Customer service message optimization method and device, storage medium and computer equipment

Abstract

The invention relates to the technical field of intelligent customer service, can be applied to digital medical treatment and financial science and technology business scenes, and discloses a customer service message optimization method and device, a storage medium and computer equipment, and the method comprises the steps: obtaining a customer service message from a customer service seat end; performing quality inspection on the customer service message by using a plurality of role quality inspection large models to obtain a quality inspection conclusion of the customer service message, the plurality of role quality inspection large models including a user role quality inspection large model, a supervisor role quality inspection large model and a quality inspector role quality inspection large model; and sending the quality inspection conclusion to the customer service seat end, so that the customer service seat end performs customer service message optimization based on the quality inspection conclusion. The customer service quality inspection system breaks through the limitations of single visual angle assessment, feedback lagging, insufficient semantic understanding ability and the like of a traditional customer service quality inspection system, realizes conversion from post-event sampling inspection to real-time intervention, gives consideration to user experience, brand specifications and compliance risk control, and improves the service quality of customer service communication.

Description

Customer service message optimization methods and devices, storage media, and computer equipment Technical Field

[0001] This application relates to the fields of intelligent customer service, digital healthcare, and financial technology, and in particular to a customer service message optimization method and apparatus, storage medium, and computer equipment. Background Technology

[0002] In modern customer service systems, the quality of customer service messages directly impacts customer experience, service efficiency, and a company's compliance and professional image. This is especially true in industries like fintech and digital healthcare, where accuracy, standardized expression, and communication sensitivity are paramount. Every response from customer service personnel carries the crucial responsibility of delivering accurate information, maintaining user trust, and mitigating legal risks. However, currently widely used customer service quality inspection methods still have many limitations in practical application, making it difficult to effectively support service quality assurance in these demanding business scenarios.

[0003] Traditional quality inspection methods are typically based on a single role's perspective. For example, quality inspectors mainly focus on compliance with process specifications, supervisors are more concerned with the consistency of brand messaging, while users are more concerned with the friendliness of communication and the effectiveness of problem-solving. Furthermore, most quality inspection processes are retrospective sampling, resulting in delayed feedback and an inability to provide rapid optimization suggestions before or after customer service messages are sent, impacting service timeliness and user experience. Taking fintech as an example, online customer service often handles matters involving financial compliance, such as loan applications, account security, and investment management. Any omissions or misleading statements can lead to user misunderstandings or even regulatory issues. Traditional customer service quality inspections rely heavily on preset rules or general models, lacking a deep understanding of specific business contexts. For instance, they struggle to identify subtle differences between expected returns and promised returns. Similarly, in digital healthcare scenarios, customer service personnel need to provide clear health consultation services without overstepping boundaries. Existing quality inspection systems often fail to accurately identify potential risks from a professional perspective, leading to the misdirection of potentially problematic messages or their failure to be corrected in a timely manner. Summary of the Invention

[0004] In view of this, the embodiments of this application provide a customer service message optimization method and apparatus, storage medium, and computer equipment, which overcomes the limitations of traditional customer service quality inspection systems such as single-perspective evaluation, delayed feedback, and insufficient semantic understanding capabilities. It realizes the transformation from post-event sampling to real-time intervention, while taking into account user experience, brand norms, and compliance risk control, thereby improving the professionalism, accuracy, and service quality of customer service communication.

[0005] According to one aspect of this application, a customer service message optimization method is provided, the method comprising: obtaining customer service messages from a customer service agent; performing quality inspection on the customer service messages using a multi-role quality inspection model to obtain a quality inspection conclusion for the customer service messages, wherein the multi-role quality inspection model includes a user role quality inspection model, a supervisor role quality inspection model, and a quality inspector role quality inspection model; and sending the quality inspection conclusion to the customer service agent so that the customer service agent can optimize the customer service messages based on the quality inspection conclusion.

[0006] Optionally, the step of sequentially calling multiple role-based quality inspection models to perform quality inspection on the customer service message and obtaining the quality inspection conclusion of the customer service message includes: selecting a target quality inspection model from the multiple quality inspection models; sequentially calling the target quality inspection model to perform quality inspection on the customer service message according to the preset order of the multiple role-based quality inspection models and obtaining the quality inspection conclusion of the customer service message, wherein the quality inspection result of the previous target quality inspection model is used as one of the inputs of the next target quality inspection model.

[0007] Optionally, when the target quality inspection model is a user role quality inspection model, a supervisor role quality inspection model, or a quality inspector role quality inspection model, the step of sequentially calling the target quality inspection model to inspect the customer service message according to a preset order of the multiple role quality inspection models to obtain the quality inspection conclusion of the customer service message includes: concatenating the customer service message, the context information corresponding to the customer service message, and a user role quality inspection prompt word template to obtain user role quality inspection prompt words; and performing user role perception analysis on the customer service message based on the user role quality inspection prompt words through the user role quality inspection model to obtain user role quality inspection results, wherein the user role perception analysis includes at least one of user emotional state analysis, user comprehension deviation analysis, and user potential confusion point analysis; and concatenating the customer service message, the context information corresponding to the customer service message, the user role quality inspection results, and the supervisor role quality inspection prompt word template. The process involves obtaining supervisor role quality inspection prompts, and then using the supervisor role quality inspection model to perform supervisor role evaluation analysis on the customer service message based on these prompts, resulting in a supervisor role quality inspection result. This supervisor role evaluation analysis includes at least one of language expression analysis, tone improvement analysis, and supplementary explanation suggestion analysis. The customer service message, its corresponding context information, the supervisor role quality inspection result, and the quality inspection prompt template are then concatenated to obtain the quality inspection prompts for the quality inspector role. Based on these prompts, the quality inspector role quality inspection model performs a compliance check on the customer service message, resulting in a quality inspector role quality inspection result. Finally, a quality inspection conclusion is derived by combining the user role quality inspection result, the supervisor role quality inspection result, and the quality inspector role quality inspection result. This compliance check includes at least one of enterprise script standardization check, legal compliance requirement check, and service quality standard check.

[0008] Optionally, selecting a target quality inspection model from the plurality of quality inspection models includes: obtaining the quality inspection rules of each role-based quality inspection model corresponding to the customer service agent terminal, and selecting a target quality inspection model from the plurality of quality inspection models according to the quality inspection rules. The quality inspection rules are determined as follows: based on the historical quality inspection results of each role-based quality inspection model corresponding to the customer service agent terminal, a historical quality inspection score for the customer service agent terminal is determined by each role-based quality inspection model; based on the historical quality inspection score, a baseline sampling rule for each role-based quality inspection model for the customer service agent terminal is determined; and based on the recent quality inspection results of each role-based quality inspection model for the customer service agent terminal, the baseline sampling rule is adjusted to obtain the quality inspection rules for the role-based quality inspection model corresponding to the customer service agent terminal. The recent quality inspection results include the quality inspection results of the most recent preset duration and / or the quality inspection results of the most recent preset number of times, and the quality inspection results corresponding to each role-based quality inspection model include a quality inspection score.

