Intelligent drill system, method, medium, device and program product based on multiple agents working in collaboration

Abstract

The application provides an intelligent exercise system, method, medium, device and program product based on the cooperative work of multiple agents. The application creates different types of virtual customers through a large model, and multiple agents cooperate to form a perfect training system. The quality control standard problem caused by traditional manual exercise is solved, resources are released, efficiency is improved, training data is retained, and can be traced and analyzed. At the same time, it makes up for the problem that the first generation of intelligent exercise is difficult to flexibly open, the course configuration cost is high, and the cycle is long. Intelligent exercise generates virtual customers through a large language model, creates realistic and challenging dialogue scenarios in insurance marketing, adjusts the conversation topic at any time according to the context of the dialogue, the customer's emotions and needs, and actively asks and asks virtual customers with personality and attitude, improving the natural fluency and practical experience of the dialogue. At the same time, the dialogue process provides real-time support and suggestions for the dialogue process, helping students better understand and solve customer needs. After the exercise is completed, the intelligent exercise provides analysis of customer portrait solutions and a comprehensive evaluation of the entire dialogue record, providing feedback and improvement suggestions. In this process, the student is helped to transform knowledge into ability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of large models, in particular to an intelligent rehearsal system and method based on the cooperative work of multiple agents, a medium, a device and a program product. BACKGROUND

[0002] Under the traditional insurance training mode, agents need to conduct one-on-one offline simulated interviews with supervisors to achieve the training goal of passing the rehearsal. With the standardization of training effectiveness, training efficiency, and rehearsal data retention, online rehearsal is explored in the training field. Because the first generation of intelligent rehearsal is based on task type and FAQ, there are various problems, such as inflexible pre-written passing process, training content that is not strong in practicality, heavy course updating burden, long time, etc., so users still have difficulty in improving training efficiency and effectiveness. User use: how to realize data recording and retention of the rehearsal process to facilitate subsequent data analysis and traceability. Therefore, the current insurance industry training mode at least has the following technical problems:

[0003] Traditional offline rehearsal:

[0004] (1) Rehearsal quality control standard problem: how to ensure the consistency of different supervisors' passing standards for courses to improve training effectiveness.

[0005] (2) Rehearsal efficiency problem: how to reduce the time input of supervisors in the training process and improve the rehearsal frequency of agents.

[0006] (3) Rehearsal data retention problem: how to realize data recording and retention of the rehearsal process to facilitate subsequent data analysis and traceability.

[0007] First generation of intelligent rehearsal

[0008] (1) Rehearsal effectiveness problem: pre-set interview questions and processes, inflexible experience, and students have difficulty in exerting their own initiative and creativity.

[0009] (2) Rehearsal output efficiency problem: slow course output, requiring a lot of building work, pre-setting all questions and processes, developing detailed and large-scale audit standards, and large tuning workload. SUMMARY

[0010] In view of the above-mentioned shortcomings of the prior art, the purpose of the present application is to provide a solution to the technical problems of existing insurance training modes in terms of rehearsal quality control, rehearsal efficiency, rehearsal data retention, rehearsal effectiveness, and rehearsal output efficiency.

[0011] To achieve the above object and other related objects, the first aspect of the present application provides an intelligent rehearsal system based on the cooperative work of multiple agents, comprising: a customer portrait generation agent configured to construct a structured data instruction based on insurance industry customer data to guide a large model to generate a customer portrait of a virtual insurance customer with different personality characteristics; a custom portrait inspection and polishing agent configured to inspect a custom customer portrait generated in response to a user operation and perform polishing processing for a customer portrait lacking necessary structured information; a scenario introduction generation agent configured to generate a dialogue scene of the insurance industry according to the customer portrait; a training dialogue generation agent configured to output insurance business dialogue content based on the simulated customer portrait and dialogue scene as input; a dialogue technique assistant agent configured to provide corresponding reply suggestions according to the questions raised by the virtual customer in the generated dialogue scene of the insurance industry, in combination with the customer portrait, scene description and dialogue context; and an intelligent end judgment agent configured to dynamically generate a corresponding dialogue ending according to the dialogue context, customer emotional characteristics and demand characteristics, and current interaction state.

[0012] In some embodiments of the first aspect of the present application, the intelligent rehearsal system further comprises any one or both of the following agents: a customer analysis approach agent configured to trace and analyze the virtual customer's portrait characteristics after the dialogue ends, by combining the customer portrait, scene description and dialogue record, and providing prediction questions and solution approaches through comprehensive learning and reasoning; and a rehearsal evaluation agent configured to generate evaluation information of the training rehearsal according to the dialogue record after the dialogue ends.

[0013] In some embodiments of the first aspect of the present application, the construction method of the customer portrait generation agent comprises: collecting customer data of the insurance industry and constructing a structured data instruction, using the structured data instruction to pre-process and extract features from the collected customer data; the structured data instruction is used to convert raw data into a structured format suitable for machine learning training; and inputting the extracted features into a GPT model for model training to construct the customer portrait generation agent.

[0014] In some embodiments of the first aspect of the present application, the inspection content of the custom portrait inspection and polishing agent on the customer portrait comprises: completeness inspection configured to verify whether the customer portrait input by the user contains all necessary structured information, and generate a series of questions or prompts according to the missing information points to guide the user to supplement the missing information, and recheck the completeness of the customer portrait after the user supplements the information; personal expression inspection configured to check whether the customer portrait uses second-person expression; and key information point inspection configured to check whether the customer portrait contains at least the following key information points: personal identity, relationship with the agent, personal situation, attitude towards insurance, and purchased insurance, etc.

[0015] In some embodiments of the first aspect of the application, the construction of the scenario introduction generation agent includes: using NLP technology to analyze the customer portrait, identifying and extracting key information points of the virtual customer; using a machine learning algorithm to extract deep features containing the implicit needs and personality characteristics of the virtual customer from the customer portrait; inputting the extracted key information points and deep features of the virtual customer into the GPT model for training, so that it can understand and simulate the behavior and language style of the customer; according to the characteristics of the customer portrait, one or more dialogue frameworks are constructed, which cover the interaction paths of different business scenarios; in each dialogue framework, the trained GPT model is used to generate dialogue content that meets the customer's characteristics; the evaluation and feedback information of the user in the dialogue process is obtained, and the GPT model is optimized through multiple iterations accordingly.

[0016] In some embodiments of the first aspect of the application, the construction of the training dialogue generation agent includes: receiving the customer portrait of the virtual customer from the customer portrait generation agent, receiving the polished user-defined customer portrait from the custom portrait checking and polishing agent, and receiving the simulated dialogue scene content from the scenario introduction generation agent, preprocessing and feature extraction of the received data; the GPT model understands the customer's intention and demand according to the input features and context information; using the language patterns and structures learned in its training, it constructs dialogue content that meets the role and scenario and outputs; wherein the language patterns and structures of the insurance industry include: use of professional terms, inquiry and question and answer mode, risk assessment and suggestion, claim process description, compliance and standardization, customer service language.

