Intelligent customer service training method and device based on dynamic skill mapping, equipment and medium

Abstract

The invention discloses an intelligent customer service training method and device based on dynamic skill mapping, equipment and a medium, and the method comprises the steps: extracting business scenes and skill chain association data in historical work orders of customer service personnel, and generating a fine adjustment data set; training a skill identification Agent based on the fine tuning data set, and outputting a dynamic skill gap matrix of the CSR; constructing a three-dimensional CSR portrait at least comprising a cognitive state, behavior characteristics and learning preferences; generating a priority training task sequence and a special training task by the path planning Agent; generating a personalized test case by a dialogue simulation Agent by using the priority training task sequence, the special training task and the three-dimensional CSR portrait; and distributing the personalized test case to a target customer service staff, and obtaining business index data of the target customer service staff to quantify a business skill value of the target customer service staff.

Description

Technical Field

[0001] This application relates to the field of intelligent training applications, and particularly to an intelligent customer service training method, device, equipment and medium with dynamic skill mapping. Background Art

[0002] Currently, customer service training mainly relies on traditional static courses, questionnaire evaluations and manual supervision. Although these methods can improve the skill level of customer service representatives (CSRs) to a certain extent, there are many deficiencies in actual applications, including: lagging skill diagnosis: traditional questionnaire evaluations and manual supervision cannot capture the skill shortfalls of CSRs in real business scenarios in real time, resulting in insufficient timeliness and pertinence of training. Rigid training content: static course libraries are difficult to dynamically adjust according to the basic capabilities and job objectives of CSRs, and cannot meet the personalized needs of different positions (such as pre-sales, after-sales, complaint handling). Delayed effect feedback: manual supervision and questionnaire evaluations are time-consuming and difficult to quantify the relevance between training effects and business metrics (such as customer satisfaction, first-time resolution rate), resulting in difficult evaluation and optimization of training effects. Lack of personalization: existing training methods cannot be customized according to the cognitive states, behavioral characteristics and learning preferences of CSRs, resulting in uneven training effects. Lack of dynamic update mechanism: existing training methods lack a dynamic update mechanism and cannot adjust training content and strategies in real time according to the actual performance of CSRs and business changes, resulting in the disconnection between training content and actual work requirements. These defects make it difficult for traditional customer service training methods to meet the enterprise's needs for efficient and accurate customer service training. Summary of the Invention

[0003] The purpose of this application is to provide an intelligent customer service training method, device, equipment and medium with dynamic skill mapping, so as to at least solve the problem that current traditional customer service training methods are difficult to meet the enterprise's needs for efficient and accurate customer service training.

[0004] To solve the above technical problems, this application provides an intelligent customer service training method with dynamic skill mapping, including: Extracting the business scenario and skill chain association data in the historical work orders of customer service representatives to generate a fine-tuning data set; Training a skill recognition Agent based on the fine-tuning data set and outputting a dynamic skill gap matrix of CSRs; According to the dynamic skill gap matrix, combining the simulation dialogue test results and real work order behavior data to construct a three-dimensional CSR portrait including at least cognitive state, behavioral characteristics and learning preferences; Based on the three-dimensional CSR portrait and the dynamic skill gap matrix, a path planning Agent generates a priority training task sequence and special training tasks; Generate personalized test cases by the dialogue simulation Agent using the prioritized training task sequence, special training tasks, and three-dimensional CSR portraits; Allocate the personalized test cases to the target customer service staff, and obtain the business metric data of the target customer service staff to quantify the business skill value of the target customer service staff.

[0005] Optionally, training the skill recognition Agent based on the fine-tuning dataset and outputting the dynamic skill gap matrix of CSRs includes: Disassemble the skill chain in the fine-tuning training set using the chain-of-thought annotation technique; Input the disassembled skill chain into the skill recognition Agent to train the skill recognition Agent; Train the skill recognition Agent to map the training objectives to sub-skills and generate the dynamic skill gap matrix of CSRs.

[0006] Optionally, before constructing the three-dimensional CSR portrait including at least cognitive state, behavioral characteristics, and learning preferences based on the dynamic skill gap matrix, it further includes: Configure the simulated dialogue test and analyze the results of the simulated dialogue test to evaluate the cognitive state of the CSR; Analyze the behavioral data in the real work orders to determine the behavioral characteristics.

[0007] Optionally, the configuration of the simulated dialogue test includes: Based on the skill defect weights in the dynamic skill gap matrix, retrieve specific texts in the knowledge base through the RAG technique to generate initial dialogue templates related to the specific texts; Inject industry-specific variables into the initial dialogue template according to the characteristics of the real business scenario to form a simulated dialogue test script adapted to the CSR portrait.

[0008] Optionally, after constructing the three-dimensional CSR portrait including at least cognitive state, behavioral characteristics, and learning preferences, it further includes: Periodically collect the behavioral data of the three-dimensional CSR in real work orders, and calculate the deviation value from the simulated dialogue test based on the behavioral data; When the deviation value exceeds the preset threshold, trigger the portrait correction mechanism and reallocate the priority weights of the cognitive state and behavioral characteristics in the three-dimensional CSR portrait.

[0009] Optionally, the method further includes: Conduct a simulation test on the test cases to obtain the simulation results; Input the simulation results into the KPI prediction model, and output the predicted value of the improvement amplitude of the customer satisfaction after training through the KPI prediction model; Perform feedback association between the predicted value and the business skill value of the target customer service staff to optimize the test cases.

[0010] Optionally, the step of inputting the simulation results into the KPI prediction model and outputting the predicted value of the improvement amplitude of the customer satisfaction after training through the KPI prediction model includes: Normalize the skill defect values in the dynamic skill gap matrix, and calculate the influence coefficients of each skill on the KPI in combination with the changes in customer satisfaction in the historical training data; Perform weighted summation of the influence coefficients and the current skill defect values, and output the quantitative predicted value of the improvement amplitude of the customer satisfaction after training.

[0011] To solve the above technical problems, the present application also provides an intelligent customer service training device for dynamic skill mapping, including: A data extraction module, configured to extract the business scenario and skill chain association data in the historical work orders of customer service staff, and generate a fine-tuning data set; A skill matrix module, configured to train a skill recognition Agent based on the fine-tuning data set, and output the dynamic skill gap matrix of CSR; A portrait construction module, configured to construct a three-dimensional CSR portrait including at least cognitive state, behavior characteristics, and learning preferences according to the dynamic skill gap matrix, in combination with the simulated dialogue test results and real work order behavior data; A task planning module, configured to generate a priority training task sequence and special training tasks by a path planning Agent based on the three-dimensional CSR portrait and the dynamic skill gap matrix; A use case generation module, configured to generate personalized test cases by a dialogue simulation Agent using the priority training task sequence, special training tasks, and three-dimensional CSR portrait; A training configuration module, configured to allocate the personalized test cases to the target customer service staff, and obtain the business index data of the target customer service staff to quantify the business skill value of the target customer service staff.

[0012] To solve the above technical problems, the present application also provides a computer device, including a memory and a processor, where computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the above-mentioned intelligent customer service training method for dynamic skill mapping.

[0013] To solve the above technical problems, the present application also provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors are caused to execute the steps of the above-described intelligent customer service training method for dynamic skill mapping.

[0014] The beneficial effects of the embodiments of the present application are as follows: By extracting the business scenario and skill chain association data in the historical work orders of customer service personnel, a fine-tuning data set is generated; Based on the fine-tuning data set, a skill recognition Agent is trained to output the dynamic skill gap matrix of CSRs; According to the dynamic skill gap matrix, combined with the simulation dialogue test results and real work order behavior data, a three-dimensional CSR portrait including at least cognitive state, behavior characteristics, and learning preferences is constructed; Based on the three-dimensional CSR portrait and the dynamic skill gap matrix, a path planning Agent generates a priority training task sequence and special training tasks; Using the priority training task sequence, special training tasks, and three-dimensional CSR portrait, a dialogue simulation Agent generates personalized test cases; The personalized test cases are assigned to the target customer service personnel, and the business metric data of the target customer service personnel is obtained to quantify the business skill value of the target customer service personnel; Through the job portrait and skill dynamic mapping technology, the skill shortboards of customer service personnel (CSRs) in different business scenarios can be accurately identified, solving the pain point of lagging skill diagnosis in traditional training and realizing real-time dynamic adjustment of training content. Secondly, the three-dimensional CSR portrait constructed by the multi-modal learner modeling technology comprehensively covers cognitive state, behavior characteristics, and learning preferences, making the training more targeted and personalized, and effectively improving the training effect. In addition, the virtual-real linkage training resource generation technology combined with the business simulation system can generate highly simulated personalized test cases, helping CSRs accumulate experience in the simulation environment and enhancing their ability to handle complex business scenarios. Finally, by quantitatively evaluating the skill value, the system can provide real-time feedback on the training effect and predict the improvement amplitude of the business KPI, solving the problem of delayed feedback on the training effect in traditional training and providing an efficient, accurate, and quantifiable solution for the customer service training of enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of the embodiments in conjunction with the accompanying drawings, where: Figure 1 is a schematic diagram of the basic process of the intelligent customer service training method for dynamic skill mapping in a specific embodiment of the present application; Figure 2 is a schematic diagram of the basic structure of the intelligent customer service training device for dynamic skill mapping in a specific embodiment of the present application; Figure 3 is a basic structural block diagram of a computer device in a specific embodiment of the present application. Detailed implementation manners

[0016] The embodiments of the present application will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and should not be construed as a limitation to the present application.