[0009] Optionally, the step of determining the baseline sampling rules for the customer service agent terminal based on the historical quality inspection scores of each role's quality inspection big model, and adjusting the baseline sampling rules based on the recent quality inspection results of each role's quality inspection big model for the customer service agent terminal, to obtain the quality inspection rules for the role's quality inspection big model corresponding to the customer service agent terminal, includes: for any role's quality inspection big model, determining the historical average quality inspection score of the customer service agent terminal based on the historical quality inspection scores; determining the target score interval hit by the historical average quality inspection score among multiple preset score intervals corresponding to the role's quality inspection big model; and determining the sampling probability of the target score interval as the sampling rate based on the correspondence between the preset score intervals corresponding to the role's quality inspection big model and the quality inspection level. The document describes the baseline sampling rules corresponding to the large-scale quality inspection model for each role. When the recent quality inspection results of any large-scale quality inspection model for the customer service agent are obtained, the document calculates the recent average quality inspection score corresponding to the recent quality inspection results based on the quality inspection scores in the recent quality inspection results. It then calculates the difference between the recent average quality inspection score and the historical average quality inspection score corresponding to any large-scale quality inspection model for each role. Based on the correspondence between the preset difference range and the probability adjustment value, it determines the target probability adjustment value corresponding to the preset difference range where the difference is located. The document then adjusts the sampling probability corresponding to any large-scale quality inspection model for each role based on the target probability adjustment value. Finally, it updates the baseline sampling rules corresponding to any large-scale quality inspection model for each role based on the adjusted sampling probability to obtain the quality inspection rules.

[0010] Optionally, the method further includes: monitoring the input information of the customer service agent on the customer service dialogue interface; dynamically displaying the input information in the input information display area of ​​the quality inspection display window, and extracting customer service message fragments from the customer service agent based on the input information, wherein the quality inspection display window is above the customer service dialogue interface; performing quality inspection on the customer service message fragments using a multi-role quality inspection model to obtain the fragment quality inspection conclusions corresponding to the customer service message fragments; sending the fragment quality inspection conclusions to the customer service agent, so that the customer service agent displays the fragment quality inspection conclusions in the quality inspection information display area of ​​the quality inspection display window, wherein the input information in the input information display area and the quality inspection conclusions in the quality inspection information display area are displayed in a matched state, and when any input information is selected, the fragment quality inspection conclusions of the corresponding customer service message fragments are displayed differently, and when any input information is updated, the fragment quality inspection conclusions of the corresponding customer service message fragments are updated accordingly.

[0011] Optionally, obtaining customer service messages from the customer service agent includes: listening to customer service messages sent by the customer service agent through the customer service dialogue interface; after sending the quality inspection conclusion to the customer service agent, the method further includes: displaying the quality inspection conclusion in the quality inspection information display area, deleting the fragment of quality inspection information, so that the customer service message and the quality inspection conclusion are displayed in a matched state, and displaying the corresponding quality inspection conclusion differently when any customer service message is selected.

[0012] According to another aspect of this application, a customer service message optimization device is provided. The device includes: a message acquisition module for acquiring customer service messages from a customer service agent; a message quality inspection module for performing quality inspection on the customer service messages using a multi-role quality inspection model to obtain a quality inspection conclusion for the customer service messages, wherein the multi-role quality inspection model includes a user role quality inspection model, a supervisor role quality inspection model, and a quality inspector role quality inspection model; and a message optimization module for sending the quality inspection conclusion to the customer service agent so that the customer service agent can optimize the customer service messages based on the quality inspection conclusion.

[0013] Optionally, the message quality inspection module is further configured to: select a target quality inspection model from the plurality of quality inspection models; sequentially call the target quality inspection model to perform quality inspection on the customer service message according to the preset order of the plurality of role quality inspection models, and obtain the quality inspection conclusion of the customer service message, wherein the quality inspection result of the previous target quality inspection model is used as one of the inputs of the next target quality inspection model.

[0014] Optionally, when the target quality inspection model is a user role quality inspection model, a supervisor role quality inspection model, or a quality inspector role quality inspection model, the message quality inspection module is further configured to: concatenate the customer service message, the context information corresponding to the customer service message, and the user role quality inspection prompt template to obtain user role quality inspection prompts; and perform user role perception analysis on the customer service message based on the user role quality inspection prompts using the user role quality inspection model to obtain user role quality inspection results, wherein the user role perception analysis includes at least one of user emotional state analysis, user comprehension deviation analysis, and user potential confusion point analysis; concatenate the customer service message, the context information corresponding to the customer service message, the user role quality inspection results, and the supervisor role quality inspection prompt template to obtain supervisor role quality inspection prompts; and perform user role quality inspection on the customer service message based on the user role quality inspection prompt template to obtain supervisor role quality inspection prompts. The large model performs supervisor role evaluation analysis on the customer service message based on the supervisor role quality inspection prompts to obtain supervisor role quality inspection results. The supervisor role evaluation analysis includes at least one of language expression analysis, tone improvement analysis, and supplementary explanation suggestion analysis. The customer service message, its corresponding context information, the supervisor role quality inspection results, and the quality inspection prompt template are concatenated to obtain quality inspection prompts. The quality inspector role quality inspection large model then performs compliance checks on the customer service message based on these prompts to obtain quality inspector role quality inspection results. Finally, a quality inspection conclusion is derived by combining the user role quality inspection results, the supervisor role quality inspection results, and the quality inspector role quality inspection results. The compliance checks include at least one of corporate communication style standard checks, legal compliance requirement checks, and service quality standard checks.

[0015] Optionally, the message quality inspection module is further configured to: obtain the quality inspection rules of each role's quality inspection big model corresponding to the customer service agent terminal, and select a target quality inspection big model from the multiple quality inspection big models according to the quality inspection rules, wherein the quality inspection rules are determined by: determining the historical quality inspection score of each role's quality inspection big model for the customer service agent terminal based on the historical quality inspection results of each role's quality inspection big model, determining the benchmark sampling inspection rules of each role's quality inspection big model for the customer service agent terminal based on the historical quality inspection scores, and adjusting the benchmark sampling inspection rules based on the recent quality inspection results of each role's quality inspection big model for the customer service agent terminal, so as to obtain the quality inspection rules of the role's quality inspection big model corresponding to the customer service agent terminal, wherein the recent quality inspection results include the quality inspection results of the most recent preset duration and / or the quality inspection results of the most recent preset number of times, and the quality inspection results corresponding to each role's quality inspection big model include the quality inspection score.