[0017] To achieve the above-mentioned and other related purposes, the second aspect of the present application provides an intelligent simulation method based on the cooperative work of multiple agents, which constructs multiple agents based on the GPT model to cooperatively work as follows: based on the structured data instructions constructed from the insurance industry customer data, guide the large model to generate customer portraits of virtual insurance customers with different personality characteristics; check the custom customer portraits generated in response to user operations, and polish the customer portraits that lack necessary structured information; according to the customer portrait, generate a dialogue scene of the insurance industry; based on the simulated customer portrait and dialogue scene as input, output the insurance business dialogue content; in the generated dialogue scene of the insurance industry, according to the questions raised by the virtual customer, combined with the customer portrait, scene description and the context of the dialogue, provide corresponding reply suggestions; according to the context of the dialogue, customer emotional characteristics and demand characteristics, and the current interaction state, dynamically generate corresponding dialogue closing words.

[0018] To achieve the above object and other related objects, the third aspect of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the intelligent rehearsal method based on the collaborative work of multiple agents.

[0019] To achieve the above object and other related objects, the fourth aspect of the present application provides a computer program product, which includes computer program codes, and when the computer program codes are run on a computer, the computer is caused to implement the intelligent rehearsal method based on the collaborative work of multiple agents.

[0020] To achieve the above object and other related objects, the fifth aspect of the present application provides a computer device, which includes a memory, a processor and a computer program stored in the memory; the processor executes the computer program to implement the intelligent rehearsal method based on the collaborative work of multiple agents.

[0021] As described above, the intelligent rehearsal system, method, medium, device and program product based on the collaborative work of multiple agents of the present application have the following beneficial effects:

[0022] (1) Different types of virtual customers are created through large models, multiple agents collaborate to form a perfect training system, and the training efficiency is greatly improved. All data generated during the training process can be well preserved for tracing and analysis.

[0023] (2) Through large models, realistic and challenging dialogue scenarios are created, and appropriate closing remarks can be dynamically generated according to the context of the dialogue, the emotions and needs of the customer, and the current interactive state, improving the natural flow and service experience of the dialogue.

[0024] (3) Through large models, virtual customers are automatically generated, and users can also customize customer portraits and polish them, providing diverse interactive rehearsal opportunities for trainees. At the same time, the dialogue assistant provides real-time support and suggestions during the dialogue to help trainees better understand and respond to the needs of customers. The entire dialogue record is also comprehensively evaluated to provide feedback and improvement suggestions, further improving the training effect. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The structure diagram of the intelligent rehearsal system based on the collaborative work of multiple agents in an embodiment of the present application is shown.

[0026] Figure 2 The diagram of the customer portrait of the virtual customer in an embodiment of the present application is shown.

[0027] Figure 3A process diagram of generating a customer portrait by a customer portrait generation agent in an embodiment of the present application is shown.

[0028] Figure 4 A GUI interface diagram of checking items of a customer portrait in an embodiment of the present application is shown.

[0029] Figure 5 A process diagram of generating a simulated dialogue scene by a scenario introduction generation agent in an embodiment of the present application is shown.

[0030] Figure 6 A flow diagram of an intelligent drill method based on the collaborative work of multiple agents in an embodiment of the present application is shown.

[0031] Figure 7 A structural diagram of a computer device in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0032] The embodiments of the present application will be described in detail by specific concrete examples, and those skilled in the art can easily understand other advantages and effects of the present application from the disclosure of the present specification. The present application can also be implemented or applied by other different specific embodiments, and the details in the present specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be noted that the following examples and features in the examples can be combined with each other without conflict.

[0033] Before the present application is further described, the terms and terminology used in the embodiments of the present application are explained, and the terms and terminology used in the embodiments of the present application are applicable to the following explanations:

[0034] <1> GPT (Generative Pre-trained Transformer): a natural language processing (NLP) model based on the Transformer architecture, the GPT model learns the patterns and structures of language by pre-training a large amount of text data, and can then be fine-tuned on various language tasks such as text generation, translation, question answering, etc. The core of the GPT model is the Transformer, which is an attention mechanism model that can process sequence data and capture long-range dependencies.

[0035] <2> Agent: a concept in the field of computer science and artificial intelligence, referring to an entity that can perceive the state of the environment and make decisions to achieve specific goals in a specific environment. Agents can be in the form of software, such as chatbots, recommendation systems, or in the form of hardware, such as autonomous vehicles, robots.

[0036] The present application aims to simulate real customer conversation scenarios through the collaborative work of multiple agents. These agents are carefully designed to perform specific tasks such as understanding, responding, and sentiment analysis, etc., together forming an efficient training environment. Through this multi-model collaborative work, the system can provide a more natural, lively, and challenging communication experience, effectively improving users' communication skills and problem-solving abilities.

[0037] Figure 1 A structural diagram of an intelligent rehearsal system based on the collaborative work of multiple agents in an embodiment of the present application is shown. The intelligent rehearsal system in this embodiment includes a customer portrait generation agent 101, a self-defined portrait checking and polishing agent 102, a situation introduction generation agent 103, a training conversation generation agent 104, a dialogue technique assistant agent 105, an intelligent end judgment agent 106, a customer analysis thought agent 107, and a rehearsal evaluation agent 108.

[0038] The customer portrait generation agent 101 is used to construct structured data instructions based on insurance industry customer data to guide large models to generate customer portraits of virtual insurance customers with different personality characteristics. These customer portraits will simulate customers with different personalities, providing students with diverse interactive rehearsal opportunities. For example, the customer portrait generation agent can generate a customer portrait of a young single professional who may have high requirements for health insurance and accidental injury insurance; or generate a customer portrait of a middle-aged housewife who may be more concerned about endowment insurance and children's education insurance, etc. In this way, students can interact with different types of customers in a simulated environment, improving their communication skills and sales abilities.

[0039] The process of generating a customer portrait of a virtual insurance customer by the customer portrait generation agent 101 is shown in Figure 3

[0040] Step S3a: Collect customer data in the insurance industry and construct structured data instructions, use the structured data instructions to preprocess and feature extract the collected customer data; the customer data is simulated customer data based on real data structure and has been executed desensitization operation.