[0017] Those skilled in the art of the present technology can understand that, unless specifically stated otherwise, the singular forms "a", "an", "the", and "said" used herein may also include the plural forms. It should be further understood that the term "comprising" used in the specification of the present application means the presence of the described features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0018] Those skilled in the art of the present technology can understand that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as the general understanding of those of ordinary skill in the art to which the present application belongs. It should also be understood that terms such as those defined in a general dictionary should be understood as having a meaning consistent with the meaning in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless specifically defined as here.

[0019] Those skilled in the art can understand that the "terminal" used herein includes both a device with a wireless signal receiver that only has the ability to receive and no ability to transmit, and a device with receiving and transmitting hardware that has the receiving and transmitting hardware capable of performing two-way communication on a two-way communication link. Such devices may include: cellular or other communication devices with a single-line display or a multi-line display or cellular or other communication devices without a multi-line display; PCS (Personal Communications Service), which can combine voice, data processing, fax, and / or data communication capabilities; PDA (Personal Digital Assistant), which may include a radio frequency receiver, a pager, Internet / intranet access, a web browser, a notepad, a calendar, and / or a GPS (Global Positioning System) receiver; conventional laptop and / or palm computers or other devices with and / or including a radio frequency receiver. The "terminal" used herein can be portable, transportable, installed in a vehicle (air, sea, and / or land), or suitable for and / or configured to operate locally and / or in a distributed manner at any other location on the earth and / or in space. The "terminal" used herein can also be a communication terminal, an Internet access terminal, a music / video playback terminal, such as a PDA, a MID (Mobile Internet Device), and / or a mobile phone with music / video playback functions, or can also be devices such as a smart TV and a set-top box.

[0020] The hardware referred to by names such as "server", "client", and "service node" in this application is essentially an electronic device with the equivalent capabilities of a personal computer, which is a hardware device with the necessary components disclosed by the von Neumann principle, including a central processing unit (including an arithmetic unit and a controller), a memory, an input device, and an output device. The computer program is stored in its memory, and the central processing unit loads the program stored in the external memory into the memory for execution, executes the instructions in the program, and interacts with the input / output devices to complete specific functions.

[0021] It should be noted that the concept of "server" in this application can similarly be extended to the case of server clusters. According to the network deployment principles understood by those skilled in the art, the servers should be logically divided. Physically, these servers can either be independent of each other but can be invoked through interfaces, or integrated into a physical computer or a set of computer clusters. Those skilled in the art should understand this flexibility and should not be restricted by this when implementing the network deployment method of this application.

[0022] One or several technical features of this application, unless expressly specified, can either be deployed on the server and accessed by the client remotely invoking the online service interface provided by the server, or directly deployed and run on the client for access.

[0023] The AI models cited or possibly cited in this application, unless expressly specified, can either be deployed on a remote server and remotely invoked on the client, or deployed on a client capable of handling the device and directly invoked. In some embodiments, when it runs on the client, its corresponding intelligence can be obtained through transfer learning to reduce the requirements for the client's hardware operating resources and avoid over-occupying the client's hardware operating resources.

[0024] All kinds of data involved in this application, unless expressly specified, can either be remotely stored on the server or stored on the local terminal device, as long as it is suitable for being invoked by the technical solution of this application.

[0025] Those skilled in the art should be aware of this: Although the various methods of this application are described based on the same concept and thus show commonality with each other, unless otherwise specified, these methods can all be executed independently. Similarly, for each embodiment disclosed in this application, they are all proposed based on the same inventive concept. Therefore, for concepts with the same expression, as well as concepts that are only appropriately transformed for convenience although the concept expressions are different, they should be equivalently understood.

[0026] For each embodiment to be disclosed in this application, unless expressly stated that there is a mutually exclusive relationship between them, otherwise, the relevant technical features involved in each embodiment can be cross-combined to flexibly construct new embodiments, as long as this combination does not deviate from the creative spirit of this application and can meet the needs in the prior art or solve certain deficiencies in the prior art. Those skilled in the art should be aware of this flexibility.

[0027] Please refer to Figure 1 , Figure 1 which is a schematic diagram of the basic process of the intelligent customer service training method for dynamic skill mapping in this embodiment.

[0028] As Figure 1 shown, it includes: S1100, extracting business scenarios and skill chain-related data from historical work orders of customer service personnel to generate a fine-tuning data set; This embodiment can be applied to customer service personnel training scenarios in various fields such as finance, e-commerce, insurance, education, medical care, law, and hotel management. In this embodiment, by configuring a customer service training system, it is used to customize and implement a systematic and professional training plan for customer service personnel. The system first obtains the historical work order data of customer service personnel from the enterprise database. The historical work order data includes complete conversation texts, customer problem descriptions, customer service personnel's response content, and final solutions. In order to extract the associated data between business scenarios and skill chains, the system uses natural language processing (NLP) technology to analyze the conversation text. Through pre-trained language models (such as BERT or Transformer architecture), the system can identify key business scenarios in the conversation (such as "product consultation", "refund processing", "complaint processing", etc.). At the same time, the system can also extract the skill chain corresponding to each business scenario in combination with job descriptions (JD) and quality inspection records. For example, in the "refund processing" scenario, the skill chain may include "payment system operation", "customer complaint appeasement speech", and "platform policy interpretation". Through annotation technology (such as thinking chain annotation CoT), the system maps these skill chains to business scenarios one by one to generate a fine-tuning data set. This dataset not only contains the correspondence between business scenarios and skill chains, but also records in detail the specific manifestations of each skill in the conversation through annotations, providing high-quality training samples for the subsequent training of skill recognition agents.

[0029] It should be pointed out that the system of this embodiment can also introduce multi-source data fusion technology. In addition to historical work order data, the system can also extract relevant information from the customer service staff's performance evaluation records, customer feedback, and training records. For example, the performance evaluation records may contain the performance scores of customer service staff in specific business scenarios. These scores can serve as a supplement to the skill chain associated data to help the system more accurately evaluate the actual skill level of customer service staff. Through the fusion of multi-source data, the generated fine-tuning data set will be more comprehensive and accurate, and can better reflect the skill requirements and performance of customer service staff in different business scenarios.

[0030] It should be pointed out that the "business scenario" in this embodiment refers to the specific customer problems or request types encountered by customer service personnel in their work, such as "product consultation", "order inquiry", "after-sales service", etc. The "skill chain" refers to a series of skill combinations required to complete a specific business scenario. For example, in the "after-sales service" scenario, the skill chain may include "product knowledge", "communication skills", "problem-solving ability", etc. By associating business scenarios with skill chains, the system can provide customer service personnel with more targeted training content to help them improve their skills in specific scenarios.

[0031] S1200. Train a skill recognition Agent based on the fine-tuning dataset and output a dynamic skill gap matrix for CSR. After extracting the business scenario and skill chain association data in the historical work orders of customer service staff to generate a fine-tuning dataset, train a skill recognition Agent based on the fine-tuning dataset and output a dynamic skill gap matrix for CSR. Specifically, the system uses the fine-tuning dataset generated in step S1100 to train the skill recognition Agent. The core of this Agent is a deep learning-based natural language processing model, and its goal is to identify the usage of specific skills from the conversation text of customer service staff. Specifically, the system uses a pre-trained language model with a Transformer architecture (such as BERT or GPT) and fine-tunes it on this basis to enable it to understand the corresponding relationship between business scenarios and skill chains in customer service conversations.

[0032] During the training process, the system takes the conversation text in the fine-tuning dataset as input and the labeled skill chain as the output label, and trains the Agent through supervised learning. To improve the generalization ability and accuracy of the model, the system uses data augmentation techniques, such as generating more diverse training samples by synonym replacement, sentence restructuring, etc. In addition, the system also introduces a multi-task learning mechanism to enable the Agent to simultaneously learn to identify the usage frequency, proficiency level of skills, and their matching degree with business scenarios. After the training is completed, the skill recognition Agent can analyze the real-time conversations of customer service staff and output their performance scores on each skill. The system compares these scores with the preset job skill standards to generate a dynamic skill gap matrix. This matrix not only shows the gap between the current skills of customer service staff and the target skills, but also can be dynamically adjusted according to changes in business scenarios, providing a basis for subsequent personalized training.

[0033] It should be noted that the system in this embodiment can further optimize the training process of the skill recognition Agent. For example, introduce transfer learning technology to transfer the existing general language model knowledge to the customer service field, reducing training time and resource consumption. At the same time, the system can adopt a reinforcement learning mechanism to enable the Agent to continuously try and optimize in a simulated environment, improving its adaptability to complex business scenarios. In addition, the system can also introduce an expert knowledge base to incorporate the experience and knowledge of domain experts into the training process, further improving the accuracy of the Agent.

[0034] It should be noted that the "skill recognition Agent" in this embodiment refers to a natural language processing model based on deep learning, whose function is to identify the specific skills used by customer service personnel from customer service dialogue texts. The "dynamic skill gap matrix" refers to a matrix that is updated in real time, which shows the gap between the current skill level of customer service personnel and the target skills of the position. This matrix can be dynamically adjusted according to changes in business scenarios, providing accurate guidance for personalized training. In this way, the system can provide a customized training plan for each customer service personnel, helping them quickly improve their skill levels and better meet the job requirements.