[0016] The message quality inspection module is further configured to: for any role quality inspection big model, determine the historical average quality inspection score of the customer service agent based on the historical quality inspection scores, determine the target score interval hit by the historical average quality inspection score among multiple preset score intervals corresponding to the role quality inspection big model, and determine the sampling probability of the target score interval as the benchmark sampling rule corresponding to the role quality inspection big model based on the correspondence between the preset score interval and the quality inspection level corresponding to the role quality inspection big model; when a recent quality inspection result of any role quality inspection big model for the customer service agent is obtained, calculate the recent average quality inspection score corresponding to the recent quality inspection result based on the quality inspection scores in the recent quality inspection result, calculate the difference between the recent average quality inspection score and the historical average quality inspection score corresponding to the any role quality inspection big model, determine the target probability adjustment value corresponding to the preset difference interval where the difference is located based on the correspondence between the preset difference interval and the probability adjustment value, adjust the sampling probability corresponding to the any role quality inspection big model based on the target probability adjustment value, and update the benchmark sampling rule corresponding to the any role quality inspection big model based on the adjusted sampling probability to obtain the quality inspection rule.

[0017] Optionally, the message acquisition module is further configured to: monitor the input information of the customer service agent on the customer service dialogue interface; the message quality inspection module is further configured to: dynamically display the input information in the input information display area of ​​the quality inspection display window, and extract customer service message fragments from the customer service agent based on the input information, wherein the quality inspection display window is above the customer service dialogue interface; perform quality inspection on the customer service message fragments using a large-scale quality inspection model with multiple roles, and obtain the fragment quality inspection conclusion corresponding to the customer service message fragment; the message optimization module is further configured to: send the fragment quality inspection conclusion to the customer service agent, so that the customer service agent displays the fragment quality inspection conclusion in the quality inspection information display area of ​​the quality inspection display window, wherein the input information in the input information display area and the quality inspection conclusion in the quality inspection information display area are displayed in a matched state, and when any input information is selected, the fragment quality inspection conclusion of the corresponding customer service message fragment is displayed differently, and when any input information is updated, the fragment quality inspection conclusion of the corresponding customer service message fragment is updated accordingly.

[0018] Optionally, the message acquisition module is further configured to: listen to customer service messages sent by the customer service agent through the customer service dialogue interface; the message optimization module is further configured to: display the quality inspection conclusion in the quality inspection information display area, delete the fragment of quality inspection information, so that the customer service message and the quality inspection conclusion are displayed in a matched state, and differentiate the display of the corresponding quality inspection conclusion when any customer service message is selected.

[0019] According to another aspect of this application, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described customer service message optimization method.

[0020] According to another aspect of this application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described customer service message optimization method.

[0021] By employing the above technical solutions, this application provides a customer service message optimization method, apparatus, storage medium, and computer equipment. This method acquires messages from customer service agents in real time and utilizes a large-scale model employing the perspectives of users, supervisors, and quality inspectors to conduct multi-dimensional quality checks on the messages. Finally, the quality inspection conclusions are fed back to the customer service end to support its immediate optimization. This application overcomes the limitations of traditional customer service quality inspection systems, such as single-perspective evaluation, delayed feedback, and insufficient semantic understanding capabilities. It achieves a shift from post-event sampling to real-time intervention, while simultaneously considering user experience, brand standards, and compliance risk control. This improves the professionalism, accuracy, and service quality of customer service communication, and has broad applicability and practical value in highly sensitive industries such as fintech and digital healthcare.

[0022] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0023] The accompanying drawings, which are included to provide a further understanding of this application and constitute a part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 shows a schematic flowchart of a customer service message optimization method provided by an embodiment of this application; Figure 2 shows a schematic flowchart of another customer service message optimization method provided by an embodiment of this application; Figure 3 shows a schematic structural diagram of a customer service message optimization device provided by an embodiment of this application. Detailed Implementation

[0024] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0025] This embodiment provides a customer service message optimization method, as shown in Figure 1. The method includes: step 101, obtaining customer service messages from the customer service agent's terminal.

[0026] In industries like fintech and digital healthcare, where service quality and compliance requirements are increasingly stringent, traditional customer service quality inspection methods are struggling to meet the complex and ever-changing service demands due to their limited perspective, delayed feedback, and insufficient semantic understanding. To address this, this embodiment provides a customer service message optimization method based on a multi-role quality inspection model. This method acquires messages sent from customer service agents in real time and conducts a comprehensive evaluation from three key roles: user, supervisor, and quality inspector. The quality inspection conclusions are then fed back to the customer service end to support immediate message content optimization. This method enhances service professionalism and compliance while ensuring communication quality. First, it acquires customer service messages from agents, collecting messages that agents are about to send or have already sent. For example, in fintech, when a customer service representative answers a user's question about credit card annual fee waiver policies, the system automatically captures that message; similarly, in digital healthcare, when a customer service representative assists a user with an initial diagnosis description and recommends a department, the message collection process can be triggered.

[0027] Step 102: Use a multi-role quality inspection model to perform quality inspection on the customer service message and obtain the quality inspection conclusion of the customer service message. The multi-role quality inspection model includes a user role quality inspection model, a supervisor role quality inspection model, and a quality inspector role quality inspection model.

[0028] Traditional quality inspection often relies on a single model or rule base. This application's embodiment introduces three different comprehensive quality inspection models, simulating the perspectives of users, supervisors, and quality inspectors, respectively, to achieve multi-dimensional evaluation of message quality. Specifically, the user-role quality inspection model focuses on whether the message is clear and easy to understand, whether it solves the user's actual problem, and whether the tone is friendly; the supervisor-role quality inspection model emphasizes the consistency of brand language, the matching degree of marketing strategies, and the maintenance of corporate image; the quality inspector-role quality inspection model focuses more on compliance, professionalism, and risk control, such as whether there are misleading statements or violations of regulatory provisions. By integrating the judgment results of these three roles, the system can generate a more comprehensive quality inspection conclusion, avoiding the service quality decline caused by one-sided judgments. This multi-role collaborative quality inspection mechanism is particularly suitable for industries such as fintech and digital healthcare, which highly value both compliance and user experience. For example, in fintech scenarios, if customer service personnel omit the warning about investment risks when introducing financial products, a quality inspector model can identify compliance deficiencies, a supervisor model can point out inconsistencies in wording, and a user-centric model might deem the expression insufficiently clear and understandable. Combining these three perspectives helps to formulate comprehensive improvement suggestions. In digital healthcare scenarios, if customer service personnel use diagnostic language such as "your symptoms may indicate a certain disease," a quality inspector model can identify the risk of unauthorized medical practice, a user model can point out a lack of reassurance in the language, and a supervisor-centric model might detect a failure to direct the user to a platform-designated doctor. The combined feedback from these three sources can effectively mitigate potential legal risks and improve service quality.

[0029] Step 103: Send the quality inspection conclusion to the customer service agent terminal so that the customer service agent terminal can optimize customer service messages based on the quality inspection conclusion.