[0041] Figure 2 ​A schematic diagram of a customer portrait of a virtual customer in an embodiment of the present application is shown. The customer data of the insurance industry includes but is not limited to customer basic information, policy details, claim record information, etc. The customer basic information includes personal information (such as name, gender, age, occupation, education background, etc.), contact information (such as phone number, email address, etc.), family and financial status (such as family structure, income level, assets and liabilities, etc.), which helps to assess the customer's payment ability and insurance needs. The policy details include the policy type (such as the type of insurance purchased by the customer, such as medical insurance, critical illness insurance, annuity insurance, etc.), policy terms (such as insurance liability, exemption terms, insurance amount, etc.), policy history (such as purchase time, policy term, renewal situation, etc. purchase history information), etc. The claim record information includes the number of claims, the amount of claims, the reason for claims (such as illness, accident), etc. It is worth noting that the training data used in the embodiments of the present application is preferably desensitized data, which is simulated customer data based on real data structure, aiming to protect the security of the real information of the customers, and to ensure that such operation complies with relevant regulatory requirements. In this way, the present application can fully utilize data to optimize the model and system without revealing the personal information of the customers.

[0042] The structured data instruction refers to a series of rules and steps defined in the data preprocessing and feature engineering stage, which is used to convert the original data into a structured format suitable for machine learning model training. For example, the structured data instruction can include: data cleaning instruction, feature selection instruction, feature conversion instruction, feature construction instruction, data encoding instruction, data segmentation instruction, data resampling instruction, time series feature extraction instruction, etc. The data cleaning instruction is used to delete or fill in missing values, remove or correct outliers, standardize date formats, etc. The feature selection instruction is used to select features highly related to the target variable (such as insurance claim amount), and exclude features irrelevant or redundant to the target variable. The feature conversion instruction is used to uniquely encode categorical variables, such as converting "male" and "female" into two binary features, and standardizing or normalizing continuous variables. The feature construction instruction is used to create new features, such as extracting year, month, etc. from date, or deriving new features from original features, such as "young male", "middle-aged female" and other categories from age and gender features. The data encoding instruction is used to convert text data into numerical data. The data segmentation instruction is used to divide the data set into training set and test set, such as dividing according to the proportion of 70% training set and 30% test set. The data resampling instruction is used to balance the classes using oversampling or undersampling techniques for unbalanced data sets. The time series feature extraction instruction is used to extract trend, seasonality, etc. features for time series data.

[0043] Step S3b: input the extracted features into the GPT model for model training to build the customer portrait generation agent.

[0044] In some examples, the extracted features are combined to form a dataset, and the dataset is divided into a training set and a validation set according to a certain proportion. The training set is input into the GPT model for training, so as to obtain a customer portrait generation agent. In the process of training the GPT model using the training set, the model parameters are optimized by minimizing the loss function, and the model learns how to predict the next most suitable token (basic unit in text, such as word, character, string, etc.) according to the given context. Gradient descent algorithm is used to update the model weights to reduce the difference between the predicted value and the actual value. After training is completed, the GPT model can generate a series of tokens according to the given context, and these tokens combined form the customer portrait of the virtual insurance customer. Finally, the validation set is used to evaluate the performance of the model, check the accuracy and relevance of the generated customer portrait, and adjust the model parameters or training strategy according to the evaluation results, so as to improve the generalization ability of the model.

[0045] It should be understood that the GPT model is a pre-trained language model based on the Transformer architecture, and its core is the self-attention mechanism, which enables the model to capture the dependency between any two positions in the input sequence, regardless of the distance between them. In the training process, the model learns how to predict the next most likely token according to the given context, and this ability makes the GPT model very suitable for text generation tasks. In generating virtual customer portraits, the GPT model serves as a powerful text generator that can generate coherent, relevant, and personalized text descriptions based on the input context (which may be the basic attributes or behavior patterns of the customer). This generation process is achieved by maximizing the probability of the next token, and the model will continuously adjust its predictions to match the patterns in the training data. In this way, the GPT model can learn how to generate detailed customer portraits based on customer data, including their possible preferences, behaviors, and needs.

[0046] The custom portrait checking and polishing agent 102 is used to check the custom customer portrait generated in response to user operations and polish the customer portrait that lacks necessary structured information.

[0047] It should be noted that the "user" referred to in the present application refers to an agent who uses this set of intelligent rehearsal system for training, so "agent" in the following is actually the same as "user". The "customer" referred to in the present application refers to a virtual customer generated by the GPT model or formed by the user's self-definition.

[0048] When the user chooses to input the customer portrait by himself / herself, in order to ensure the completeness and richness of the portrait, the GPT model-based custom portrait checking and polishing intelligent agent 102 analyzes the user input text, compares it with the pre-defined necessary information standard, and checks whether all necessary structured information is included or the information is not detailed enough.

[0049] In some examples, the GPT model-based custom portrait checking and polishing intelligent agent 102 checks the customer portrait in terms of completeness, personal pronoun expression, and key information points. The completeness check verifies whether the user input customer portrait includes all necessary structured information, and generates a series of questions or prompts based on the missing information points to guide the user to supplement the missing information. After the user supplements the information, the customer portrait is checked again for completeness and natural language processing to ensure the detail of the information and the accuracy of the expression. The personal pronoun expression check checks whether the customer portrait uses second person pronouns. The key information point check checks whether the customer portrait includes at least the following key information points: personal identity, relationship with the agent, personal situation, attitude towards insurance, etc. Among them, during the key information point check, the GPT model-based custom portrait checking and polishing intelligent agent 102 provides prompt information to the user for supplement or automatically supplements according to the information provided by the user for the customer portrait that lacks key information points or the information detail does not meet the standard. For example, if the user only inputs the customer's name and age, the custom portrait checking and polishing intelligent agent 102 can automatically add some common information points such as occupation, income level, family status, etc., so that the portrait is closer to the real customer.

[0050] For example, the GUI interface shown is as follows: Figure 4 The checking items of the customer portrait include:

[0051] 1. Whether it is a second person expression, such as "You are Ms. Wang".

[0052] 2. Whether it contains more than 3 of the following information points:

[0053] (1) Who are you? Or how do you usually call yourself. Such as "You are Ms. Wang".

[0054] (2) How did you meet the insurance agent? Former acquaintance? Pure stranger? Introduced by others?

[0055] (3) What is your personal situation, such as whether you are married and have children, income level, etc., and what is your personality?

[0056] (4) What is your attitude towards insurance? Do you recognize insurance? Do you have insurance configuration?

