.

[0123] It is understandable that the above-mentioned modem may not be integrated into the processor 1101, but may be implemented by a separate chip.

[0124] Memory 1105 may include either RAM or ROM. Optionally, memory 1105 may include non-transitory computer-readable media. Memory 1105 may be used to store instructions, programs, codes, code sets, or instruction sets. Memory 1105 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, sound playback function, image playback function, etc.), instructions for implementing the aforementioned method embodiments, etc.; the data storage area may store data related to the aforementioned method embodiments, etc. Memory 1105 may also optionally be at least one storage device located remotely from the aforementioned processor 1101. Memory 1105, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and application programs. Processor 1101 may be configured to invoke the application programs stored in memory 1105 and execute the methods described in the aforementioned embodiments.

[0125] The embodiments of this specification also provide a computer-readable storage medium having instructions stored therein that, when executed on a computer or processor, cause the computer or processor to perform the steps of the aforementioned embodiments. If the components of the aforementioned electronic device are implemented as software functional units and sold or used as independent products, they may be stored in the computer-readable storage medium.

[0126] The embodiments of this specification also provide a computer program product, including a computer program, which implements multiple steps in the above embodiments when executed by a processor.

[0127] In the absence of conflict, the technical features in this embodiment and implementation scheme can be combined arbitrarily.

[0128] In the above embodiments, all or part of the embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product comprises multiple computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted via the computer-readable storage medium. The computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium accessible by a computer or a data storage device such as a server or data center that integrates multiple available media. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital versatile disc (DVD)), or a semiconductor medium (eg, a solid state drive (SSD)).

[0129] When implemented via hardware or firmware, the aforementioned method flow is programmed into the hardware circuit to obtain the corresponding hardware circuit structure and realize the corresponding function. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit, whose logical function is determined by the user's device programming. Designers can "integrate" a digital system on a PLD through self-programming, eliminating the need for chip manufacturers to design and manufacture dedicated integrated circuit chips. Moreover, today, instead of manually manufacturing integrated circuit chips, this programming is often performed using "logic compiler" software. This is similar to the software compiler used in program development. Before compilation, the original code must also be written in a specific programming language, called a hardware description language (HDL). There are not just one HDL, but many. Those skilled in the art will also understand that simply by programming the method flow in one of the aforementioned hardware description languages ​​and programming it into the integrated circuit, a hardware circuit that implements the logical method flow can be easily obtained.

[0130] The embodiments described above are merely preferred embodiments of this specification and are not intended to limit the scope of this specification. Without departing from the design spirit of this specification, various modifications and improvements made to the technical solutions of this specification by ordinary technicians in this field should fall within the scope of protection determined by the claims of this specification.

Claims

1. A customer service call quality control method based on speech recognition, characterized in that: Includes steps: Read the voice of the customer service call in real time, and identify the voice as the customer's voice and the customer service voice according to the timbre of the voice; Recognize the texts of the customer voice and the customer service voice, record them as customer text and customer service text respectively, and identify the conversation topic according to the customer text and customer service text; According to the customer text and a pre-configured answer library, a number of optional answers are obtained, and the correlation between the customer service text and the several optional answers is calculated respectively; When the correlation between the customer service text and several optional replies is lower than a preset threshold, intercepting subsequent customer service voice; Identify the speech emotion of the subsequent customer service speech, compare the speech emotion with a pre-configured screening emotion set, identify the text of the subsequent customer service speech as a subsequent customer service text, and compare the subsequent customer service text with a pre-configured screening vocabulary; When the voice emotion does not belong to the screening emotion set and the subsequent customer service text does not include words in the screening vocabulary, identifying the topic of the subsequent customer service text; When the relevance between the subsequent customer service text and the conversation topic is higher than a preset threshold, or when the relevance between the topic of the subsequent customer service text and the conversation topic is higher than a preset threshold, the subsequent customer service voice is released.