[0030] Finally, the quality inspection results are sent to the customer service agents, allowing them to optimize customer service messages based on these results. This transforms quality inspection from mere post-event monitoring into an embedded part of the customer service personnel's actual operational processes, enabling real-time intervention and optimization. Customer service personnel can receive prompts before sending and modify message content, or quickly adjust subsequent communication strategies after sending, improving communication efficiency and service quality. For example, in a fintech platform, when a customer service representative answers a user's question about loan interest rates, the system can detect ambiguity in the fixed interest rate description before the user clicks send and prompt them to supplement the explanation of the term and floating mechanism, thus avoiding misleading the user. In digital healthcare scenarios, if a customer service representative misuses medical terminology, causing user confusion, the system can immediately prompt them to change to a more colloquial expression, improving communication effectiveness. This not only achieves real-time, multi-dimensional quality inspection of customer service messages but also provides personalized, scenario-based optimization suggestions, improving the quality and compliance of customer service communication. Compared to traditional quality inspection methods, this approach is closer to actual business needs, especially in the highly sensitive fintech and digital healthcare industries, effectively reducing legal risks, enhancing user trust, and improving brand image.

[0031] By applying the technical solution of this embodiment, messages from customer service agents are acquired in real time, and a large-scale model from the perspectives of users, supervisors, and quality inspectors is used to conduct multi-dimensional quality inspections of these messages. The inspection results are then fed back to the customer service side to support immediate optimization. This embodiment overcomes the limitations of traditional customer service quality inspection systems, such as single-perspective evaluation, delayed feedback, and insufficient semantic understanding capabilities. It achieves a shift from post-event sampling to real-time intervention, while simultaneously considering user experience, brand standards, and compliance risk control. This improves the professionalism, accuracy, and service quality of customer service communication, and has broad applicability and practical value in highly sensitive industries such as fintech and digital healthcare.

[0032] In this embodiment of the application, optionally, the step of sequentially calling multiple role-based quality inspection models to perform quality inspection on the customer service message and obtaining the quality inspection conclusion of the customer service message includes: selecting a target quality inspection model from the multiple quality inspection models; sequentially calling the target quality inspection model to perform quality inspection on the customer service message according to a preset order of the multiple role-based quality inspection models and obtaining the quality inspection conclusion of the customer service message, wherein the quality inspection result of the previous target quality inspection model is used as one of the inputs of the next target quality inspection model.

[0033] In this embodiment, a target quality inspection model is selected from three large-scale quality inspection models: the user role model, the supervisor role model, and the quality inspector role model. These models are then sequentially invoked according to a preset order to perform layered quality inspections on customer service messages. The quality inspection result of the previous model serves as one of the inputs to the next model, forming a progressive quality inspection process. This design ensures that the quality inspection models for each role are no longer independent judgments, but rather a collaborative evaluation mechanism with contextual dependence and information transmission, thereby improving the coherence and depth of the quality inspection conclusions. For example, in a fintech scenario, when a customer service representative replies to a user about "explanation of the returns on investment products," the user role model is first invoked to determine whether the message is easy to understand and meets the user's comprehension needs. Subsequently, the supervisor role model, based on the results of the previous step, further evaluates whether the wording aligns with the brand's promotional strategy. Finally, the quality inspector role model, based on the combined conclusions of the first two models, focuses on identifying any misleading statements or compliance risks. This layered, progressive quality inspection approach helps to ensure professionalism and compliance while also considering user experience and communication effectiveness. Similarly, in digital healthcare scenarios, when customer service personnel respond to users' questions about "medication precautions," the user role model can first determine whether the response is clear and reassuring; then, the supervisor role model can assess whether it conforms to the platform's unified script standards; finally, the quality inspector model, combined with the preceding judgments, can identify whether there is any unauthorized medical advice or missing information. This structured quality inspection process not only improves the logic and consistency of the model's judgments but also enhances its ability to understand and respond to complex semantics.

[0034] In this embodiment of the application, optionally, when the target quality inspection model is a user role quality inspection model, a supervisor role quality inspection model, or a quality inspector role quality inspection model, the step of sequentially calling the target quality inspection model to perform quality inspection on the customer service message according to the preset order of the multiple role quality inspection models, and obtaining the quality inspection conclusion of the customer service message, includes: concatenating the customer service message, the context information corresponding to the customer service message, and a user role quality inspection prompt word template to obtain user role quality inspection prompt words, and performing user role perception analysis on the customer service message based on the user role quality inspection prompt words through the user role quality inspection model to obtain user role quality inspection results, wherein the user role perception analysis includes at least one of user emotional state analysis, user comprehension deviation analysis, and user potential confusion point analysis; concatenating the customer service message, the context information corresponding to the customer service message, the user role quality inspection result, and the supervisor role quality inspection prompt word template... The templates are spliced ​​together to obtain supervisor role quality inspection prompts. Based on these prompts, the supervisor role quality inspection model is used to perform supervisor role evaluation analysis on the customer service message, resulting in a supervisor role quality inspection result. This supervisor role evaluation analysis includes at least one of language expression analysis, tone improvement analysis, and supplementary explanation suggestion analysis. The customer service message, its corresponding context information, the supervisor role quality inspection result, and the quality inspection prompt template are then spliced ​​together to obtain quality inspection prompts for the quality inspector role. Based on these prompts, the quality inspector role quality inspection model is used to perform compliance checks on the customer service message, resulting in a quality inspector role quality inspection result. Finally, a quality inspection conclusion is drawn by combining the user role quality inspection result, the supervisor role quality inspection result, and the quality inspector role quality inspection result. This compliance check includes at least one of corporate communication style standardization check, legal compliance requirement check, and service quality standard check.

[0035] In this embodiment, firstly, for the user role quality inspection model, the customer service message and its contextual information are concatenated with a specifically designed user role quality inspection prompt template to generate user role quality inspection prompts. Based on these prompts, the user role quality inspection model can perform user role perception analysis, including but not limited to user emotional state analysis, user comprehension bias analysis, and user potential confusion point analysis. This analysis helps to deeply understand the user's emotional reactions and cognitive situation, thereby ensuring that customer service responses accurately address the user's needs and provide appropriate reassurance or clarification. Next, after obtaining the user role quality inspection results, they are concatenated together with the original customer service message, contextual information, and supervisor role quality inspection prompt template to form supervisor role quality inspection prompts. Using these prompts, the supervisor role quality inspection model can perform supervisor role evaluation analysis, covering aspects such as language expression analysis, tone improvement analysis, and supplementary explanation suggestion analysis. This step aims to ensure that customer service messages not only convey the correct information but also adopt appropriate language style and expression, enhancing the effectiveness and professionalism of communication. Finally, combining all the preliminary analysis results (i.e., customer service messages, contextual information, user role quality inspection results, and supervisor role quality inspection results) with the quality inspection role prompt template, quality inspection prompts for each role are generated. The quality inspector role's large-scale quality inspection model uses these prompts to conduct compliance checks on customer service messages, including checks on corporate communication style, legal compliance requirements, and service quality standards. The prompt templates for each role's large-scale quality inspection model not only contain quality inspection guidance information for that role but also prompts to avoid generating contradictory or redundant quality inspection results from other role's large-scale quality inspection models. This step effectively identifies and corrects potential compliance risks, ensuring that customer service communication meets the highest standards of professionalism and legality. This embodiment of the application, through this progressive quality inspection process, not only strengthens the connection and collaboration between each quality inspection stage but also improves the quality and reliability of customer service messages by comprehensively considering user experience, expression effectiveness, and compliance requirements.