[0057] …

[0058] It should be understood that,Figure 4 The content shown is only an exemplary example to assist in understanding the relevant concepts, and is intended to provide an intuitive reference framework to help those skilled in the art better understand. However, the specific situation may be different when actually applied, so it should not be considered as a fixed and unchangeable limitation. Figure 4

[0059] The scenario introduction generating agent 103 is used to generate a dialogue scenario for customer profiling in the insurance industry. The scenario introduction generating agent 103 uses advanced GPT large model technology to generate realistic simulated dialogue scenarios based on customer profiling. It not only captures the basic information of the customer, but also deeply mines their implicit needs and personality characteristics through deep learning algorithms, creating a realistic and challenging simulation environment.

[0060] It is worth noting that in the context of the insurance industry, the scenario introduction generating agent 103 can identify and strategically hide certain real information of the customer, such as personality tendencies, lifestyle habits, etc., in order to design more realistic and variable dialogue scenarios. For example, even if the agent knows that the customer is actually introverted, it may set a scenario that requires extroverted interaction, forcing the trainee to respond in an incomplete information situation, exercising their adaptability and insight. In addition, the scenario introduction generating agent 103 will take into account the diversity and complexity of insurance products, ensuring that each generated scenario can specifically train the trainee's sales skills and problem-solving ability for specific products or services. In this way, the trainee can face various possible challenges in a safe environment, preparing them for real situations encountered in actual work.

[0061] In the embodiments of the present application, the scenario introduction generating agent 103 realizes the process of generating realistic simulated dialogue scenarios based on customer profiling based on the GPT large model as shown in Figure 5

[0062] Step S5a: Use NLP technology to analyze the customer profile and identify and extract key information points of the virtual customer; use machine learning algorithms to extract deep features from the customer profile that contain the implicit needs and personality characteristics of the virtual customer; input the extracted key information points and deep features of the virtual customer into the GPT model for training, so that it can understand and simulate the behavior and language style of the customer.

[0063] NLP technology refers to Natural Language Processing technology, which is a branch of artificial intelligence and linguistics, and is committed to enabling computers to understand, interpret and generate human language content. The key information points extracted through NLP technology include personal identity, relationship with the agent, personal situation, attitude towards insurance, etc. ​​

[0064] The machine learning algorithm is used to extract deep features containing the implicit needs and personality characteristics of the virtual customer from the customer portrait. The process includes selecting features related to implicit needs and personality characteristics from the original data of the customer portrait, such as purchase frequency, product preference, feedback content, interaction frequency, etc. According to the selected features, the historical data is used to train the machine learning model, such as decision tree, random forest, neural network, etc. The deep learning model can identify the implicit needs and personality characteristics from the features. It should be understood that implicit needs refer to needs that are not directly expressed or even not realized by the customer, which are hidden in the customer's behavior, preferences and feedback, and need to be discovered through analysis and mining. Generally speaking, implicit needs include preferences and interests, potential problem solutions, and future demand forecasts. Preferences and interests refer to preferences for certain products or services that are not explicitly expressed, potential problem solutions refer to problems that customers face but have not yet sought solutions, and future demand forecasts refer to predicting future possible needs by analyzing the customer's current behavior.

[0065] Step S5b: According to the features of the customer portrait, one or more dialogue frameworks are constructed, which cover the interaction paths of different business scenarios; in each dialogue framework, the trained GPT model is used to generate dialogue content consistent with the customer's characteristics.

[0066] For ease of understanding, the following three dialogue frameworks and corresponding interaction paths are used as examples for illustration.

[0067] Dialogue framework one: customer consultation and service.

[0068] Business scenario: the customer has questions about the insurance product and needs to consult the specific insurance clauses, fees, claim process, etc.

[0069] Interaction path:

[0070] 1. The customer raises specific questions.

[0071] 2. The virtual dialogue system provides accurate information and solutions according to the customer's questions.

[0072] 3. The customer provides feedback according to the information provided, which may require further explanation or help.

[0073] 4. The virtual dialogue system provides further help or solutions according to the customer's feedback.

[0074] Dialogue Content Generation Method: Utilize the natural language processing capabilities of the GPT model to generate detailed answers to specific customer questions, such as insurance clause explanations, cost calculations, etc. By simulating real conversations, the model is trained to provide more personalized and professional services, for example: "Hello, regarding the health insurance you mentioned, it covers hospitalization costs, surgery costs, etc. The specific terms are as follows...".

[0075] Dialogue Framework Two: Claim Handling

[0076] Business Scenario: Customers need to submit a claim or consult the claim process.

[0077] Interaction Path:

[0078] 1. The customer raises a claim requirement or question.

[0079] 2. The virtual dialogue system guides the customer through the claim process, providing necessary guidance and information.

[0080] 3. The customer provides the necessary claim information and documents according to the guidance.

[0081] 4. The virtual dialogue system confirms the information and submits the claim application.

[0082] Dialogue Content Generation Method: Use the GPT model to generate guidance dialogues for the claim process, such as: "To help you complete the claim quickly, please provide the following information: date of visit, hospital, and insurance policy number."

[0083] Dialogue Framework Three: Sales and Recruitment Training

[0084] Business Scenario: Insurance agents need to improve their sales and recruitment capabilities, and conduct simulated training through a virtual dialogue system.

[0085] Interaction Path:

[0086] 1. The agent enters the virtual dialogue system and selects a sales or recruitment scenario.

[0087] 2. The system generates a simulated customer dialogue based on the selected scenario.

[0088] 3. The agent interacts with the simulated customer to try to achieve sales or recruitment goals.

[0089] 4. The system provides feedback and suggestions for improvement based on the agent's performance.

[0090] Dialogue Content Generation: Use the GPT model to generate simulated customer dialogues that simulate real sales scenarios, such as: "Hello, I have been considering purchasing a life insurance policy recently. Can you introduce it to me?" The system provides feedback based on the agent's response to help them improve their communication skills and sales strategies.

[0091] Step S5c: Obtain the evaluation and feedback information of the user on the dialogue scene during the dialogue process, and optimize the GPT model through multiple rounds of iteration according to the evaluation and feedback information.

[0092] According to the collected user feedback and the results of the automatic scoring model, the GPT model is optimized, including adjusting the parameters of the model, improving the structure of the model, or adding more training data. The parameters of the GPT model that can be optimized include but are not limited to: learning rate, batch size, optimizer, regularization parameter, loss function weight, sequence length, context length, etc. The learning rate is a parameter that controls the step size of parameter updates during the training process of the model, and the learning process can be accelerated or slowed down by adjusting the learning rate. The batch size determines the amount of data input into the model each time the model is trained, and adjusting the batch size can affect the convergence speed and stability of the model. Optimizers such as AdamW have different effects on the training efficiency and final performance of the model. Regularization parameters such as dropout rate are used to prevent overfitting of the model, and the generalization ability of the model can be balanced by adjusting these parameters. In multi-task learning, the loss function weights of different tasks can be adjusted to optimize the performance of the model on each task. For the GPT model, the sequence length determines the length of the text that the model can process, and adjusting the sequence length can affect the model's ability to process long text. In multi-round dialogue, the context length determines the amount of dialogue history information that the model can remember, which is crucial for generating coherent dialogue.