[0035] S1300. According to the dynamic skill gap matrix, combined with the simulation dialogue test results and real work order behavior data, construct a three-dimensional CSR portrait that at least includes cognitive state, behavior characteristics, and learning preferences; After training the skill recognition Agent based on the fine-tuning dataset and outputting the dynamic skill gap matrix of CSRs, according to the dynamic skill gap matrix, combined with the simulation dialogue test results and real work order behavior data, construct a three-dimensional CSR portrait that at least includes cognitive state, behavior characteristics, and learning preferences. Specifically, first, the system uses the dynamic skill gap matrix output by the skill recognition Agent to analyze the gap between the performance of CSRs in various skills and the target skills, so as to evaluate their cognitive state. For example, through simulation dialogue tests, the system can measure the response accuracy rate of CSRs to policy terms, the compliance of the conversation scripts, etc. These data reflect the understanding and mastery of business knowledge by CSRs. At the same time, the system extracts behavior characteristic data from real work orders, such as response speed, transfer rate, repeated consultation rate, etc. These data are processed by the data analysis module to evaluate the behavior performance of CSRs in actual work. For example, the response speed can reflect the efficiency of CSRs, while the transfer rate can reveal their ability to handle complex problems. In addition, the system also identifies their learning preferences by analyzing the behavior of CSRs on the training platform, such as course click stream, learning time distribution, etc. For example, by analyzing the number of clicks and the stay time of CSRs on video, graphic or scenario simulation courses, the system can determine which learning method they prefer. To achieve the above functions, the system adopts multi-modal data fusion technology to integrate text data, behavior data, and preference data. Through machine learning algorithms, the system can automatically identify the key features in the CSR portrait and dynamically update the portrait according to real-time data. Finally, the generated three-dimensional CSR portrait can not only comprehensively reflect the current state of CSRs, but also provide accurate basis for subsequent personalized training.

[0036] It should be noted that the system of this embodiment can be further optimized for the construction of the CSR portrait by introducing deep learning technology. For example, the neural network is used to perform semantic analysis on the simulation dialogue test results to more accurately evaluate the cognitive state of the CSR. At the same time, the system can introduce time series analysis technology to dynamically monitor the behavior data in the real work orders and timely detect changes in the CSR behavior patterns. In addition, the system can also allow the CSR to participate in the portrait calibration process through the user feedback mechanism to improve the accuracy and practicality of the portrait.

[0037] It should be noted that the "cognitive state" refers to the CSR's understanding and mastery of business knowledge, which is evaluated through simulation dialogue tests and skill gap matrices. The "behavior characteristics" refer to the CSR's performance in actual work, such as response speed, transfer rate, etc., which are obtained by analyzing real work order data. The "learning preference" refers to the CSR's tendency towards different learning methods during the learning process, which is obtained by analyzing its behavior data on the training platform. By constructing a three-dimensional CSR portrait, the system can provide personalized training programs for each CSR to help them improve their business capabilities.

[0038] S1400. Based on the three-dimensional CSR portrait and the dynamic skill gap matrix, the path planning Agent generates a priority training task sequence and special training tasks; After constructing a three-dimensional CSR portrait that at least includes cognitive state, behavioral characteristics, and learning preferences based on the dynamic skills gap matrix, combined with the simulated conversation test results and real work order behavior data, based on the three-dimensional CSR portrait and the dynamic skills gap matrix, the path planning Agent generates a priority training task sequence and special training tasks. Specifically, the core of the path planning Agent in this embodiment is an intelligent decision-making model based on reinforcement learning, and its goal is to plan the optimal training path according to the current skill level of the CSR and the job target. The path planning Agent first analyzes the cognitive state, behavioral characteristics, and learning preferences in the CSR portrait, and combines the skill shortboards identified in the dynamic skills gap matrix to determine the training priority. For example, if there is a large gap in the "emotion recognition" skill of the CSR in the "customer complaint handling" scenario and their learning preference is video learning, then the path planning Agent will give priority to arranging relevant video training courses. When generating the training task sequence, the system adopts a hierarchical planning strategy. First, for the key skill shortboards, short-term special training tasks are generated to quickly improve the CSR's ability in specific scenarios. Subsequently, combined with the long-term career development path of the CSR, a long-term skill improvement sequence is generated. For example, for newly recruited CSRs, the system will first arrange basic "product knowledge" and "communication skills" training, and then gradually introduce more complex "customer relationship management" and "problem-solving ability" training. To ensure the effectiveness of the training tasks, the path planning Agent will monitor the training progress and performance of the CSR in real time. If the CSR performs poorly in a special training task, the system will automatically adjust the task difficulty or add relevant training modules. In addition, the system will also dynamically adjust the training task sequence according to the CSR's performance in real work orders to ensure that the training content is closely aligned with the actual work requirements.

[0039] It should be noted that the system in this embodiment can further optimize the decision-making process of the path planning Agent by introducing a multi-agent cooperation mechanism. For example, multiple Agents can be responsible for task planning in different skill fields respectively, and generate a more comprehensive training path through collaborative work. At the same time, the system can introduce a user feedback mechanism to allow the CSR to evaluate the difficulty and practicality of the training tasks, so as to optimize the subsequent training content.

[0040] It should be noted that the "path planning Agent" in this embodiment refers to an intelligent decision-making model based on reinforcement learning, whose role is to generate personalized training paths according to the skill gaps and portrait features of CSRs. The "priority training task sequence" refers to an ordered list of short-term and long-term training tasks generated according to the skill shortages of CSRs and job requirements. The "special training task" refers to a targeted training module designed for the skill shortages of CSRs in specific business scenarios. Through these technical means, the system can provide efficient and personalized training programs for CSRs to help them quickly improve their business capabilities.

[0041] S1500. Generate personalized test cases by the dialogue simulation Agent using the priority training task sequence, special training tasks, and three-dimensional CSR portraits; After the path planning Agent generates the priority training task sequence and special training tasks based on the three-dimensional CSR portrait and the dynamic skill gap matrix, the dialogue simulation Agent generates personalized test cases using the priority training task sequence, special training tasks, and three-dimensional CSR portraits. The system generates personalized test cases through the dialogue simulation Agent to help customer service representatives (CSRs) improve their skills in a simulated environment. In one embodiment, the dialogue simulation Agent determines the difficulty and complexity of the test cases based on the cognitive state and behavioral characteristics in the CSR portrait. For example, if the CSR has weak emotion recognition skills in the "customer complaint handling" scenario, the Agent will generate customer conversations containing emotional language to simulate real complaint scenarios. At the same time, the Agent will adjust the form of the test cases according to the learning preferences of the CSR. For example, if the CSR prefers video learning, the Agent can generate test cases containing video conversations. When generating test cases, the Agent will extract accurate policy terms and product information from the enterprise knowledge base to ensure the business accuracy of the test cases. In addition, the Agent will also combine with the business simulation system to simulate different types of customers, such as customers with dialect accents or complex complaint combinations, to increase the diversity and challenge of the test cases. The test cases generated by the dialogue simulation Agent not only include dialogue texts but also can contain context information, such as customer background, historical order records, etc., to help CSRs better understand customer needs. After the test cases are generated, the system will dynamically adjust the test difficulty according to the real-time performance of the CSR to ensure that the test process is challenging and in line with the current skill level of the CSR.

[0042] It should be noted that the system of this embodiment can further enhance the realism and interactivity of test cases by introducing multi-modal interaction technology. For example, by combining speech recognition and synthesis technologies, the CSR can communicate with the simulated customer through speech to improve their communication skills. At the same time, the system can introduce sentiment analysis technology to evaluate the CSR's emotional responses in the conversation in real time, helping them better manage their emotions.

[0043] It should be noted that the "dialogue simulation Agent" in this embodiment refers to a natural language generation model based on deep learning, whose function is to generate personalized test cases according to the portrait of the CSR and the training tasks. "Personalized test cases" refer to simulated dialogue scenarios generated according to the CSR's skill deficiencies and learning preferences, aiming to help the CSR improve specific skills. Through these technical means, the system can provide a highly realistic training environment for the CSR to better handle various customer problems in actual work.

[0044] S1600. Assign the personalized test case to the target customer service staff, and obtain the business metric data of the target customer service staff to quantify the business skill value of the target customer service staff.

[0045] After generating personalized test cases by the dialogue simulation Agent using the priority training task sequence, special training tasks, and three-dimensional CSR portraits, the personalized test cases are assigned to the target customer service staff, and the business metric data of the target customer service staff is obtained to quantify the business skill value of the target customer service staff. The system assigns the personalized test cases generated in step S1500 to the target customer service staff (CSR) and quantitatively evaluates the business skill value of the CSR through a series of technical means. First, the system pushes the test cases to the CSR through an automated assignment mechanism and records various data during their processing, including key business metrics such as response time, problem-solving efficiency, and customer satisfaction. Specifically, to quantify the business skill value of the CSR, the system adopts a multi-dimensional evaluation model. This model combines the performance of the CSR in the test cases and its historical data in real work orders to generate a comprehensive score. For example, the system can analyze the dialogue text of the CSR through natural language processing technology to evaluate its compliance of speech, problem-solving ability, and emotion management ability. At the same time, the system will also calculate its efficiency metrics based on the speed and accuracy of the CSR in processing test cases. In addition, the system introduces a business KPI prediction model to conduct correlation analysis between the performance of the CSR in the test cases and actual business metrics (such as customer satisfaction, first-time resolution rate, etc.). Through machine learning algorithms, the system can predict the improvement amplitude of the business performance of the CSR after training. For example, if the CSR performs well in the test cases, the system can predict the probability of improvement in customer satisfaction in actual work. The system also supports a real-time feedback mechanism. After the CSR completes the test cases, they can immediately obtain a detailed skill evaluation report. The report not only shows the comprehensive score of the CSR but also provides specific improvement suggestions and subsequent training directions. In this way, the system can help the CSR quickly understand their own skill level and continuously optimize the training effect.