[0036] In this embodiment of the application, optionally, the step of selecting a target quality inspection model from the plurality of quality inspection models includes: obtaining the quality inspection rules of each role-based quality inspection model corresponding to the customer service agent terminal, and selecting a target quality inspection model from the plurality of quality inspection models according to the quality inspection rules. The quality inspection rules are determined in the following manner: based on the historical quality inspection results of each role-based quality inspection model corresponding to the customer service agent terminal, a historical quality inspection score for the customer service agent terminal by each role-based quality inspection model is determined; based on the historical quality inspection score, a baseline sampling rule for the customer service agent terminal by each role-based quality inspection model is determined; and based on the recent quality inspection results of the customer service agent terminal by each role-based quality inspection model, the baseline sampling rule is adjusted to obtain the quality inspection rules of the role-based quality inspection model corresponding to the customer service agent terminal. The recent quality inspection results include the quality inspection results of the most recent preset duration and / or the quality inspection results of the most recent preset number of times, and the quality inspection results corresponding to each role-based quality inspection model include a quality inspection score.

[0037] In this embodiment, during the selection of a target quality inspection model from multiple role-based quality inspection models, a dynamic evaluation mechanism can be used to intelligently select an appropriate combination of quality inspection models based on the historical and recent quality inspection performance of the customer service agent. Specifically, the quality inspection rules of each quality inspection model associated with the current customer service agent, such as the user role, supervisor role, and quality inspector role, are obtained, and these rules are used to determine which quality inspection models should be used for evaluation. The process of formulating these quality inspection rules is highly adaptive and personalized. First, based on the historical quality inspection results of each role-based quality inspection model on the customer service agent's end, its corresponding historical quality inspection score is calculated, thereby evaluating the customer service representative's performance level in different dimensions. For example, if a customer service representative has received good user feedback in past conversations but has a low compliance score, the quality inspector role model may be prioritized in subsequent quality inspections to strengthen the monitoring of compliance content. Building upon this foundation, the existing baseline sampling rules are dynamically adjusted by combining the recent quality inspection results of each agent with the comprehensive quality inspection model for each role. This includes quality inspection scores over a recent period (e.g., one week, one month) or a certain number of services (e.g., the last 10 services). This recent performance feedback mechanism allows the quality inspection strategy to respond in real time to changes in customer service quality, avoiding resource waste or blind spots caused by a static quality inspection process. Quality inspection rules determined in this way can be precisely adapted to different agents and their varying skill gaps, thereby improving the efficiency and relevance of quality inspections. For example, in a fintech scenario, if a customer service representative has repeatedly scored low on "completeness of risk warnings" recently, the weight of the quality inspector role model will be automatically increased in the quality inspection process, ensuring that such issues are closely monitored in subsequent conversations. In a digital healthcare scenario, if an agent performs poorly in user reassurance and communication guidance, the intervention frequency of the user role quality inspection model will be increased to help them optimize their communication methods.

[0038] In this embodiment of the application, optionally, the step of determining the benchmark sampling rules for the customer service agent terminal based on the historical quality inspection scores of each role's quality inspection big model, and adjusting the benchmark sampling rules based on the recent quality inspection results of the customer service agent terminal by each role's quality inspection big model to obtain the quality inspection rules for the role's quality inspection big model corresponding to the customer service agent terminal, includes: for any role's quality inspection big model, determining the historical average quality inspection score of the customer service agent terminal based on the historical quality inspection scores, determining the target score interval hit by the historical average quality inspection score among multiple preset score intervals corresponding to the role's quality inspection big model, and determining the sampling probability of the target score interval based on the correspondence between the preset score intervals corresponding to the role's quality inspection big model and the quality inspection level. The rate serves as the benchmark sampling rule corresponding to the large-scale quality inspection model for the role. When the recent quality inspection results of any large-scale quality inspection model for the customer service agent are obtained, the recent average quality inspection score corresponding to the recent quality inspection result is calculated based on the quality inspection score in the recent quality inspection result. The difference between the recent average quality inspection score and the historical average quality inspection score corresponding to any large-scale quality inspection model for the role is calculated. The target probability adjustment value corresponding to the preset difference interval where the difference is located is determined according to the correspondence between the preset difference interval and the probability adjustment value. The sampling probability corresponding to any large-scale quality inspection model for the role is adjusted according to the target probability adjustment value. The benchmark sampling rule corresponding to any large-scale quality inspection model for the role is updated according to the adjusted sampling probability to obtain the quality inspection rule.

[0039] In this embodiment, based on the historical and recent quality inspection performance of customer service agents, the sampling rules of the role-based quality inspection model are dynamically generated and adjusted, thereby achieving personalization and intelligence in the quality inspection strategy. Specifically, firstly, for each role-based quality inspection model (such as user role, supervisor role, and quality inspector role quality inspection model), a historical average quality inspection score is calculated based on the historical quality inspection scores of the customer service agent under that role. Then, this average score is compared with multiple preset scoring intervals to find which target scoring interval it matches. Each scoring interval corresponds to a preset quality inspection level, and each quality inspection level corresponds to a baseline sampling probability, for example: low-risk agents have a low sampling probability, and high-risk agents have a high sampling probability. Through this process, an initial baseline sampling rule is determined for each role-based quality inspection model, that is, the probability that the role model will be called in subsequent quality inspection processes. This mechanism allows the system to automatically adjust the quality inspection intensity based on the historical performance of customer service agents. Subsequently, after obtaining the recent quality inspection results (such as quality inspection scores in recent periods or several services) of the same customer service agent from the quality inspection big data model for that role, a recent average quality inspection score is further calculated. This recent average score is compared with the previous historical average score, and the difference between the two is calculated. Based on the preset difference range in which this difference falls, the corresponding probability adjustment value is found, and the original sampling probability is adjusted accordingly. For example, if a customer service agent's recent average score under the quality inspector role is significantly higher than the historical average score, it indicates that their compliance performance has improved, so the sampling probability of that role model is lowered; conversely, if the recent score decreases, the sampling probability is increased to strengthen monitoring. This dynamic adjustment mechanism ensures that the quality inspection strategy can flexibly adapt to changes in customer service agent performance.