[0093] The training dialogue generation agent 104 is trained to generate insurance business dialogue content based on the customer portrait simulated by the customer portrait generation agent 101, the customer portrait polished by the custom portrait check and polishing agent 102, and the dialogue scene generated by the context introduction generation agent 103 as input.

[0094] The training dialogue generation agent 104 uses advanced GPT large model technology to build the entire dialogue process and defines the way of interaction. When playing a specific role, GPT not only clarifies its focus, but also carefully designs the way of communication to facilitate more effective information exchange. In this way, we aim to provide a more personalized and efficient service experience. In the insurance industry, effective communication is the key to building customer trust and facilitating transactions. The training dialogue agent 104 simulates real sales dialogue to help students learn how to guide the dialogue, how to handle customer objections, and how to effectively promote products. In addition, the agent can provide feedback and suggestions to help students continuously improve their communication skills.

[0095] In the embodiments of the present application, the implementation process of the training dialogue generation agent 104 is as follows:

[0096] First, the GPT model-based training dialogue generation agent 104 preprocesses and extracts features from data such as customer profiles and dialogue scenario content. It should be understood that these inputs can be multi-modal data, such as text, speech, or other forms of data. Preprocessing includes but is not limited to text cleaning, tokenization, stop word removal, etc. to facilitate better understanding and processing by the model. Feature extraction refers to extracting key features such as keywords, phrases, semantic roles, etc. from the preprocessed data, which are used for subsequent dialogue generation. These comprehensive inputs make the training dialogue more realistic and allow students to practice and improve their sales skills in a risk-free environment. At the same time, the agent can also provide immediate feedback and guidance based on the student's performance, so that they can grow into a good insurance agent more quickly.

[0097] Second, the GPT model understands the customer's intentions and needs based on the input features and context information; uses the language patterns and structures learned in its training to build dialogue content that conforms to the role and scenario and outputs.

[0098] It is worth noting that in the insurance industry, language patterns and structures generally refer to the expressions, sentence patterns, professional terms, and dialogue processes commonly used in conversations. These patterns and structures can make the conversation more fluent, professional, and consistent with industry characteristics. For example:(1) Use of professional terms: The insurance industry has its own professional terms, such as "premiums", "policies", "claims", "risk assessment", etc. The GPT model will learn to use these professional terms during the training process, so as to accurately use them when generating conversations.(2) Question and answer mode: In the insurance consulting scenario, there is usually a mode of customer inquiry and insurance consultant answer. For example, the customer may ask: "What does this insurance product cover?" And the insurance consultant needs to give specific coverage content and terms. The GPT model will learn this question and answer mode and imitate this structure when generating conversations.(3) Risk assessment and recommendations: Insurance consultants will conduct risk assessments based on customer situations and provide appropriate insurance recommendations when conversing with customers. The GPT model will learn this language pattern of assessment and recommendations to ask for personalized insurance needs in the conversation.(4) Claims process explanation: In the claims process, insurance consultants need to explain the claims process and required materials to customers. The GPT model will learn this language pattern of process explanation to simulate the real reactions of customers and ask questions about the claims steps.(5) Compliance and standardization: Conversations in the insurance industry need to comply with strict compliance and standardization requirements. The GPT model will learn these requirements and ensure that the generated content complies with laws and regulations and company policies.(6) Customer service language: The insurance industry emphasizes customer service, so the GPT model will simulate various personalities of customers to train students to learn how to use polite, empathetic, and professional language to communicate with customers to improve customer satisfaction. Through the learning of these language patterns and structures, the GPT model can generate fluent, natural, and professional conversation content in the insurance industry, meet the needs of customers and provide high-quality services.

[0099] The dialogue assistant agent 105 is used to provide corresponding reply suggestions according to the questions raised by the virtual customer in the generated insurance industry dialogue scenario, combined with the customer portrait, scene description and context of the dialogue.

[0100] The customer portrait includes the customer's basic information, personality characteristics, consumption habits, etc., the scene description refers to the background information when the customer asks questions, such as pre-purchase consultation, after-sales service or questions in product use, etc., and the context information refers to the previous communication records of the customer, which is crucial to understanding the customer's current questions.

[0101] The script assistant agent 105 uses a GPT model to provide real-time response suggestions for the trainee. The GPT model is a pre-trained language model based on the Transformer architecture, which learns the general patterns and structures of language by pre-training on large-scale text data. The GPT model is fine-tuned using customer profiles, scenario descriptions, and context data, allowing it to learn how to predict the most appropriate response approach and suggestions based on the input customer data. For example, if a customer expresses doubts about a certain insurance product, the script assistant agent will suggest that the trainee first understand the customer's concerns and then provide relevant information and data to alleviate the customer's doubts. Or if the customer is sensitive to price, the agent will suggest that the trainee emphasize the value and long-term benefits of the product rather than just focusing on price, and so on. In addition, the script assistant agent also teaches the trainee some practical script skills and strategies, such as how to use open-ended questions to guide the conversation, how to use affirmative language to enhance persuasiveness, and so on. These skills and strategies can help the trainee better communicate with customers and improve sales results.

[0102] After the GPT model is trained into a script assistant agent, it can generate response ideas and suggestions based on new customer questions and feedback, relying on the model's generation capabilities. That is, when a customer asks a new question or provides feedback, the script assistant agent will analyze the input data and generate one or more response suggestions. These suggestions can be direct text responses or response ideas and frameworks for further refinement by human customer service representatives. In addition, an evaluation mechanism can be set up to evaluate the effectiveness of the responses based on user satisfaction surveys, response success rates, and other indicators, and to further optimize the GPT model accordingly.

[0103] In some examples, the training process of the script assistant agent based on the GPT model includes:

[0104] Step 1. Collect customer profiles, scenario descriptions, and context information, and preprocess these data as the dataset for model training. Preprocessing includes data cleaning, formatting, and splitting the dataset into training and test sets.

[0105] Step 2. Input the preprocessed training set into the GPT model for fine-tuning. During the fine-tuning process, a loss function such as cross-entropy loss function is used to calculate the difference between the model's predicted value and the actual value, and an optimization algorithm such as Adam or SGD is used to update the model's weights to minimize the loss function.

[0106] Step 3. After the model training is completed, the GPT model obtained by training is tested and evaluated using the test set, and the model indicators such as F1 score, precision, or recall are used for evaluation. Based on the evaluation results, it is determined whether further adjustment of the model's hyperparameters is needed.

