Insurance inspection report interpretation method and device, electronic equipment and storage medium

Abstract

The invention provides an insurance inspection report interpretation method and device, electronic equipment and a storage medium, and relates to the technical field of data processing, and the method comprises the steps: collecting family insurance related multi-source data, carrying out the preprocessing and structural processing of the multi-source data, and generating a structural family risk and guarantee data set, and generating an insurance interpretation result and outputting the insurance interpretation result based on the structured data set and a user request collaborative decision through a hybrid agent decision engine which comprises coordinator agents and each agent performs asynchronous communication and knowledge sharing by means of message-oriented middleware, the problems that in the prior art, due to the fact that data are collected and processed manually, the workload is large, the error and omission rate is high, or due to the fact that a single big data automatic analysis framework is dependent on, a dynamic task matching mechanism is lacked, the intelligent decision-making and real-time response capacity is insufficient, and module collaboration is poor can be solved.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and in particular to a method and apparatus for interpreting insurance inspection reports, electronic equipment, and storage medium. Background Technology

[0002] The intelligent interpretation technology for family insurance review reports is a fusion of insurtech and artificial intelligence, primarily used for family risk protection assessment. It has evolved from traditional manual analysis to automated big data parsing. Current technology mainly includes two core solutions: first, manual interpretation, where professionals analyze key information such as the policy, policyholder, insured, and coverage amount to complete risk assessment and report generation; second, automated big data parsing, which systematically organizes policy data through rule engines and actuarial algorithms to automatically generate reports.

[0003] However, neither system has formed a multi-dimensional capability integration architecture, exhibiting significant limitations: manual information interpretation and processing is labor-intensive, averaging over 48 hours with an error rate of 3.2%; automatic big data analysis relies on static rule bases, making it difficult to adapt to dynamic policy adjustments, handle multi-dimensional cross-analysis scenarios, and lack the ability to quantitatively evaluate multiple solutions. Furthermore, due to the lack of an intelligent agent collaboration mechanism, existing systems suffer from problems such as a single decision-making path and lagging knowledge updates when dealing with complex asset portfolios of high-net-worth clients and policy updates, affecting service response speed and solution professionalism. Summary of the Invention

[0004] This disclosure provides a method, apparatus, electronic device, and storage medium for interpreting insurance review reports. Its main objective is to at least partially address one of the technical problems in the related art.

[0005] According to the first aspect of this disclosure, a method for interpreting insurance review reports is provided, including: Collect multi-source data related to family insurance; The multi-source data is preprocessed and structured to generate a structured family risk and protection dataset; Through a hybrid agent decision engine, based on the structured family risk and protection dataset and user requests, multiple agents with different functions make collaborative decisions to generate insurance interpretation results. Output the insurance interpretation results; The hybrid agent decision engine includes a coordinator agent, which dynamically allocates tasks to other agents based on task attributes. The agents communicate asynchronously and share knowledge through a message middleware.

[0006] Optionally, the collection of multi-source data related to family insurance includes: Key information is extracted from policy images using image recognition technology; Data can be integrated with the insurance business system through the application programming interface to obtain policy status and historical records in real time. The system receives family information and insurance needs data input by users through a user interaction interface.

[0007] Optionally, the preprocessing and structuring of the multi-source data includes: The data is cleaned, missing and outlier values ​​are removed, and the data format is standardized. Extract insurance-related entities and relationships from text and data to construct a structured risk and protection data model; A preliminary assessment of household risk is conducted based on an actuarial model, and a risk gap indicator is calculated.

[0008] Optionally, the plurality of intelligent agents with different functions include: Reactive parsing agents are used to quickly respond to standardized queries; A goal-driven analytical agent is used to perform deep analysis and generate multiple optimization schemes. A utility optimization agent is used to evaluate and select the best solution based on a preset utility function; Learning-adaptive agents are used to dynamically adjust decision-making models and parameters based on system operation data and user feedback.

[0009] Optionally, the output of the insurance interpretation result includes: Generate structured insurance interpretation reports; Render visual charts to illustrate coverage gaps and product comparison information; Generate a natural language summary to concisely explain the interpretation results; Provide an interactive Q&A interface for subsequent user queries.

[0010] Optional, also includes: Based on user feedback on the interpretation results and system decision logs, the rule base, model parameters, and utility function weights of each agent in the hybrid agent decision engine are dynamically optimized through online learning algorithms.