[0040] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to fully illustrate the specific implementation process of this embodiment, another customer service message optimization method is provided, as shown in Figure 2. The method includes: step 201, listening to the input information of the customer service agent on the customer service dialogue interface; dynamically displaying the input information in the input information display area of ​​the quality inspection display window, and extracting customer service message fragments from the customer service agent based on the input information, wherein the quality inspection display window is above the customer service dialogue interface.

[0041] In this embodiment, by monitoring the input behavior of customer service agents, the message content being edited is displayed in real time in a separate quality inspection display window on the customer service interface. Simultaneously, corresponding customer service message segments are extracted based on the input content, providing data input for subsequent quality inspection. For example, customer service message segments can be extracted based on punctuation marks or character length. For instance, in a digital healthcare scenario, when a customer service representative inputs "You may have high blood pressure; it is recommended that you seek medical attention as soon as possible," the system also displays this segment in real time and provides input content for subsequent quality inspection.

[0042] Step 202: Use a multi-role quality inspection model to perform quality inspection on the customer service message fragment, and obtain the fragment quality inspection conclusion corresponding to the customer service message fragment.

[0043] In this embodiment, during the customer service representative's input process, a comprehensive quality inspection model encompassing the user, supervisor, and quality inspector roles is invoked to evaluate the current input segment from multiple dimensions and generate a segment quality inspection conclusion. This multi-role collaborative quality inspection mechanism overcomes the limitations of traditional quality inspection models' single-perspective approach, making the quality inspection conclusions more comprehensive and targeted. For example, in a digital healthcare scenario, the user role model can identify whether the expression is clear and whether the reassurance is adequate; the supervisor model determines whether it conforms to the platform's unified communication style; and the quality inspector model focuses on identifying whether diagnostic suggestions have been provided beyond their authority.

[0044] Step 203: Send the segment quality inspection conclusion to the customer service agent terminal, so that the customer service agent terminal displays the segment quality inspection conclusion in the quality inspection information display area of ​​the quality inspection display window. The input information in the input information display area and the quality inspection conclusion in the quality inspection information display area are displayed in a matched state. When any input information is selected, the segment quality inspection conclusion of the corresponding customer service message segment is displayed differently. When any input information is updated, the segment quality inspection conclusion of the corresponding customer service message segment is updated accordingly.

[0045] In this embodiment, the quality inspection conclusions are fed back to the quality inspection information display area of ​​the customer service interface in real time, and are matched one-to-one with the content in the input information display area. When a customer service representative selects a piece of input content, the corresponding quality inspection conclusion is displayed in a differentiated manner; when the input content is modified, the quality inspection conclusion is also dynamically updated accordingly. This design achieves synchronous linkage between quality inspection information and input behavior, providing customer service representatives with immediate optimization suggestions. For example, after a customer service representative modifies "You may have high blood pressure" to "Based on your described symptoms, it is recommended that you consult a professional doctor for diagnosis," the quality inspection conclusion is automatically updated, prompting "The expression is more standardized, it is recommended to keep it."

[0046] Step 204: Listen to customer service messages sent by the customer service agent through the customer service dialogue interface.

[0047] In this embodiment of the application, the message content finally sent by the customer service personnel is captured to ensure that the quality inspection process covers the entire process from input to sending.

[0048] Step 205: Use a multi-role quality inspection model to perform quality inspection on the customer service message and obtain the quality inspection conclusion of the customer service message. The multi-role quality inspection model includes a user role quality inspection model, a supervisor role quality inspection model, and a quality inspector role quality inspection model.

[0049] In this embodiment, after the customer service message is sent, three quality inspection models—user, supervisor, and quality inspector—are invoked again to comprehensively evaluate the entire message and form a final quality inspection conclusion. This process supplements and confirms the quality inspection at the input stage, ensuring the consistency of message quality before and after sending. For example, in fintech, if a message sent by a customer service representative omits a risk warning, the quality inspector model will clearly point out the compliance deficiency; in digital healthcare, if a message contains diagnostic statements, the quality inspector model can identify the risk of exceeding authority, while the user model can point out that the expression is not reassuring enough.

[0050] Step 206: Send the quality inspection conclusion to the customer service agent's terminal, display the quality inspection conclusion in the quality inspection information display area, delete the fragment of quality inspection information, so that the customer service message and the quality inspection conclusion are displayed in a matching state, and display the corresponding quality inspection conclusion differently when any customer service message is selected.

[0051] In this embodiment, after the message is sent, the customer service agent updates the quality inspection information display area, clearing fragmented quality inspection information and retaining only the final sent message and its corresponding quality inspection conclusion. This allows for matching and differentiated selection display. This design makes the quality inspection information structure clear and easy to understand, aiding customer service personnel in reviewing and learning, and improving subsequent service quality. For example, when customer service personnel view historical conversations, they can click on a sent message, and the system will highlight its corresponding quality inspection conclusion, helping them quickly identify problems and optimization directions.

[0052] By applying the technical solution of this embodiment, a complete closed-loop system is constructed, encompassing input monitoring, segment quality inspection, dynamic feedback, sending quality inspection, and conclusion display, thereby enhancing the quality control capabilities of customer service messages. Its core advantages lie in: Deep multi-role analysis: Introducing three roles—user, supervisor, and quality inspector—to analyze customer service message content layer by layer, forming a closed-loop quality inspection process and avoiding the limitations of traditional single-point quality inspection. Intelligent dynamic optimization: Combining the powerful understanding and generation capabilities of large-scale models, real-time and intelligent optimization suggestions are provided for customer service expressions, helping agents improve communication quality instantly. Chained process ensures continuity: Through intelligent process orchestration and chained context management protocols, the system ensures contextual consistency and logical coherence at each analysis stage, guaranteeing the traceability and systematic nature of quality inspection results. Compared with existing technologies, this application achieves more granular, intelligent, and closed-loop customer service message optimization and quality inspection, effectively improving customer satisfaction and enterprise service quality.

[0053] Furthermore, as a specific implementation of the method in Figure 1, this application embodiment provides a customer service message optimization device, as shown in Figure 3. The device includes: a message acquisition module, used to acquire customer service messages from a customer service agent; a message quality inspection module, used to perform quality inspection on the customer service messages using a multi-role quality inspection model to obtain the quality inspection conclusion of the customer service messages, wherein the multi-role quality inspection model includes a user role quality inspection model, a supervisor role quality inspection model, and a quality inspector role quality inspection model; and a message optimization module, used to send the quality inspection conclusion to the customer service agent so that the customer service agent can optimize the customer service messages based on the quality inspection conclusion.