[0107] Intelligent ending judgment The intelligent ending judgment agent 106 is configured to dynamically generate a corresponding conversation ending according to the context of the conversation, the customer emotional features and demand features, and the current interaction state.

[0108] In the embodiments of the present application, the training process of the intelligent ending judgment agent 106 based on the GPT model includes:

[0109] Step 1. Collect and preprocess data such as context, customer emotional features and demand features, and current interaction state. Preprocessing includes data cleaning, formatting, and splitting the dataset into training and test sets.

[0110] Step 2. Use the preprocessed training set to input the GPT model for fine-tuning, so that it learns how to identify the natural end point of the conversation and provide appropriate conversation ending methods according to the input conversation context, emotion and demand. During the fine-tuning process, a loss function (such as cross-entropy loss function) is used to calculate the difference between the model's predicted value and the actual value, and an optimization algorithm such as Adam or SGD is used to update the model's weights to minimize the loss function.

[0111] In some examples, the process of the GPT model identifying the natural end point of the conversation includes:

[0112] First, the GPT model processes the input sequence, which is converted into a numerical form that the model can understand, i.e., through tokenization, the text is converted into a sequence of index numbers. If the length of the text exceeds the maximum sequence length set by the model (e.g., 1024), only the first 1024 tokens are kept and the remaining tokens are deleted; if the length is insufficient, special markers (such as <|endoftext|>) are used to fill in.

[0113] Second, the GPT model uses a self-attention mechanism to process the context information in the input sequence. This mechanism allows the model to generate the next word based on all the words in the input sequence, not just the previous word. The self-attention mechanism calculates the weight score of each word in the input sequence to determine the influence of each word on generating the next word.

[0114] Subsequently, based on the current context, the GPT model generates a probability distribution of the next word. This distribution is based on the language patterns learned by the model during training, and the model predicts the probability of each possible next word.

[0115] Next, the GPT model selects an output token from the probability distribution. This selection can be done through different strategies, such as greedy search (always choosing the word with the highest probability), Beam Search (keeping multiple candidate word sequences and choosing the one with the highest score), or sampling (randomly choosing the next word according to the probability distribution).

[0116] Next, the generated word is added to the input sequence, and the above steps are repeated until a complete sentence is generated or a set generation length is reached. This process continues until the model identifies a natural end point of the conversation, i.e., the generated output token meets the stopping condition. The stopping condition for the GPT model to identify the end of the conversation can be a preset generation length, a specific end token (such as a period, question mark, etc.), or the probability of the end token in the probability distribution of the next word predicted by the model being significantly higher than that of other words. In this way, the model can determine that the conversation has naturally ended.

[0117] Step 3. After the model training is completed, the GPT model trained is tested and evaluated using the test set, and the model indicators such as F1 score, precision or recall are used for evaluation. According to the evaluation results, it is decided whether to further adjust the hyperparameters of the model.

[0118] It is worth noting that traditional procedural endings are usually based on fixed logic and rules, such as the conversation reaching a certain number of turns, the appearance of specific keywords, or preset time limits. This ending method lacks a deep understanding of the context of the conversation and the perception of the user's emotions, so it may not be able to adapt to complex conversation scenarios and changes in user needs. For example, if the user raises a new question or shows new needs in the conversation, the procedural ending may not be able to capture this point, resulting in the conversation ending when the user still needs help. The agent used in this invention uses advanced GPT large model technology, which can dynamically generate appropriate endings according to the context of the conversation, the emotions and needs of the customer, and the current interaction state. This way not only sounds more natural and smooth, but also better meets the expectations of customers and improves the overall service experience.

[0119] The customer analysis agent 107 is used to trace and analyze the virtual customer's problem characteristics after the conversation ends by combining the customer portrait, scene description and conversation record.

[0120] The customer analysis idea agent 107 based on the GPT model uses customer portraits, scene descriptions, and dialogue records for model fine-tuning training. Using the trained GPT model, combined with customer portraits, scene descriptions, and dialogue records, the key information and potential needs of the customer are identified. According to historical data and industry knowledge, the GPT model can predict the customer's possible questions and concerns, helping the agent to prepare in advance. Finally, the analysis results are organized to generate a report, which includes the customer's key information, potential needs, and predicted questions. In addition, the GPT model uses the feedback of the agent and the subsequent interaction data of the customer for continuous learning and optimization of the model, to improve the accuracy of the prediction and the relevance of the suggestions.

[0121] The drill evaluation agent 108 is used to generate evaluation information for training drills based on dialogue records after the dialogue ends.

[0122] Specifically, the GPT model is fine-tuned using annotated dialogue record data, and the model can generate evaluation information based on the dialogue content. Using the fine-tuned GPT model, input the dialogue record, and the model generates evaluation information step by step through self-recurrence, and controls the diversity and accuracy of the generated evaluation information through beam search and other techniques. The generated evaluation information is checked for grammar, logical coherence, and other checks, and the generated evaluation information is compared with actual evaluations to collect feedback for further optimization of the model. Finally, the generated evaluation information is organized into a report and provided to the agent or relevant personnel, which can be displayed in the form of charts, dashboards, and other forms for more intuitive display.

[0123] It should be understood that the regression generation is a text generation method that relies on the model's ability to predict the next word based on the current sequence of generated words. In the GPT model, this generation method is implemented through the following steps: input the dialogue record as an input sequence into the model; the model uses self-attention mechanism to understand the context and predict the next most likely word; the model generates text word by word, each step relying on the previously generated words; this process is repeated until a complete evaluation information is generated, or the preset word limit is reached. Beam search is a technique commonly used when generating text, which retains multiple candidate words (forming a "beam") at each step and considers the combined probability of these candidate words. This method can improve the quality and relevance of the generated text, as it considers multiple possible word sequences rather than relying solely on the single highest probability word. By adjusting the size of the beam, the exploration level during generation and the quality of the final result can be controlled. Therefore, by combining self-recurrence generation and techniques to control the quality of generation, the GPT model can generate evaluation information that is consistent with the dialogue content and has a certain accuracy and diversity. The application of these techniques enables the model to better adapt to different evaluation needs and scenarios.

[0124] In the above, the intelligent agents of the intelligent rehearsal system in the embodiments of the present application are explained in detail. In the following, the intelligent training method, device, medium, etc. will be further described in combination with the drawings.

[0125] As shown in Figure 6 , a flowchart of an intelligent rehearsal method based on the collaborative work of multiple intelligent agents in the embodiments of the present application is shown. The intelligent rehearsal method constructs multiple intelligent agents based on the GPT model, so as to collaboratively work to execute the following:

[0126] Step S61: Constructing structured data instructions based on insurance industry customer data to guide the large model to generate customer portraits of virtual insurance customers with different personality characteristics.