[0046] Furthermore, in some scenarios, the system can further improve the accuracy and objectivity of skill evaluation by introducing an external evaluation mechanism. For example, inviting domain experts to conduct manual reviews on the performance of the CSR in the test cases and adjusting the system evaluation results in combination with expert opinions. At the same time, the system can introduce a customer feedback mechanism to let simulated customers evaluate the service quality of the CSR and further improve the evaluation system.

[0047] It should be noted that the "business indicator data" in this embodiment refers to the key data generated by the CSR during the process of processing test cases, such as response time, problem-solving efficiency, customer satisfaction, etc., which are used to quantitatively evaluate the business skill level of the CSR. The "business skill value" refers to the comprehensive skill score of the CSR calculated through a multi-dimensional evaluation model, which reflects the ability level of the CSR in a specific business scenario. In this way, the system can provide accurate skill evaluation and improvement suggestions for the CSR to help them better handle customer problems in actual work.

[0048] In a specific implementation scenario, assume that an e-commerce enterprise needs to improve the ability of its customer service team to handle order inquiries and after-sales issues during the "Double Eleven Promotion" period. First, the system extracts the historical ticket data of customer service representatives (CSRs) from the enterprise database, including customer consultation conversations, order processing records, and after-sales feedback, etc. Through natural language processing technology, the system analyzes these data, extracts the corresponding relationships between business scenarios (such as "order query", "return and exchange processing") and required skill chains (such as "order system operation", "customer appeasement skills"), and generates a fine-tuning data set. Based on the fine-tuning data set, the system trains a skill recognition Agent. The Agent analyzes the performance of the CSR in historical tickets, identifies their performance in various skills, and compares it with the job skill standards to generate a dynamic skill gap matrix. For example, the system finds that a certain CSR has a large gap in the skill of "return and exchange policy interpretation". The system combines the dynamic skill gap matrix, the results of simulated dialogue tests, and real ticket behavior data to construct a three-dimensional portrait of the CSR. The portrait shows that the CSR's cognitive state in the "return and exchange processing" scenario is "insufficient policy understanding", the behavioral characteristic is "slow response speed", and the learning preference is "video tutorials". The path planning Agent generates a priority training task sequence and special training tasks according to the CSR portrait and the skill gap matrix. The system arranges a special training task of "return and exchange policy interpretation" for this CSR and recommends that they first learn the policy knowledge through video courses and then conduct conversation skill training. The dialogue simulation Agent generates personalized test cases according to the priority training task sequence and the CSR portrait. The test case simulates a return and exchange scenario during the "Double Eleven Promotion" period, where the customer is emotional and the CSR needs to accurately interpret the policy and appease the customer's emotions in the conversation. After the CSR completes the test case, the system analyzes their conversation text through natural language processing technology, evaluates the compliance of their conversation and problem-solving ability, and quantifies their business skill value in combination with business indicators such as response time. The system generates a detailed skill evaluation report, showing the improvement of the CSR in the "return and exchange policy interpretation" skill and predicting the improvement range of customer satisfaction in actual work. Through this complete training process, the CSR can handle customer problems more efficiently during the "Double Eleven Promotion" period, improving customer satisfaction and business efficiency.

[0049] In the above embodiments, by extracting the business scenario and skill chain association data in the historical work orders of customer service staff, a fine-tuning data set is generated; based on the fine-tuning data set, a skill recognition Agent is trained to output the dynamic skill gap matrix of CSRs; according to the dynamic skill gap matrix, combined with the simulation dialogue test results and real work order behavior data, a three-dimensional CSR portrait including at least cognitive state, behavior characteristics, and learning preferences is constructed; based on the three-dimensional CSR portrait and the dynamic skill gap matrix, a path planning Agent generates a priority training task sequence and special training tasks; using the priority training task sequence, special training tasks, and three-dimensional CSR portrait, a dialogue simulation Agent generates personalized test cases; the personalized test cases are assigned to the target customer service staff, and the business index data of the target customer service staff is obtained to quantify the business skill value of the target customer service staff; through the job portrait and skill dynamic mapping technology, the skill shortboards of customer service staff (CSRs) in different business scenarios can be accurately identified, solving the pain point of lagging skill diagnosis in traditional training and realizing the real-time dynamic adjustment of training content. Secondly, the three-dimensional CSR portrait constructed by the multi-modal learner modeling technology comprehensively covers cognitive state, behavior characteristics, and learning preferences, making the training more targeted and personalized and effectively improving the training effect. In addition, the virtual-real linkage training resource generation technology combined with the business simulation system can generate highly simulated personalized test cases to help CSRs accumulate experience in the simulation environment and enhance their ability to handle complex business scenarios. Finally, by quantitatively evaluating the skill value, the system can provide real-time feedback on the training effect and predict the improvement range of business KPIs, solving the problem of delayed feedback on the training effect in traditional training and providing an efficient, accurate, and quantifiable solution for the customer service training of enterprises.

[0050] In some embodiments, S1200 training the skill recognition Agent based on the fine-tuning data set to output the dynamic skill gap matrix of CSRs includes: S1211. Disassembling the skill chain in the fine-tuning training set by using the thought chain annotation technology; In this embodiment, in the process of training the skill recognition Agent based on the fine-tuning dataset and outputting the dynamic skill gap matrix of CSR, first, the thought chain annotation technique is used to disassemble the skill chain in the fine-tuning training set. The core of this process is to decompose complex business scenarios into a series of specific skill points so that the skill recognition Agent can more accurately identify and evaluate the skill levels of CSRs. First, the system extracts the correspondence between business scenarios and skill chains from the fine-tuning dataset. For example, in the business scenario of "handling customer complaints", the skill chain may include multiple skill points such as "emotion recognition", "problem localization", and "solution provision". Through the thought chain annotation technique, the system disassembles and annotates these skill points in a logical order. For example, it is annotated as "emotion recognition → problem localization → solution provision", forming a clear skill disassembly trajectory. During the annotation process, the system uses natural language processing techniques to analyze the dialogue text and identify the specific manifestation forms of each skill point in the dialogue. For example, in the skill point of "emotion recognition", the annotator will mark how the CSR in the dialogue recognizes the customer's emotional state (such as through tone, keywords, etc.). This annotation method not only helps the system understand the logical order of skill points but also provides rich training samples for the skill recognition Agent.

[0051] It should be noted that, in order to improve the accuracy and efficiency of annotation, the system introduces a semi-automated annotation tool. The annotator can quickly locate the key skill points in the dialogue text through the tool and perform annotations. At the same time, the system will automatically learn the annotation rules based on the annotated data to assist the annotator in completing subsequent annotation tasks. In this way, the system can quickly generate high-quality annotated data, providing a solid foundation for the training of the skill recognition Agent.

[0052] It should be noted that the system can also involve domain experts to further improve the accuracy and practicality of skill chain disassembly. Domain experts can refine and optimize the skill chain according to actual business needs. For example, for the skill point of "solution provision", the expert can further disassemble it into sub-skill points such as "policy interpretation", "operation guidance", and "follow-up", making the skill chain more detailed and specific.

[0053] In addition, the system can adopt a multi-round annotation and verification mechanism to ensure the quality of the annotated data. After the annotator completes the preliminary annotation, the system will randomly select some of the annotated data for cross-verification. If inconsistent annotations are found, the system will organize the annotators to discuss and correct them until the annotation results are consistent. In this way, the system can generate high-quality annotated data, providing reliable input for the training of the skill recognition Agent.

[0054] S1212. Input the disassembled skill chain into the skill recognition Agent to train the skill recognition Agent; After disassembling the skill chain in the fine-tuning training set using the chain-of-thought annotation technique, this embodiment further inputs the disassembled skill chain into a skill recognition Agent to train the skill recognition Agent. The core of the skill recognition Agent is a natural language processing model based on deep learning, and its purpose is to accurately identify the specific skill points demonstrated by customer service representatives (CSRs) in the dialogue text by learning the annotated skill chain data. In this embodiment, the skill recognition Agent adopts a pre-trained language model with a Transformer architecture (such as BERT or GPT) and performs fine-tuning on this basis. The fine-tuning process includes the following key steps: a. Data preprocessing: Clean and format the annotated skill chain data to ensure the quality of the input data. For example, convert the dialogue text and the corresponding skill labels into a format acceptable to the model.

[0055] b. Model training: Input the preprocessed data into the pre-trained model and train the model through supervised learning. The goal of the model is to predict the corresponding skill labels based on the dialogue text. For example, when inputting the dialogue text "The customer is emotional, and the CSR relieves the customer's emotion through soothing words", the model needs to output skill labels such as "emotion recognition" and "soothing skills".

[0056] c. Loss function optimization: Use the cross-entropy loss function to measure the difference between the model's predicted values and the true labels, and optimize the model parameters through the backpropagation algorithm to improve the model's accuracy and generalization ability.

[0057] d. Model validation and adjustment: Evaluate the model through the validation set, and adjust the hyperparameters of the model according to the evaluation results, such as the learning rate, batch size, etc. At the same time, introduce an early stopping mechanism to prevent the model from overfitting.

[0058] During the training process, the system also introduces data augmentation techniques to generate more diverse training samples through methods such as synonym replacement and sentence restructuring, further enhancing the robustness of the model. In addition, the system can adopt a multi-task learning mechanism to enable the skill recognition Agent to simultaneously learn the usage frequency, proficiency level of skills, and their matching degree with business scenarios, thereby improving the comprehensive performance of the model.