[0011] According to a second aspect of this disclosure, an insurance review report interpretation device is provided, comprising: The data collection unit is used to collect multi-source data related to family insurance. The generation unit is used to preprocess and structure the multi-source data to generate a structured family risk and protection dataset. The decision-making unit is used to generate insurance interpretation results by using multiple agents with different functions to make collaborative decisions based on the structured family risk and protection dataset and user requests through a hybrid agent decision-making engine. The output unit is used to output the insurance interpretation result; The hybrid agent decision engine includes a coordinator agent, which dynamically allocates tasks to other agents based on task attributes. The agents communicate asynchronously and share knowledge through a message middleware.

[0012] Optionally, the acquisition unit is also used for: Key information is extracted from policy images using image recognition technology; Data can be integrated with the insurance business system through the application programming interface to obtain policy status and historical records in real time. The system receives family information and insurance needs data input by users through a user interaction interface.

[0013] Optionally, the generating unit is also used for: The data is cleaned, missing and outlier values ​​are removed, and the data format is standardized. Extract insurance-related entities and relationships from text and data to construct a structured risk and protection data model; A preliminary assessment of household risk is conducted based on an actuarial model, and a risk gap indicator is calculated.

[0014] Optionally, the plurality of intelligent agents with different functions include: Reactive parsing agents are used to quickly respond to standardized queries; A goal-driven analytical agent is used to perform deep analysis and generate multiple optimization schemes. A utility optimization agent is used to evaluate and select the best solution based on a preset utility function; Learning-adaptive agents are used to dynamically adjust decision-making models and parameters based on system operation data and user feedback.

[0015] Optionally, the output unit is also used for: Generate structured insurance interpretation reports; Render visual charts to illustrate coverage gaps and product comparison information; Generate a natural language summary to concisely explain the interpretation results; Provide an interactive Q&A interface for subsequent user queries.

[0016] Optional, also includes: The optimization unit is used to dynamically optimize the rule base, model parameters, and utility function weights of each agent in the hybrid agent decision engine based on user feedback data on the interpretation results and system decision logs through an online learning algorithm.

[0017] According to a third aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect above.

[0018] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect above.

[0019] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0020] The insurance review report interpretation method, apparatus, electronic device, and storage medium disclosed herein collect multi-source data related to family insurance, preprocess and structure the multi-source data to generate a structured family risk and protection dataset, and then use a hybrid agent decision engine that includes a coordinator agent and where agents communicate asynchronously and share knowledge through message middleware to generate and output insurance interpretation results based on the aforementioned structured dataset and user requests. Therefore, it can solve the problems of existing technologies that rely solely on manual data collection and processing, resulting in large workloads and high error rates, or that rely on a single big data automatic parsing architecture that lacks dynamic task matching mechanisms, intelligent decision-making and real-time response capabilities, and poor module collaboration. It achieves efficient and accurate processing of family insurance data, precise adaptation of tasks and agents, and efficient collaboration among agents, thereby quickly generating insurance interpretation results that meet user needs, improving the efficiency, accuracy, real-time response capability, and system scalability of insurance interpretation.

[0021] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0022] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 A flowchart illustrating an insurance inspection report interpretation method provided in this embodiment of the disclosure; Figure 2 This is a schematic diagram of the structure of an insurance inspection report interpretation device provided in an embodiment of the present disclosure; Figure 3 A schematic block diagram of an example electronic device provided for embodiments of this disclosure. Detailed Implementation

[0023] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0024] The following description, with reference to the accompanying drawings, outlines an embodiment of the insurance inspection report interpretation method, apparatus, electronic device, and storage medium.

[0025] Figure 1 This is a flowchart illustrating an insurance inspection report interpretation method provided in an embodiment of this disclosure.

[0026] like Figure 1 As shown, the method includes the following steps: Step 101: Collect multi-source data related to family insurance.

[0027] In the embodiments of this disclosure, the core objective of step 101 is to acquire comprehensive basic data to support the intelligent interpretation of family insurance review reports. This is achieved by collecting multi-source data related to family insurance through a data acquisition mechanism adapted to multiple dimensions of family insurance data. The multi-source data covers various information carriers related to the entire lifecycle of a family insurance policy, including but not limited to visualized data corresponding to policy certificates, structured policy data stored in the insurance company's business system, and information related to the family protection needs and actual situation provided by the user. The data acquisition process is implemented through technical means that are compatible with different data formats and adaptable to different acquisition scenarios, so as to ensure that the data coverage meets the core requirements of family insurance review for information integrity. As one implementation method, data can be acquired by collecting policy images through visual processing technology, obtaining data by connecting with the insurance company's system based on standardized interfaces, or receiving family insurance-related data manually input by the user through an interactive interface.