[0127] Step S62: Checking the custom customer portraits generated in response to user operations, and polishing the customer portraits that lack necessary structured information.

[0128] Step S63: Generating a dialogue scene of the insurance industry according to the customer portrait.

[0129] Step S64: Based on the simulated customer portrait and dialogue scene as input, output the insurance business dialogue content.

[0130] Step S65: In the generated dialogue scene of the insurance industry, according to the questions raised by the virtual customer, combined with the customer portrait, scene description and the context of the dialogue, provide corresponding reply suggestions.

[0131] Step S66: According to the context of the dialogue, the emotional characteristics and demand characteristics of the customer, and the current interactive state, dynamically generate corresponding dialogue ending words.

[0132] It should be understood that the specific process of each module executing the above corresponding steps has been described in detail in the above system embodiments, and for the sake of brevity, will not be repeated here.

[0133] It should be noted that in the embodiments of the present application, the words such as "exemplary" or "for example" represent an example, illustration or description. Any embodiment or design scheme described as "exemplary" or "for example" in the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concept in a specific manner.

[0134] In the embodiments of the present application, "at least one" means one or more, and "multiple" means two or more. The "and / or" describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the following three cases: A exists alone, A and B exist together, and B exists alone, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the front and rear associated objects. "At least one of the following" or similar expressions means any combination of these items, including any combination of single item or multiple items. For example, at least one of a, b or c can represent a, b, c, a-b, a-c, b-c or a-b-c, where a, b and c can be single or multiple.

[0135] Figure 7 is a schematic block diagram of a computer device provided by the embodiments of the present application. As shown in Figure 7 , the computer device includes at least one processor 701, a memory 702, at least one network interface 703 and a user interface 705. The various components in the device are coupled together through a bus system 704. It can be understood that the bus system 704 is used to realize the connection communication between the components. In addition to including a data bus, the bus system 704 also includes a power bus, a control bus and a status signal bus. However, for the purpose of clear illustration, all kinds of buses are marked as the bus system in Figure 7 .

[0136] The user interface 705 can include a display, a keyboard, a mouse, a trackball, a click gun, a key, a button, a touchpad or a touch screen, etc.

[0137] It can be understood that the memory 702 can be a volatile memory or a non-volatile memory, and can also include both volatile and non-volatile memories. The non-volatile memory can be a read-only memory (ROM, Read Only Memory), a programmable read-only memory (PROM, Programmable Read-Only Memory), which is used as an external cache. By way of example but not limitation, many forms of RAM can be used, such as static random access memory (SRAM, Static Random Access Memory), synchronous static random access memory (SSRAM, Synchronous Static Random Access Memory). The memory described in the embodiments of the present application is intended to include but not limited to these and any other suitable categories of memory.

[0138] The memory 702 in the embodiment of the present application is used to store various categories of data to support the operation of the electronic terminal 700. Examples of these data include: any executable program for operating on the electronic terminal 700, such as an operating system 7021 and an application program 7022; the operating system 7021 contains various system programs, for example, a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application program 7022 can contain various application programs, for example, a media player (Media Player), a browser (Browser), etc., for implementing various application services. The XX method provided by the embodiment of the present application can be included in the application program 7022.

[0139] The method disclosed in the above embodiment of the present application can be applied to the processor 701 or implemented by the processor 701. The processor 701 can be an integrated circuit chip with a signal processing capability. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor 701 or an instruction in the form of software. The above processor 701 can be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor 701 can implement or execute the disclosed methods, steps and logic block diagrams in the embodiments of the present application. The general-purpose processor 701 can be a microprocessor or any conventional processor, etc. In combination with the steps of the accessory optimization method provided by the embodiment of the present application, the hardware decoding processor can be directly embodied to execute the above steps, or the combination of hardware and software modules in the decoding processor can be executed. The software module can be located in a storage medium, which is located in the memory. The processor reads the information in the memory and combines the hardware to complete the steps of the above method.

[0140] In the exemplary embodiment, the electronic terminal 700 can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), etc., for executing the above method.

[0141] According to the method provided by the embodiment of the present application, the present application further provides a computer program product, which comprises computer program code, when the computer program code runs on a computer, so that the computer executes the intelligent rehearsal method based on the cooperative work of multiple intelligent agents.

[0142] According to the method provided in the embodiments of the present application, the present application further provides a computer readable storage medium storing program codes, which, when executed on a computer, cause the computer to perform the above method.

[0143] The terms "component," "module," "system," and the like are used to generally refer to computer-related entities, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to being, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and / or a computer. By way of illustration, both an application running on a computing device and the computing device can be a component. One or more components can reside within a process and / or thread of execution and a component can be localized, partially localized, or distributed across two or more computers. Also, these components can execute from various computer readable media having various data structures stored thereon. The components can communicate by way of local and / or remote processes such as in accordance with a signal having one or more data packets (e.g., data from one component interacting with another component in a local system, distributed system, and / or across a network such as the Internet with other systems via the signal).

[0144] Those skilled in the art can clearly understand that the various illustrative logical blocks and steps described in connection with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0145] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0146] In several embodiments provided by the present application, it should be understood that the disclosed system, device and method can be implemented by other ways. For example, the above-described device embodiments are merely schematic, for example, the division of units is merely a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed ones can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0147] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, i.e., may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0148] In addition, the functional units in each embodiment of the present application can be integrated into one processing unit, or each unit can be physically present alone, or two or more units can be integrated into one unit.

[0149] In the above embodiments, the functions of each functional unit can be implemented by software, hardware, firmware or any combination thereof, in whole or in part. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions (programs). When the computer program instructions (programs) are loaded and executed on a computer, the flow or function according to the embodiments of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. Computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another, for example, computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media. The available media can be magnetic media (such as floppy disk, hard disk, magnetic tape), optical media (such as high-density digital video disc (digital video disc, DVD), or semiconductor media (such as solid state disk (solid state disk, SSD) and the like.

[0150] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0151] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0152] In summary, the present application provides an intelligent drill system, method, medium, device and program product based on the cooperative work of multiple agents. Different types of virtual customers are created through a large model, and multiple agents cooperate to form a perfect training system. The training effect standardization problem caused by artificial training is solved, the training efficiency is greatly improved, and all data generated during the training process can be perfectly retained for tracing and analysis. Through the large model, realistic and challenging dialogue scenarios are created. According to the context of the dialogue, the emotions and needs of the customer, and the current interaction state, appropriate closing words are dynamically generated to improve the natural flow and service experience of the dialogue. Through the large model, virtual customers are automatically generated, and users can also customize customer portraits and polish them, providing diverse interactive drill opportunities for colleges. At the same time, there is a dialogue process that provides real-time support and suggestions, helping students better understand and respond to customer needs. The entire dialogue record is also evaluated comprehensively to provide feedback and suggestions for improvement, further improving training effectiveness. Therefore, the present application effectively overcomes the shortcomings of the prior art and has high industrial utilization value.