[0059] Furthermore, the system can reduce the training time and resource consumption by introducing transfer learning techniques to transfer the existing general language model knowledge to the customer service field. For example, use a language model pre-trained on large-scale text data (such as BERT) and perform fine-tuning for specific tasks in the customer service field. This can make full use of the general language knowledge of the pre-trained model and at the same time adapt it to the specific needs of the customer service scenario.

[0060] In addition, the system can adopt a reinforcement learning mechanism to enable the skill recognition Agent to continuously trial and error and optimize in a simulated environment. For example, by interacting with a simulated customer service dialogue system, the Agent can adjust its skill recognition strategy according to the feedback, thereby improving its adaptability in complex business scenarios. Through these technical means, the system can further enhance the performance of the skill recognition Agent, making it more accurate and efficient in practical applications.

[0061] S1213. Train the skill recognition Agent to map the training objectives to sub-skills and generate a dynamic skill gap matrix for CSRs.

[0062] After inputting the disassembled skill chain into the skill recognition Agent, train the skill recognition Agent to map the training objectives to sub-skills and generate a dynamic skill gap matrix for CSRs. Specifically, first, the system decomposes these objectives into a series of sub-skills according to the training objectives set by the enterprise (such as "improving complaint handling efficiency" or "enhancing customer satisfaction"). For example, "improving complaint handling efficiency" can be decomposed into specific skills such as "emotion recognition", "problem location", and "solution provision". The skill recognition Agent can identify the specific manifestation forms of these sub-skills in the dialogue text by learning and fine-tuning the skill chains annotated in the dataset. Next, the Agent conducts a comparative analysis of the performance of CSRs in historical work orders and simulated dialogue tests with these sub-skills. The system extracts the key information of CSRs in the dialogue through natural language processing technology and matches it with the skill labels. For example, if a CSR fails to effectively recognize the customer's emotion when handling a customer complaint, the Agent will mark the deficiency of this CSR in the "emotion recognition" skill. Based on this comparative analysis, the system generates a dynamic skill gap matrix for CSRs. This matrix is presented in a table form, with the horizontal axis representing various sub-skills and the vertical axis representing the comparison between the current skill level and the target skill level of CSRs. For example, if the current level of a CSR in the "emotion recognition" skill is 30% while the target level is 80%, this skill will show a relatively large gap value in the gap matrix. The system will also dynamically adjust the gap matrix according to the performance of CSRs in different business scenarios to reflect the changes in their skill levels. Finally, the system combines the dynamic skill gap matrix with the CSR portrait to provide a basis for subsequent personalized training path planning. In this way, the system can provide a customized training plan for each CSR, helping them quickly improve their skill levels and better meet the job requirements.

[0063] To further optimize the effect, the system of this embodiment can further optimize the generation process of the skill gap matrix by introducing machine learning algorithms. For example, by using clustering analysis techniques, CSRs are grouped according to their skill levels and job requirements, and a more targeted skill gap matrix is generated for each group. This can improve the utilization efficiency of training resources and ensure that the training content highly matches the actual needs of CSRs.

[0064] In some embodiments, before constructing a three-dimensional CSR portrait including at least cognitive state, behavioral characteristics, and learning preferences based on the dynamic skill gap matrix, in combination with the results of simulated dialogue tests and real work order behavior data, it further includes: S1311. Configure simulated dialogue tests and analyze the results of the simulated dialogue tests to evaluate the cognitive state of CSRs; In this embodiment, the system evaluates the cognitive state of customer service representatives (CSRs) by configuring simulated dialogue tests. The simulated dialogue test is a virtual dialogue environment designed based on the actual business scenarios of the enterprise, aiming to evaluate the CSRs' understanding and application abilities of business knowledge. First, the system designs simulated dialogue test cases for multiple business scenarios according to the business requirements and job skill requirements of the enterprise. These test cases cover common scenarios such as customer consultations, complaint handling, and order inquiries, and embed key business knowledge points and skill points. For example, in the "product consultation" scenario, the test case may include questions about product functions, prices, usage methods, etc., to evaluate the CSRs' mastery of product knowledge. Next, the system pushes the simulated dialogue tests to the CSRs and records their performance during the tests. Through natural language processing technology, the system analyzes the dialogue texts of the CSRs and evaluates the accuracy, completeness, and logic of their answers. For example, the system can check whether the CSRs can accurately answer customers' questions about product functions, whether they can provide clear solutions, and whether they can effectively appease customers' emotions. The system also introduces a scoring mechanism to score the performance of CSRs in each test case. The scoring criteria include the accuracy of answers (such as the correct citation of policy terms), the compliance of the conversation language (such as whether the standard conversation language template is used), and the efficiency of problem-solving (such as whether the problem can be quickly located and a solution can be provided). Through comprehensive scoring, the system can quantitatively evaluate the cognitive state of CSRs and provide data support for subsequent portrait construction. In addition, the system can also generate a detailed feedback report based on the results of the simulated dialogue tests, indicating which skill points the CSRs perform well in and which skill points need further improvement. This feedback not only helps CSRs understand their own knowledge shortfalls but also provides targeted training suggestions for training managers.

[0065] It should be noted that the system of this embodiment can further enhance the realism and interactivity of the simulated dialogue test by introducing a multi-modal test environment. For example, by combining speech recognition and synthesis technologies, the CSR can communicate with virtual customers through speech to simulate a real telephone customer service scenario. At the same time, the system can introduce emotion analysis technology to evaluate the CSR's emotional response in the dialogue in real time, helping them better manage their emotions.

[0066] S1312. Analyze the behavior data in real work orders to determine the behavior characteristics.

[0067] After configuring the simulated dialogue test and analyzing the results of the simulated dialogue test to evaluate the CSR's cognitive state, analyze the behavior data in real work orders to determine the behavior characteristics. Specifically, the real work order data is a record of the CSR's interaction with customers in actual work, containing rich behavior information, which can reflect the CSR's work performance in actual business scenarios. The system extracts the real work order data processed by the CSR from the enterprise database, including key indicators such as dialogue text, processing time, customer satisfaction score, work order transfer rate, and repeated consultation rate. These data are cleaned and formatted through a data preprocessing module to ensure data accuracy and consistency. Next, the system uses data analysis techniques to perform multi-dimensional analysis on these behavior data. For example, by calculating the average response time for each work order processed by the CSR, evaluate their work efficiency; by analyzing the work order transfer rate, understand the CSR's ability to handle complex problems; by counting the customer satisfaction score, evaluate the CSR's service quality. In addition, the system can also use text analysis techniques to extract the CSR's communication style and problem-solving ability from the dialogue text, such as whether they can effectively soothe the customer's emotions, whether they can quickly locate problems and provide solutions. To more comprehensively evaluate the CSR's behavior characteristics, the system introduces machine learning algorithms such as clustering analysis and association rule mining. Clustering analysis can group CSRs according to their behavior patterns. For example, group high-efficiency and high-satisfaction CSRs into one group, and low-efficiency and high-transfer-rate CSRs into another group. Association rule mining can discover potential relationships between behavior data, such as discovering the negative correlation between response time and customer satisfaction. Finally, the system integrates the analysis results into a behavior characteristic portrait of the CSR, including key indicators such as response speed, transfer rate, repeated consultation rate, and customer satisfaction. These behavior characteristics can not only reflect the CSR's performance in actual work, but also provide an important basis for subsequent personalized training path planning.

[0068] Furthermore, the system can further improve the timeliness and accuracy of behavior feature evaluation by introducing real-time data analysis technology. For example, the system can monitor the process of CSR handling work orders in real time, detect abnormal behaviors in a timely manner and issue warnings. At the same time, the system can combine with the business simulation system to simulate complex scenarios in real work orders, and further verify the behavior features and problem-solving abilities of CSRs.

[0069] In addition, the system can introduce a customer feedback mechanism to allow customers to evaluate the service quality of CSRs. These customer feedback data can be combined with behavior data to further improve the behavior feature portrait of CSRs. For example, customers' evaluations of CSRs can reflect the performance of CSRs in aspects such as communication skills, problem-solving abilities, and emotion management. In this way, the system can more comprehensively evaluate the behavior features of CSRs and provide more accurate support for the enterprise's customer service training.

[0070] In some embodiments, configure a simulated dialogue test, including: S1321. Based on the skill defect weights in the dynamic skill gap matrix, retrieve specific texts in the knowledge base through the RAG technology, and generate an initial dialogue template related to the specific texts; In this embodiment, the system uses the Retrieval-Augmented Generation (RAG) technology, combines the skill deficiency weights in the dynamic skill gap matrix, retrieves specific texts from the enterprise knowledge base, and generates an initial dialogue template related thereto. This process aims to provide accurate and practical dialogue content for the simulated dialogue test, ensuring that the test can target the actual skill shortfalls of CSRs for targeted training. First, the system identifies, based on the skill deficiency weights in the dynamic skill gap matrix, the skills in which CSRs have significant gaps. For example, if the skill deficiency weight of the "emotion recognition" skill of CSRs in the "customer complaint handling" scenario is high, the system will take this skill point as the focus of retrieval. Next, the system uses the RAG technology to retrieve specific texts related to this skill point from the enterprise knowledge base. The knowledge base stores structured and unstructured data such as enterprise policy terms, product information, and frequently asked questions. The RAG technology quickly locates text fragments related to the skill deficiency in the knowledge base through the Retriever module, and then passes these fragments to the Generator module to generate an initial dialogue template related to the specific text. For example, the Retriever module may find a detailed description of the "customer complaint handling process" in the knowledge base, and the Generator module will generate an initial dialogue template based on this information, which includes common questions of customer complaints, standard response methods of CSRs, and emotion soothing words. The generated dialogue template not only contains business knowledge but also embeds key points of skill training, such as "how to recognize customer emotions" and "how to effectively soothe customer emotions".