[0054] In this embodiment of the application, optionally, the message quality inspection module is further configured to: select a target quality inspection model from the plurality of quality inspection models; sequentially call the target quality inspection model to perform quality inspection on the customer service message according to the preset order of the plurality of role quality inspection models, and obtain the quality inspection conclusion of the customer service message, wherein the quality inspection result of the previous target quality inspection model is used as one of the inputs of the next target quality inspection model.

[0055] In this embodiment of the application, optionally, when the target quality inspection model is a user role quality inspection model, a supervisor role quality inspection model, or a quality inspector role quality inspection model, the message quality inspection module is further configured to: concatenate the customer service message, the context information corresponding to the customer service message, and the user role quality inspection prompt word template to obtain user role quality inspection prompt words; and perform user role perception analysis on the customer service message based on the user role quality inspection prompt words through the user role quality inspection model to obtain user role quality inspection results, wherein the user role perception analysis includes at least one of user emotional state analysis, user comprehension deviation analysis, and user potential confusion point analysis; concatenate the customer service message, the context information corresponding to the customer service message, the user role quality inspection results, and the supervisor role quality inspection prompt word template to obtain supervisor role quality inspection prompt words; and perform user role perception analysis on the customer service message based on the user role quality inspection prompt words through the supervisor role quality inspection model to obtain user role quality inspection results. The role-based quality inspection model performs a supervisor role evaluation analysis on the customer service message based on the supervisor role quality inspection prompts, obtaining a supervisor role quality inspection result. The supervisor role evaluation analysis includes at least one of language expression analysis, tone improvement analysis, and supplementary explanation suggestion analysis. The customer service message, its corresponding context information, the supervisor role quality inspection result, and the quality inspection prompt template are concatenated to obtain the quality inspection prompts. The quality inspector role quality inspection model then performs a compliance check on the customer service message based on these prompts, obtaining the quality inspector role quality inspection result. Finally, a quality inspection conclusion is derived by combining the user role quality inspection result, the supervisor role quality inspection result, and the quality inspector role quality inspection result. The compliance check includes at least one of corporate communication style standardization check, legal compliance requirement check, and service quality standard check.

[0056] Optionally, in this embodiment, the message quality inspection module is further configured to: obtain the quality inspection rules of each role's quality inspection big model corresponding to the customer service agent terminal, and select a target quality inspection big model from the plurality of quality inspection big models according to the quality inspection rules, wherein the quality inspection rules are determined by: determining the historical quality inspection score of each role's quality inspection big model for the customer service agent terminal based on the historical quality inspection results of each role's quality inspection big model, determining the benchmark sampling inspection rule of each role's quality inspection big model for the customer service agent terminal based on the historical quality inspection score, and adjusting the benchmark sampling inspection rule based on the recent quality inspection results of each role's quality inspection big model for the customer service agent terminal, so as to obtain the quality inspection rules of the role's quality inspection big model corresponding to the customer service agent terminal, wherein the recent quality inspection results include the quality inspection results of the most recent preset duration and / or the quality inspection results of the most recent preset number of times, and the quality inspection results corresponding to each role's quality inspection big model include the quality inspection score.

[0057] In this embodiment, the message quality inspection module is further configured to: for any role quality inspection big model, determine the historical average quality inspection score of the customer service agent based on the historical quality inspection scores; determine the target score interval hit by the historical average quality inspection score among multiple preset score intervals corresponding to the role quality inspection big model; and determine the sampling probability of the target score interval as the benchmark sampling rule corresponding to the role quality inspection big model based on the correspondence between the preset score intervals corresponding to the role quality inspection big model and the quality inspection level; when the recent quality inspection results of any role quality inspection big model for the customer service agent are obtained... When the result is obtained, the recent quality inspection average score corresponding to the recent quality inspection result is calculated based on the quality inspection score in the recent quality inspection result. The difference between the recent quality inspection average score and the historical quality inspection average score corresponding to the quality inspection big model of any role is calculated. The target probability adjustment value corresponding to the preset difference interval where the difference is located is determined according to the correspondence between the preset difference interval and the probability adjustment value. The sampling probability corresponding to the quality inspection big model of any role is adjusted according to the target probability adjustment value. The benchmark sampling rule corresponding to the quality inspection big model of any role is updated according to the adjusted sampling probability to obtain the quality inspection rule.

[0058] Optionally, in this embodiment, the message acquisition module is further configured to: monitor the input information of the customer service agent on the customer service dialogue interface; the message quality inspection module is further configured to: dynamically display the input information in the input information display area of ​​the quality inspection display window, and extract customer service message fragments from the customer service agent based on the input information, wherein the quality inspection display window is above the customer service dialogue interface; perform quality inspection on the customer service message fragments using a large-scale quality inspection model with multiple roles, and obtain the fragment quality inspection conclusions corresponding to the customer service message fragments; the message optimization module is further configured to: send the fragment quality inspection conclusions to the customer service agent, so that the customer service agent displays the fragment quality inspection conclusions in the quality inspection information display area of ​​the quality inspection display window, wherein the input information in the input information display area and the quality inspection conclusions in the quality inspection information display area are displayed in a matched state, and when any input information is selected, the fragment quality inspection conclusions of the corresponding customer service message fragments are displayed differently, and when any input information is updated, the fragment quality inspection conclusions of the corresponding customer service message fragments are updated accordingly.

[0059] Optionally, in this embodiment of the application, the message acquisition module is further configured to: listen to customer service messages sent by the customer service agent through the customer service dialogue interface; the message optimization module is further configured to: display the quality inspection conclusion in the quality inspection information display area, delete the fragment of quality inspection information, so that the customer service message and the quality inspection conclusion are displayed in a matched state, and differentiate the display of the corresponding quality inspection conclusion when any customer service message is selected.

[0060] It should be noted that other corresponding descriptions of the functional units involved in the customer service message optimization device provided in this application embodiment can be found in the corresponding descriptions in the methods of Figures 1 and 2, and will not be repeated here.

[0061] This application also provides a computer device, specifically a personal computer, server, network device, etc. The computer device includes a bus, processor, memory, and communication interface, and may also include input / output interfaces and a display device. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores location information. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the various method embodiments.

[0062] Those skilled in the art will understand that the structure of the computer device described above is only a partial structure related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. A specific computer device may include more or fewer components, or combine certain components, or have different component arrangements.

[0063] In one embodiment, a computer-readable storage medium is provided, which may be non-volatile or volatile, having stored thereon a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0064] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0065] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0066] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, graphics processors, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0067] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0068] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for optimizing customer service messages, characterized in that, The method includes: obtaining customer service messages from customer service agents; performing quality inspection on the customer service messages using a multi-role quality inspection model to obtain a quality inspection conclusion for the customer service messages, wherein the multi-role quality inspection model includes a user role quality inspection model, a supervisor role quality inspection model, and a quality inspector role quality inspection model; and sending the quality inspection conclusion to the customer service agent so that the customer service agent can optimize the customer service messages based on the quality inspection conclusion.