2. The customer service call quality control method based on speech recognition according to claim 1 is characterized in that: The method for identifying the conversation topic according to the customer text and the customer service text includes: Try to group the plurality of customer texts and customer service texts into one group; Identify several keywords of the customer text and customer service text in the group; Compare the keywords with topics in a preset topic library to obtain topic relevance and the maximum relevance value among the topic relevances; Comparing the topic corresponding to the maximum correlation value with the previously identified conversation topic to obtain a comparison correlation degree; When the maximum correlation value is greater than a preset threshold or the comparison correlation degree is greater than a preset threshold, the topic corresponding to the maximum correlation value is used as the conversation topic; When the maximum correlation value is not greater than a preset threshold and the comparison correlation degree is not greater than a preset threshold, an attempt is made to group the plurality of customer texts and customer service texts again.

3. The customer service call quality control method based on speech recognition according to claim 2 is characterized in that: The method of comparing the topic corresponding to the maximum correlation value with the previously identified conversation topic to obtain the comparison correlation includes: Read the voice of historical customer service calls, recognize the text of the voice, and record it as historical text; Identifying a number of keywords of the historical text, and grouping the historical text according to the keywords; According to the comparison between the grouped keywords and a preset topic library, the topic of the grouped historical text is obtained and recorded as the grouped topic; All grouped topics are arranged in chronological order to obtain a historical topic sequence; Repeat the above steps to obtain multiple historical topic sequences; Pairing two adjacent historical topics in the historical topic sequence as a topic pair; Obtain the relevance of historical topics according to the frequency of topic pairs appearing in the entire historical topic sequence; The topic corresponding to the maximum correlation value and two historical topics closest to the previously identified conversation topic are respectively obtained, and a comparison correlation is obtained according to the correlation between the two closest historical topics.

4. The customer service call quality control method based on speech recognition according to claim 3 is characterized in that: The method for obtaining the relevance between the topic of the customer service text and the call topic includes: Two historical topics closest to the topic of the customer service text and the topic of the conversation are respectively obtained, and the relevance between the topic of the customer service text and the topic of the conversation is obtained according to the relevance between the two closest historical topics.

5. The customer service call quality control method based on speech recognition according to any one of claims 1 to 4, characterized in that: Methods for configuring the reply library include: Read the voice of historical customer service calls, recognize the text of the voice, and record it as historical text; Extracting customer text and customer service text from the historical text; The keywords of the customer text are mentioned as customer keywords, and the customer text is classified according to the customer keywords; Obtaining customer questions based on the customer texts of the same category; Obtaining a plurality of customer service texts that are adjacent to the customer text of the same category, and obtaining a plurality of answers to the customer question based on the plurality of customer service texts; According to the customer questions and the corresponding plurality of replies, a reply library is obtained.

6. The customer service call quality control method based on speech recognition according to claim 5 is characterized in that: According to the customer text of the same category, the method of obtaining the customer's question includes: The customer questions are generated based on the keywords of the customer text of the same category and the pre-connected large language model.

7. The customer service call quality control system based on speech recognition is characterized by: include: A reading module reads the voice of the customer service call in real time, and identifies the voice as the customer's voice and the customer service voice according to the timbre of the voice; A recognition module, which recognizes the text of the customer voice and the customer service voice, records them as customer text and customer service text respectively, and recognizes the conversation topic according to the customer text and the customer service text; A reply module, which obtains a plurality of optional replies according to the customer text and a pre-configured reply library, and respectively calculates the relevance between the customer service text and the plurality of optional replies; An interception module, when the correlation between the customer service text and a plurality of the optional replies is lower than a preset threshold, intercepts the customer service voice; A comparison module, which identifies the voice emotion of the customer service voice, compares the voice emotion with a pre-configured screening emotion set, and compares the customer service text with a pre-configured screening vocabulary; A topic module, when the voice emotion does not belong to the screening emotion set and the customer service text does not include words in the screening vocabulary, identifies the topic of the customer service text; The management and control module releases the customer service voice when the relevance between the customer service text and the call topic is higher than a preset threshold, or the relevance between the topic of the customer service text and the call topic is higher than a preset threshold.

8. An electronic device, characterized in that including a processor and a memory; The processor is connected to the memory; The memory is used to store executable program code; The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. Computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images