[0153] The above embodiments only exemplarily illustrate the principles and effects of the present application, and are not intended to limit the present application. Any person skilled in the art can modify or change the above embodiments without departing from the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by those skilled in the art without departing from the spirit and technical idea disclosed in the present application should be covered by the claims of the present application.

Claims

1. An intelligent drill system based on the collaborative work of a plurality of intelligent agents, characterized in that, The method comprises the following steps: A customer portrait generation intelligent agent is used to construct structured data instructions based on insurance industry customer data to guide a large model to generate customer portraits of virtual insurance customers with different personality characteristics; The insurance industry customer data includes customer basic information, policy details, and claim record information; A self-defined portrait checking and polishing intelligent agent is used to check a self-defined customer portrait generated in response to a user operation and polish the customer portrait that lacks necessary structured information; wherein the checking content of the self-defined portrait checking and polishing intelligent agent includes: completeness checking, which is used to verify whether the customer portrait input by the user contains all necessary structured information, and generate a series of questions or prompts according to missing information points to guide the user to supplement the missing information, and check the customer portrait after the user supplements the information again for completeness; personal expression checking, which is used to check whether the customer portrait adopts a second-person expression; key information point checking, which is used to check that the customer portrait at least contains the following key information points: personal identity, relationship with an agent, personal situation, and attitude towards insurance; A scenario introduction generation intelligent agent is used to generate a dialog scene in the insurance industry according to the customer portrait; A training dialog generation intelligent agent is used to output insurance business dialog content based on the simulated customer portrait and dialog scene as input; A dialogue skill assistant intelligent agent is used to provide corresponding reply suggestions in the generated dialog scene in the insurance industry according to the questions raised by the virtual customer, in combination with the customer portrait, scene description, and context of the dialog; An intelligent end judgment intelligent agent is used to dynamically generate a corresponding dialog ending according to the context of the dialog, customer emotional characteristics and demand characteristics, and current interaction state. 2.The intelligent drill system based on the cooperation of multiple intelligent agents according to claim 1, wherein, The intelligent rehearsal system further comprises any one or both of the following intelligent agents: A customer analysis idea intelligent agent is used to trace and analyze the virtual customer's question characteristics in combination with the customer portrait, scene description, and dialog record after the dialog ends; A rehearsal evaluation intelligent agent is used to generate evaluation information of the training rehearsal according to the dialog record after the dialog ends. 3.The intelligent drill system based on the cooperation of multiple intelligent agents according to claim 1, wherein, The construction method of the customer portrait generation intelligence comprises: Collecting customer data in the insurance industry and constructing structured data instructions, and using the structured data instructions to pre-process and extract features from the collected customer data; the structured data instructions are used to convert raw data into a structured format suitable for machine learning training; The extracted features are input into a GPT model for model training to construct the customer portrait generation intelligent agent.

4. The intelligent drill system based on the collaboration of multiple intelligent agents according to claim 1, wherein, The construction method of the scenario introduction generation intelligent agent comprises: Using NLP technology to analyze the customer portrait, identify and extract key information points of the virtual customer; using a machine learning algorithm to extract deep features containing the implicit needs and personality characteristics of the virtual customer from the customer portrait; inputting the extracted key information points and deep features of the virtual customer into a GPT model for training, so that the GPT model can understand and simulate the behavior and language style of the customer; According to the characteristics of the customer portrait, one or more dialogue frameworks are constructed, which cover the interaction paths of different business scenarios; in each dialogue framework, a trained GPT model is used to generate dialogue content that meets the characteristics of the customer; Obtain the evaluation and feedback information of the user in the dialogue process, and optimize the GPT model through multiple rounds of iteration accordingly.

5. The intelligent drill system based on the collaboration of multiple intelligent agents according to claim 1, wherein, The training dialogue generation agent is constructed in the following manner: Receive the customer portrait of the virtual customer from the customer portrait generation agent, receive the polished customer portrait of the user from the custom portrait checking and polishing agent, and receive the simulated dialogue scene content from the scenario introduction generation agent, and pre-process and feature extract the received data; The GPT model understands the customer's intent and needs based on the input features and context information; it uses the language patterns and structures learned in its training to construct dialogue content that meets the role and scenario and outputs it; wherein the language patterns and structures of the insurance industry include: use of professional terms, inquiry and question and answer mode, risk assessment and suggestion, claim process description, compliance and standardization, customer service language.

6. An intelligent drill method based on the cooperative work of a plurality of intelligent agents, characterized by, Based on the GPT model, multiple agents are constructed to work together to perform the following: Based on the insurance industry customer data, structured data instructions are constructed to guide the large model to generate customer portraits of virtual insurance customers with different personality characteristics; the customer data is simulated customer data based on real data structure and has been desensitized; The insurance industry customer data includes customer basic information, policy details, and claim record information. Check the custom customer portrait generated in response to user operations and polish the customer portrait that lacks necessary structured information; wherein the custom portrait checking and polishing agent checks the customer portrait, including: completeness check, which verifies whether the customer portrait input by the user contains all necessary structured information, and generates a series of questions or prompts based on missing information points to guide the user to supplement the missing information, and checks the customer portrait after the user supplements the information again for completeness; person expression check, which checks whether the customer portrait uses second person expression; key information point check, which checks that the customer portrait contains at least the following key information points: personal identity, relationship with the agent, personal situation, attitude towards insurance; According to the customer portrait, a dialogue scene in the insurance industry is generated; Based on the simulated customer portrait and dialogue scene as input, output the insurance business dialogue content; In the generated dialogue scene of the insurance industry, according to the questions raised by the virtual customer, combine the customer portrait, scene description and context of the dialogue to provide corresponding reply suggestions in real time; According to the context of the dialogue, the emotional characteristics and demand characteristics of the customer, and the current interaction state, dynamically generate corresponding dialogue ending words.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to realize the intelligent rehearsal method based on the collaborative work of multiple agents in claim 6.

8. A computer program product, characterised in that, The computer program product comprises computer program code, and when the computer program code runs on a computer, the computer program code causes the computer to implement the intelligent rehearsal method based on the collaborative work of multiple agents as claimed in claim 6.

9. A computer apparatus comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program, when executed by the processor, causes the processor to perform the method of any one of claims 1 to 8. The processor executes the computer program to implement the intelligent rehearsal method based on the collaborative work of multiple agents as claimed in claim 6.

Images