[0071] In this embodiment, to ensure the quality of the generated dialogue template, the system introduces an expert review mechanism. Domain experts will review and optimize the generated initial dialogue template to ensure that it conforms to the actual business scenario and training requirements. In this way, the system can generate high-quality initial dialogue templates, providing a solid foundation for subsequent simulated dialogue tests.

[0072] In addition, the system of this embodiment can further improve the quality of retrieval and generation by introducing the multi-source knowledge base fusion technology. For example, in addition to the enterprise internal knowledge base, the system can integrate external resources such as industry standard knowledge bases and frequently asked question libraries to enrich the content sources of retrieval. This can ensure that the generated dialogue template not only contains enterprise-specific knowledge but also covers industry-wide best practices.

[0073] S1322. According to the characteristics of the real business scenario, inject industry-specific variables into the initial dialogue template to form a simulated dialogue test script adapted to the CSR profile.

[0074] After retrieving specific texts from the knowledge base through the RAG technology based on the skill deficiency weights in the dynamic skill gap matrix and generating an initial dialogue template related to the specific texts, industry-specific variables are injected into the initial dialogue template according to the characteristics of the real business scenario to form a simulated dialogue test script adapted to the CSR portrait. Specifically, the system generates a simulated dialogue test script adapted to the CSR portrait by injecting industry-specific variables into the initial dialogue template. This process aims to ensure that the simulated dialogue test can highly reproduce the real business scenario and be customized for the personalized needs of the CSR. First, the system analyzes the characteristics of the real business scenario and extracts key industry-specific variables. These variables may include industry terms, business processes, customer types, common problems, etc. For example, in the e-commerce industry, the variables may include "Double Eleven promotion", "Return and exchange policy", "Logistics information query", etc.; in the financial industry, the variables may include "Financial product recommendation", "Loan application process", "Risk warning", etc. Next, the system makes personalized adjustments to the initial dialogue template according to the cognitive state, behavioral characteristics, and learning preferences in the CSR portrait. For example, if the CSR is weak in the "Emotion recognition" skill in the "Customer complaint handling" scenario and has a learning preference for video tutorials, the system will inject more tips and example conversations about emotion recognition into the dialogue template and add relevant video links or animation demonstrations to the test script.

[0075] Furthermore, the system also introduces template engine technology to efficiently inject industry-specific variables into the initial dialogue template. The template engine can dynamically replace variable placeholders in the template according to preset rules and logic. For example, in a dialogue template about "Customer complaint handling", the system can inject the variable "The customer is emotionally excited" into the dialogue to generate a simulated scenario where the customer is emotionally excited and require the CSR to identify and soothe the customer's emotions in the conversation.

[0076] In addition, the system dynamically adjusts the difficulty of the dialogue script according to the CSR's skill level and training objectives. For example, for beginners, the script may provide more tips and guidance; while for experienced CSRs, the script will be more complex, including more variable combinations and unexpected situations. In this way, the system can generate highly personalized and challenging simulated dialogue test scripts to help CSRs improve their skill levels in the simulated environment.

[0077] Furthermore, the system can further enhance the realism and interactivity of the simulated dialogue test script by introducing multimodal interaction technology. For example, by combining speech recognition and synthesis technologies, the CSR can have a conversation with the virtual customer through speech to simulate a real phone customer service scenario. At the same time, the system can introduce emotion analysis technology to real-time evaluate the CSR's emotional response in the conversation and help them better manage their emotions.

[0078] In some embodiments, after constructing the three-dimensional CSR portrait including at least cognitive state, behavioral characteristics, and learning preferences, the following steps are further included: S1331: Periodically collect the behavioral data of the three-dimensional CSR in real work orders, and calculate the deviation value from the simulated dialogue test according to the behavioral data; In this embodiment, after constructing the three-dimensional CSR portrait including at least cognitive state, behavioral characteristics, and learning preferences, the following steps are further included: periodically collect the behavioral data of the three-dimensional CSR in real work orders, and calculate the deviation value from the simulated dialogue test according to the behavioral data. Specifically, the system periodically collects the behavioral data of the three-dimensional CSR in real work orders and calculates the deviation value between these data and the results of the simulated dialogue test to evaluate the accuracy and timeliness of the CSR portrait. This process is a key link in dynamically updating the CSR portrait to ensure that the portrait can reflect the latest performance of the CSR in actual work. First, the system sets a reasonable data collection period, such as weekly or monthly, and regularly extracts the real work order data processed by the CSR from the enterprise database. These data include key behavioral indicators such as response time, transfer rate, customer satisfaction score, and problem-solving efficiency. At the same time, the system records the performance data of the CSR in the most recent simulated dialogue test, including test scores, skill mastery levels, etc. Next, the system calculates the deviation value between the real work order behavioral data and the results of the simulated dialogue test through the data analysis module. Multiple statistical methods can be used to calculate the deviation value, such as mean squared error (MSE) or mean absolute error (MAE). Taking the response time as an example, the system calculates the difference between the average response time of the CSR in real work orders and the response time in the simulated dialogue test. If the difference is large, it indicates that there is a deviation between the performance of the CSR in actual work and the simulated test.

[0079] Furthermore, the system also introduces a weight mechanism for the deviation value. Different weights are assigned according to the impact degree of different behavioral indicators on the CSR's work. For example, customer satisfaction may be given a higher weight because it directly affects the customer experience and corporate reputation. Through weighted calculation, the system can more accurately evaluate the overall deviation of the CSR portrait. Finally, the system compares the calculated deviation value with a preset threshold. If the deviation value exceeds the threshold, it indicates that the CSR portrait may need to be updated to reflect the latest performance of the CSR in actual work. This provides a trigger condition for the subsequent portrait correction mechanism.

[0080] S1332: When the deviation value exceeds the preset threshold, trigger the portrait correction mechanism and reassign the priority weights of the cognitive state and behavioral characteristics in the three-dimensional CSR portrait.

[0081] After periodically collecting the behavior data of the three-dimensional CSR in real work orders and calculating the deviation value from the simulated dialogue test, when the deviation value exceeds a preset threshold, the portrait correction mechanism is triggered to reallocate the priority weights of the cognitive state and behavior characteristics in the three-dimensional CSR portrait. When the system detects that the deviation value between the CSR portrait and the actual performance exceeds the preset threshold, the portrait correction mechanism will be triggered. The core of this mechanism is to re-evaluate and adjust the priority weights of the cognitive state and behavior characteristics in the CSR portrait to ensure that the portrait can more accurately reflect the actual ability level and work performance of the CSR. First, the system determines the portrait dimension that needs to be adjusted by analyzing the specific source of the deviation value. For example, if the deviation mainly comes from the response time, it indicates that the CSR may have problems in work efficiency and the response speed weight in the behavior characteristics needs to be adjusted; if the deviation comes from the customer satisfaction score, the cognitive state of the CSR, such as the compliance of the conversation skills and the emotion management ability, may need to be re-evaluated. Next, the system dynamically adjusts the weight allocation according to the size and direction of the deviation value. For example, if the customer satisfaction score of the CSR in the real work order is lower than the simulation test result, the system will increase the weights of "emotion management" and "conversation skills compliance" in the cognitive state, and at the same time reduce the weight of "response speed" in the behavior characteristics (assuming that the response speed performance is good). This adjustment reflects the skill points that the CSR needs to improve more in actual work. To achieve the dynamic adjustment of the weights, the system adopts an adaptive algorithm to automatically optimize the weight allocation according to the historical performance and real-time data of the CSR. For example, the system can introduce a reinforcement learning mechanism to find the optimal weight allocation scheme through continuous trial and error and feedback. At the same time, the system will record the results of each adjustment for subsequent analysis and optimization. After adjusting the weights, the system will regenerate the CSR portrait and assign personalized training tasks to the CSR according to the new portrait. For example, if the adjusted behavior characteristic weights show that the CSR needs to improve the "problem-solving efficiency", the system will give priority to arranging relevant special training tasks to help the CSR better handle complex problems in actual work.

[0082] In some embodiments, the method further includes: S1611: Conduct a simulation test on the test case to obtain a simulation result; S1612: Input the simulation result into the KPI prediction model, and output a predicted value of the improvement range of the customer satisfaction after training through the KPI prediction model; S1613: Feedback and associate the predicted value with the business skill value of the target customer service staff to optimize the test case.

[0083] In this embodiment, the system validates and evaluates the generated personalized test cases through simulation tests to obtain simulation results. The purpose of the simulation test is to simulate real business scenarios, evaluate the performance of CSRs in test cases, and provide data support for subsequent KPI prediction and test case optimization. First, the system constructs a business simulation environment that can simulate the interaction between real customers and CSRs. The simulation environment includes virtual customers (played by dialogue simulation agents), CSRs (played by target customer service staff), and business scenarios (such as order query, return and exchange processing, etc.). The virtual customers will have conversations with CSRs according to preset dialogue scripts, simulating the behaviors and needs of real customers. During the simulation test, the system records various performance data of CSRs in the conversations, including response time, problem-solving efficiency, customer satisfaction scores, emotion management capabilities, etc. These data are analyzed through natural language processing technology to generate a detailed simulation result report. For example, the system can analyze whether CSRs can effectively appease customers' emotions when handling customer complaints, whether they can quickly locate problems and provide solutions. To ensure the accuracy and reliability of the simulation test, the system introduces a multi-round dialogue mechanism. In each round of conversation, the virtual customers will dynamically adjust the conversation content according to the CSRs' answers, simulating the possible reactions of real customers. For example, if the CSR fails to effectively solve the problem, the virtual customer may show dissatisfaction and require the CSR to provide further solutions. In this way, the system can comprehensively evaluate the performance of CSRs in different scenarios.