2. The method according to claim 1, characterized in that, The step of sequentially calling multiple role-based quality inspection models to inspect the customer service message and obtain the quality inspection conclusion of the customer service message includes: selecting a target quality inspection model from the multiple quality inspection models; sequentially calling the target quality inspection model to inspect the customer service message according to the preset order of the multiple role-based quality inspection models and obtaining the quality inspection conclusion of the customer service message, wherein the quality inspection result of the previous target quality inspection model is used as one of the inputs of the next target quality inspection model.

3. The method according to claim 2, characterized in that, When the target quality inspection model is a user role quality inspection model, a supervisor role quality inspection model, or a quality inspector role quality inspection model, the step of sequentially calling the target quality inspection model to inspect the customer service message according to the preset order of the multiple role quality inspection models to obtain the quality inspection conclusion of the customer service message includes: concatenating the customer service message, the context information corresponding to the customer service message, and a user role quality inspection prompt word template to obtain user role quality inspection prompt words; and performing user role perception analysis on the customer service message based on the user role quality inspection prompt words through the user role quality inspection model to obtain user role quality inspection results. The user role perception analysis includes at least one of user emotional state analysis, user comprehension deviation analysis, and user potential confusion point analysis; concatenating the customer service message, the context information corresponding to the customer service message, the user role quality inspection results, and the supervisor role quality inspection prompt word template to obtain... The process involves obtaining supervisor role quality inspection prompts and performing supervisor role evaluation analysis on the customer service message based on these prompts using the supervisor role quality inspection model. This results in a supervisor role quality inspection result, where the supervisor role evaluation analysis includes at least one of language expression analysis, tone improvement analysis, and supplementary explanation suggestion analysis. The customer service message, its corresponding context information, the supervisor role quality inspection result, and the quality inspector role quality inspection prompt template are then concatenated to obtain the quality inspector role quality inspection prompts. Based on these prompts, the quality inspector role quality inspection model performs a compliance check on the customer service message to obtain the quality inspector role quality inspection result. Finally, a quality inspection conclusion is drawn by combining the user role quality inspection result, the supervisor role quality inspection result, and the quality inspector role quality inspection result. This compliance check includes at least one of enterprise script standardization check, legal compliance requirement check, and service quality standard check.

4. The method according to claim 2, characterized in that, The step of selecting a target quality inspection model from among the multiple quality inspection models includes: obtaining the quality inspection rules of each role-based quality inspection model corresponding to the customer service agent terminal, and selecting a target quality inspection model from among the multiple quality inspection models according to the quality inspection rules. The quality inspection rules are determined as follows: based on the historical quality inspection results of each role-based quality inspection model corresponding to the customer service agent terminal, a historical quality inspection score for the customer service agent terminal is determined by each role-based quality inspection model; based on the historical quality inspection score, a baseline sampling rule for each role-based quality inspection model for the customer service agent terminal is determined; and based on the recent quality inspection results of each role-based quality inspection model for the customer service agent terminal, the baseline sampling rule is adjusted to obtain the quality inspection rules for the role-based quality inspection model corresponding to the customer service agent terminal. The recent quality inspection results include the quality inspection results of the most recent preset duration and / or the quality inspection results of the most recent preset number of times, and the quality inspection results corresponding to each role-based quality inspection model include a quality inspection score.

5. The method according to claim 4, characterized in that, The process of determining the baseline sampling rules for each role's quality inspection big model on the customer service agent end based on the historical quality inspection scores, and adjusting the baseline sampling rules based on the recent quality inspection results of each role's quality inspection big model on the customer service agent end to obtain the quality inspection rules for the role's quality inspection big model corresponding to the customer service agent end, includes: for any role's quality inspection big model, determining the historical average quality inspection score of the customer service agent end based on the historical quality inspection scores; determining the target score interval hit by the historical average quality inspection score among multiple preset score intervals corresponding to the role's quality inspection big model; and determining the sampling probability of the target score interval as the corner score based on the correspondence between the preset score intervals corresponding to the role's quality inspection big model and the quality inspection level. The quality inspection rule is based on the benchmark sampling rule corresponding to the large-scale quality inspection model. When the recent quality inspection results of the large-scale quality inspection model for any role are obtained, the recent average quality inspection score corresponding to the recent quality inspection result is calculated based on the quality inspection score in the recent quality inspection result. The difference between the recent average quality inspection score and the historical average quality inspection score corresponding to the large-scale quality inspection model for any role is calculated. The target probability adjustment value corresponding to the preset difference interval where the difference is located is determined according to the correspondence between the preset difference interval and the probability adjustment value. The sampling probability corresponding to the large-scale quality inspection model for any role is adjusted according to the target probability adjustment value. The benchmark sampling rule corresponding to the large-scale quality inspection model for any role is updated according to the adjusted sampling probability to obtain the quality inspection rule.

6. The method according to any one of claims 1 to 5, characterized in that, The method further includes: monitoring the input information of the customer service agent on the customer service dialogue interface; dynamically displaying the input information in the input information display area of ​​the quality inspection display window, and extracting customer service message fragments from the customer service agent based on the input information, wherein the quality inspection display window is above the customer service dialogue interface; performing quality inspection on the customer service message fragments using a multi-role quality inspection model to obtain the fragment quality inspection conclusions corresponding to the customer service message fragments; sending the fragment quality inspection conclusions to the customer service agent, so that the customer service agent displays the fragment quality inspection conclusions in the quality inspection information display area of ​​the quality inspection display window, wherein the input information in the input information display area and the quality inspection conclusions in the quality inspection information display area are displayed in a matched state, and when any input information is selected, the fragment quality inspection conclusions of the corresponding customer service message fragments are displayed differently, and when any input information is updated, the fragment quality inspection conclusions of the corresponding customer service message fragments are updated accordingly.

7. The method according to claim 6, characterized in that, The step of obtaining customer service messages from the customer service agent includes: listening to customer service messages sent by the customer service agent through the customer service dialogue interface; after sending the quality inspection conclusion to the customer service agent, the method further includes: displaying the quality inspection conclusion in the quality inspection information display area, deleting the fragment of quality inspection information, so that the customer service message and the quality inspection conclusion are displayed in a matched state, and displaying the corresponding quality inspection conclusion differently when any customer service message is selected.

8. A customer service message optimization device, characterized in that, The device includes: a message acquisition module for acquiring customer service messages from a customer service agent; a message quality inspection module for performing quality inspection on the customer service messages using a multi-role quality inspection model to obtain a quality inspection conclusion for the customer service messages, wherein the multi-role quality inspection model includes a user role quality inspection model, a supervisor role quality inspection model, and a quality inspector role quality inspection model; and a message optimization module for sending the quality inspection conclusion to the customer service agent so that the customer service agent can optimize the customer service messages based on the quality inspection conclusion.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

10. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

Images