[0084] Furthermore, the system inputs the results of the simulation test into the KPI prediction model to predict the improvement in customer satisfaction of CSRs after completing the training. This process aims to provide a quantitative assessment and forward-looking prediction of the training effect through data analysis and machine learning techniques. First, the system extracts key metrics from the simulation test, such as the response time of CSRs in conversations, problem-solving efficiency, customer satisfaction scores, emotional management capabilities, etc. These metrics are organized into structured data and input into the KPI prediction model. The KPI prediction model is a machine learning-based prediction model, usually constructed using regression analysis, time series analysis, or deep learning algorithms (such as neural networks). The training data of the model includes historical ticket data, training records of CSRs, and actual feedback on customer satisfaction. Through these data, the model learns the customer satisfaction performance of CSRs at different skill levels and establishes a prediction model. For example, the model can learn that when the response time of CSRs is shortened by 10%, the customer satisfaction may increase by 5 percentage points; or the specific increase in customer satisfaction after the improvement of CSRs' emotional management capabilities. After inputting the simulation results, the KPI prediction model will output the predicted value of the improvement in customer satisfaction of CSRs after completing the training according to its internal prediction logic. For example, if the simulation results show a significant improvement in the emotional management capabilities of CSRs when handling customer complaints, the model may predict that the customer satisfaction will increase by 8 percentage points.

[0085] Furthermore, the system correlates the predicted value of the improvement in customer satisfaction output by the KPI prediction model with the business skill values of CSRs to optimize the test cases. This process aims to ensure, through a feedback mechanism, that the test cases can better reflect the actual improvement in the capabilities of CSRs and provide more accurate guidance for subsequent training. First, the system conducts a comparative analysis of the predicted improvement in customer satisfaction and the business skill values of CSRs in the simulation test. The business skill values are obtained through a quantitative assessment of the performance of CSRs in various skills, such as emotional management capabilities, problem-solving efficiency, and compliance of conversation scripts. The system uses correlation analysis to identify which skill improvements contribute the most to the improvement in customer satisfaction. For example, if the prediction results show that the improvement in customer satisfaction is mainly due to the improvement in the emotional management capabilities of CSRs, the system will mark this skill point as a key skill. Next, the system optimizes the test cases according to the results of the feedback correlation. If it is found that some test cases fail to effectively improve the key skills of CSRs, or the performance of CSRs in some skills does not match the prediction results, the system will adjust the content and difficulty of the test cases. For example, if the performance of CSRs in emotional management capabilities does not meet the expectations, the system may add more conversation scenarios for emotion recognition and appeasement, or adjust the complexity of the conversation scenarios to better exercise the skills of CSRs.

[0086] In this way, the system can more comprehensively evaluate the effectiveness of test cases, timely detect the deviation between test cases and actual business requirements, and provide more accurate support for the enterprise's customer service training.

[0087] In some embodiments, inputting the simulation result into the KPI prediction model and outputting the predicted value of the improvement amplitude of the customer satisfaction after training by the KPI prediction model further includes: S1621. Normalize the skill defect values in the dynamic skill gap matrix, and calculate the influence coefficient of each skill on the KPI in combination with the change of customer satisfaction in historical training data; S1622. Weightedly sum the influence coefficient and the current skill defect value, and output the quantitative predicted value of the improvement amplitude of the customer satisfaction after training.

[0088] In this embodiment, during the process of inputting the simulation results into the KPI prediction model and outputting the predicted value of the improvement range of customer satisfaction after training through the KPI prediction model, it also includes normalizing the skill defect values in the dynamic skill gap matrix, combining the changes in customer satisfaction in the historical training data, and calculating the influence coefficients of each skill on the key performance indicator (KPI). This process aims to quantitatively evaluate the potential impact of each skill on customer satisfaction and provide data support for subsequent predictions. Specifically, first, the system normalizes the skill defect values in the dynamic skill gap matrix. Normalization is to convert the skill defect values to a unified range (such as 0 to 1) so that the defect values between different skills can be directly compared. The normalization method can adopt Min-Max Normalization or Z-Score Standardization. For example, for the skill "emotion recognition", its defect values may range from 0 to 100. After normalization, all defect values are converted to the range of 0 to 1. Next, the system combines the changes in customer satisfaction in the historical training data and calculates the influence coefficients of each skill on the KPI. The historical training data includes the customer satisfaction scores of CSRs before and after training, the skill improvement situation, and the corresponding business scenarios. The system analyzes these data through the data analysis module to find out which skill improvements contribute the most to the improvement of customer satisfaction. For example, the system may find that the improvement of the "emotion recognition" skill contributes 0.8 to the improvement of customer satisfaction, while the improvement of the "problem-solving efficiency" skill contributes 0.6. To calculate the influence coefficients, the system uses correlation analysis or regression analysis methods. For example, through a linear regression model, the system can establish the relationship between skill improvement and customer satisfaction improvement. The input of the model is the change in skill defect values (such as the difference before and after training), and the output is the change in customer satisfaction. Through model training, the system can obtain the influence coefficients of each skill. These influence coefficients reflect the sensitivity of each skill to customer satisfaction and provide an important basis for subsequent predictions.

[0089] Further, in this embodiment, the influence coefficient is weighted and summed with the current skill defect value to output a quantitative prediction value of the improvement range of the customer satisfaction after training. Specifically, first, the system performs weighted summation based on the influence coefficient calculated in step S71 and combines it with the current skill defect value. Specifically, the influence coefficient of each skill is multiplied by the corresponding skill defect value, and then the results of all skills are added together. For example, assume that the influence coefficient of the "emotion recognition" skill is 0.8 and the current defect value is 0.6 (after normalization), then the contribution of this skill to the improvement of customer satisfaction is 0.8 × 0.6 = 0.48. Similarly, after calculating the contribution values of all skills, these values are added together to obtain the total contribution value. Next, the system converts the total contribution value into a quantitative prediction value of the improvement range of customer satisfaction. For example, if the total contribution value is 1.5, the system can map this value to a specific percentage of the improvement of customer satisfaction according to historical data and business experience. Assume that historical data shows that a total contribution value of 1.5 corresponds to a 10 percentage point increase in customer satisfaction, then the predicted value output by the system is 10%. To ensure the accuracy and reliability of the predicted value, the system introduces a calibration mechanism. The calibration mechanism adjusts the influence coefficient and weight allocation by comparing the historical predicted value with the actual improvement of customer satisfaction. For example, if the historical predicted value is too high or too low, the system will adjust the influence coefficient according to the actual data to improve the accuracy of the prediction. In addition, the system can introduce a user feedback mechanism to allow training managers and CSRs to participate in the evaluation process of the predicted value. For example, the system can display the predicted value to the training manager and solicit their opinions. The training manager can adjust the predicted value according to the actual business needs and experience. In this way, the system can generate a predicted value that better meets the actual business needs, improving the training effect and user experience.

[0090] Through the above implementation manner, the system can provide more accurate and timely prediction results for training managers, helping them better plan training content and evaluate training effects.

[0091] For details, please refer to Figure 2 , Figure 2 which is the basic structural schematic diagram of the intelligent customer service training device for dynamic skill mapping in this embodiment.

[0092] As shown in Figure 2As shown in the figure, an intelligent customer service training device for dynamic skill mapping includes: a data extraction module 1100, which is used to extract the business scenario and skill chain association data in the historical work orders of customer service personnel and generate a fine-tuning data set; a skill matrix module 1200, which is used to train a skill recognition Agent based on the fine-tuning data set and output a dynamic skill gap matrix of CSRs; a portrait construction module 1300, which is used to construct a three-dimensional CSR portrait including at least cognitive state, behavior characteristics, and learning preferences according to the dynamic skill gap matrix, combined with the results of simulated dialogue tests and real work order behavior data; a task planning module 1400, which is used to generate a priority training task sequence and a special training task by a path planning Agent based on the three-dimensional CSR portrait and the dynamic skill gap matrix; a use case generation module 1500, which is used to generate personalized test cases by a dialogue simulation Agent using the priority training task sequence, the special training task, and the three-dimensional CSR portrait; a training configuration module 1600, which is used to assign the personalized test cases to target customer service personnel and obtain the business index data of the target customer service personnel to quantify the business skill value of the target customer service personnel.

[0093] The above intelligent customer service training device for dynamic skill mapping extracts the business scenario and skill chain association data in the historical work orders of customer service personnel to generate a fine-tuning data set; trains a skill recognition Agent based on the fine-tuning data set and outputs a dynamic skill gap matrix of CSRs; constructs a three-dimensional CSR portrait including at least cognitive state, behavior characteristics, and learning preferences according to the dynamic skill gap matrix, in combination with the simulated dialogue test results and real work order behavior data; generates a priority training task sequence and special training tasks by a path planning Agent based on the three-dimensional CSR portrait and the dynamic skill gap matrix; generates personalized test cases by a dialogue simulation Agent using the priority training task sequence, special training tasks, and three-dimensional CSR portrait; allocates the personalized test cases to target customer service personnel, obtains the business metric data of the target customer service personnel to quantify the business skill value of the target customer service personnel; through the job portrait and skill dynamic mapping technology, it can accurately identify the skill deficiencies of customer service representatives (CSRs) in different business scenarios, solve the pain point of lagging skill diagnosis in traditional training, and realize the real-time dynamic adjustment of training content. Secondly, the three-dimensional CSR portrait constructed by the multi-modal learner modeling technology comprehensively covers cognitive state, behavior characteristics, and learning preferences, making the training more targeted and personalized, and effectively improving the training effect. In addition, the virtual-real linkage training resource generation technology combined with the business simulation system can generate highly simulated personalized test cases to help CSRs accumulate experience in the simulated environment and enhance their ability to handle complex business scenarios. Finally, by quantitatively evaluating the skill value, the system can provide real-time feedback on the training effect and predict the improvement range of business KPIs, solve the problem of delayed feedback on the training effect in traditional training, and provide an efficient, accurate, and quantifiable solution for the customer service training of enterprises.

[0094] To solve the above technical problems, an embodiment of the present application further provides a computer device. For details, please refer to Figure 3 , Figure 3 which is the basic structural block diagram of the computer device in this embodiment.

[0095] As Figure 3As shown, it is a schematic diagram of the internal structure of a computer device. The computer device includes a processor, a non-volatile storage medium, a memory, and a network interface connected through a system bus. Among them, the non-volatile storage medium of the computer device stores an operating system, a database, and computer-readable instructions, and control information sequences can be stored in the database. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. Computer-readable instructions can be stored in the memory of the computer device. When the computer-readable instructions are executed by the processor, the processor can execute an intelligent customer service training method for dynamic skill mapping. The computer-readable instructions are stored in the form of a computer program. The network interface of the computer device is used to connect and communicate with a terminal. Those skilled in the art can understand, Figure 3 The structure shown in is only a block diagram of some structures 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. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0096] In this embodiment, the processor is used to execute Figure 2 the specific functions of the data extraction module 1100, the skill matrix module 1200, the portrait construction module 1300, the task planning module 1400, the use case generation module 1500, and the training configuration module 1600 in. The memory stores the program codes and various types of data required to execute the above modules. The network interface is used for data transmission between the user terminal and the server. The memory in this embodiment stores the program codes and data required to execute all sub-modules in the intelligent customer service training device for dynamic skill mapping. The server can call the program codes and data of the server to execute the functions of all sub-modules.

[0097] The computer device generates a fine-tuning dataset by extracting the business scenario and skill chain association data from the historical work orders of customer service staff; trains a skill recognition Agent based on the fine-tuning dataset and outputs a dynamic skill gap matrix of CSRs; constructs a three-dimensional CSR portrait including at least cognitive state, behavior characteristics, and learning preferences according to the dynamic skill gap matrix, in combination with the simulation dialogue test results and real work order behavior data; generates a priority training task sequence and special training tasks by a path planning Agent based on the three-dimensional CSR portrait and the dynamic skill gap matrix; generates personalized test cases by a dialogue simulation Agent using the priority training task sequence, special training tasks, and the three-dimensional CSR portrait; allocates the personalized test cases to the target customer service staff to obtain the business metric data of the target customer service staff to quantify the business skill value of the target customer service staff; through the job portrait and skill dynamic mapping technology, it can accurately identify the skill shortboards of customer service staff (CSRs) in different business scenarios, solve the pain point of lagging skill diagnosis in traditional training, and realize the real-time dynamic adjustment of training content. Secondly, the three-dimensional CSR portrait constructed by the multi-modal learner modeling technology comprehensively covers cognitive state, behavior characteristics, and learning preferences, making the training more targeted and personalized, and effectively improving the training effect. In addition, the virtual-real linkage training resource generation technology combined with the business simulation system can generate highly simulated personalized test cases, helping CSRs accumulate experience in the simulation environment and enhancing their ability to handle complex business scenarios. Finally, by quantifying the evaluation of skill values, the system can provide real-time feedback on the training effect and predict the improvement range of business KPIs, solving the problem of delayed feedback on the training effect in traditional training, and providing an efficient, accurate, and quantifiable solution for the customer service training of enterprises.

[0098] The present application also provides a storage medium storing computer-readable instructions, which when executed by one or more processors, cause the one or more processors to execute the steps of the intelligent customer service training method for dynamic skill mapping according to any one of the above embodiments.

[0099] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0100] Those skilled in the art can understand that the various operations, methods, steps, measures, and solutions in the processes discussed in this application can be alternated, changed, combined, or deleted. Further, other steps, measures, and solutions in the various operations, methods, and processes discussed in this application can also be alternated, changed, rearranged, decomposed, combined, or deleted. Further, those in the prior art that have steps, measures, and solutions in the various operations, methods, and processes disclosed in this application can also be alternated, changed, rearranged, decomposed, combined, or deleted.

[0101] The above are only partial embodiments of this application. It should be noted that for those of ordinary skill in the art, without departing from the principle of this application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of this application.

Claims

1. An intelligent customer service training method for dynamic skill mapping, characterized in that Including: Extract the business scenario and skill chain association data from the historical work orders of customer service staff to generate a fine-tuning dataset; Train a skill recognition Agent based on the fine-tuning dataset and output the dynamic skill gap matrix of CSRs; According to the dynamic skill gap matrix, combined with the simulation dialogue test results and real work order behavior data, construct a three-dimensional CSR portrait that at least includes cognitive state, behavior characteristics, and learning preferences; Based on the three-dimensional CSR portrait and the dynamic skill gap matrix, generate a priority training task sequence and special training tasks by a path planning Agent; Using the priority training task sequence, special training tasks, and three-dimensional CSR portrait, generate personalized test cases by a dialogue simulation Agent; Assign the personalized test cases to the target customer service staff and obtain the business metric data of the target customer service staff to quantify the business skill value of the target customer service staff.

2. The intelligent customer service training method for dynamic skill mapping according to claim 1, characterized in that The training of the skill recognition Agent based on the fine-tuning dataset and outputting the dynamic skill gap matrix of CSRs includes: Adopt the thought chain annotation technique to disassemble the skill chain in the fine-tuning training set; Input the disassembled skill chain into the skill recognition Agent to train the skill recognition Agent; Train the skill recognition Agent to map the training objectives to sub-skills and generate the dynamic skill gap matrix of CSRs.

3. The intelligent customer service training method for dynamic skill mapping according to claim 1, wherein Before constructing the three-dimensional CSR portrait that at least includes cognitive state, behavior characteristics, and learning preferences according to the dynamic skill gap matrix, combined with the simulation dialogue test results and real work order behavior data, it further includes: Configure the simulation dialogue test and analyze the simulation dialogue test results to evaluate the cognitive state of CSRs; Analyze the behavior data in the real work orders to determine the behavior characteristics.

4. The intelligent customer service training method for dynamic skill mapping according to claim 3, wherein The configuration of the simulation dialogue test includes: Based on the skill defect weights in the dynamic skill gap matrix, retrieve specific texts in the knowledge base through the RAG technique to generate an initial dialogue template related to the specific texts; According to the real business scenario characteristics, inject industry-specific variables into the initial dialogue template to form a simulation dialogue test script adapted to the CSR portrait.

5. The intelligent customer service training method for dynamic skill mapping according to claim 4, characterized in that After constructing the three-dimensional CSR portrait that at least includes cognitive state, behavior characteristics, and learning preferences, it further includes: Periodically collect the behavior data of the three-dimensional CSR in real work orders and calculate the deviation value from the simulation dialogue test according to the behavior data; When the deviation value exceeds the preset threshold, trigger the portrait correction mechanism and reassign the priority weights of the cognitive state and behavior characteristics in the three-dimensional CSR portrait.

6. The intelligent customer service training method for dynamic skill mapping according to claim 1, wherein The method further includes: Conduct a simulation test on the test cases to obtain the simulation results; Input the simulation results into a KPI prediction model and output the predicted value of the improvement amplitude of customer satisfaction after training through the KPI prediction model; Perform feedback association between the predicted value and the business skill value of the target customer service staff to optimize the test cases.

7. The intelligent customer service training method for dynamic skill mapping according to claim 6, wherein The inputting the simulation results into a KPI prediction model and outputting the predicted value of the improvement amplitude of customer satisfaction after training through the KPI prediction model includes: Normalize the skill defect values in the dynamic skill gap matrix, and calculate the influence coefficient of each skill on the KPI in combination with the change in customer satisfaction in the historical training data; Weighted sum the influence coefficient and the current skill defect value, and output the quantitative prediction value of the improvement amplitude of the customer satisfaction after training.

8. An intelligent customer service training device for dynamic skill mapping, characterized in that, It includes: A data extraction module for extracting the business scenario and skill chain association data in the historical work orders of customer service personnel to generate a fine-tuning data set; A skill matrix module for training a skill recognition Agent based on the fine-tuning data set and outputting the dynamic skill gap matrix of CSR; A portrait construction module for constructing a three-dimensional CSR portrait including at least cognitive state, behavior characteristics, and learning preferences according to the dynamic skill gap matrix, in combination with the simulation dialogue test results and real work order behavior data; A task planning module for generating a priority training task sequence and special training tasks by a path planning Agent based on the three-dimensional CSR portrait and the dynamic skill gap matrix; A use case generation module for generating personalized test cases by a dialogue simulation Agent using the priority training task sequence, special training tasks, and three-dimensional CSR portrait; A training configuration module for allocating the personalized test cases to target customer service personnel and obtaining the business indicator data of the target customer service personnel to quantify the business skill value of the target customer service personnel.

9. A computer device, characterized in that, It includes a memory and a processor. When the computer-readable instructions stored in the memory are executed by the processor, the processor executes the steps of the intelligent customer service training method for dynamic skill mapping according to any one of claims 1 to 7.

10. A storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the intelligent customer service training method for dynamic skill mapping according to any one of claims 1 to 7.

Images