[0028] By collecting multi-source data related to family insurance, the problem of incomplete information caused by a single data source can be avoided, ensuring that subsequent data processing can be carried out based on complete information, and providing reliable data support for generating accurate and comprehensive insurance interpretation results.

[0029] Step 102: Preprocess and structure the multi-source data to generate a structured family risk and protection dataset.

[0030] In the embodiments disclosed herein, the core of step 102 is to transform the collected multi-source family insurance data into a standardized data format suitable for subsequent intelligent decision-making. First, preprocessing operations are used to address potential missing, anomaly, or format inconsistencies in the multi-source data, ensuring basic data usability. Then, through structured processing, key information related to family insurance review (such as policyholder / insured identity information, policy sum insured, scope of insurance liability, family structure, and risk-related attributes) is extracted from the preprocessed data. The relationships between these information are then organized according to a preset data logic, ultimately generating a structured dataset integrating the core content of family risk and protection dimensions. This dataset must meet the requirements for data standardization and completeness in subsequent decision-making processes. As one implementation method, a rule engine can be used to set cleaning rules to handle data anomalies, named entity recognition technology can be used to extract key insurance entity information, and basic risk association logic can be combined to initially match family risks with existing protection information, forming a structured family risk and protection dataset.

[0031] Preprocessing eliminates the clutter and instability of multi-source data, preventing poor-quality data from affecting subsequent analysis; structured processing transforms non-standardized data into standardized information that the system can directly access, providing reliable data support for the hybrid intelligent agent decision engine to accurately carry out collaborative decision-making, and ensuring the efficiency and accuracy of subsequent insurance interpretation work.

[0032] Step 103: Using a hybrid agent decision engine, based on the structured family risk and protection dataset and user requests, multiple agents with different functions make collaborative decisions to generate insurance interpretation results.

[0033] The hybrid agent decision engine includes a coordinator agent, which dynamically allocates tasks to other agents based on task attributes. The agents communicate asynchronously and share knowledge through a message middleware.

[0034] In the embodiments of this disclosure, step 103 is the core decision-making step that transforms structured data and user needs into insurance interpretation results. Relying on a hybrid intelligent agent decision engine, and using a structured family risk and protection dataset as the basic information source, combined with specific user requests (such as protection gap analysis, policy information queries, etc.), multiple intelligent agents with differentiated functions collaboratively conduct decision-making calculations. In this hybrid intelligent agent decision engine, the coordinator intelligent agent, as the core scheduling unit, first analyzes task attributes (including task complexity, response time requirements, required information dimensions, etc.), and then dynamically matches and allocates tasks to other intelligent agents with suitable functions. Simultaneously, each intelligent agent establishes an asynchronous communication mechanism through a message middleware to achieve real-time transmission of data and intermediate results during the decision-making process. Furthermore, information exchange is achieved through a shared knowledge system, avoiding information silos in the decision-making process. Finally, through the collaborative calculation of multiple intelligent agents, an insurance interpretation result that meets the user's needs is generated. As one implementation method, if the user request is a simple policy coverage inquiry, the coordinator agent can assign the task to an agent that focuses on rapid response. If it is a deep analysis of family protection gaps, it can be assigned to an agent with deep computing capabilities. Each agent transmits analysis parameters and results through a message middleware and shares knowledge such as insurance terms and risk assessment benchmarks to ensure consistent decision-making logic.

[0035] By dynamically allocating tasks through coordinator agents, the resource waste and inefficiency caused by fixed task allocation are avoided. Combined with asynchronous communication and knowledge sharing of message middleware, information barriers between agents are eliminated, enabling multi-agent collaboration to adapt to the differences in user requests while ensuring the accuracy of decisions. This effectively solves the problems of insufficient decision-making focus and low response efficiency in existing technologies, providing core support for generating insurance interpretation results that meet user needs.

[0036] Step 104: Output the insurance interpretation result.

[0037] In the embodiments of this disclosure, the core of step 104 is to present the insurance interpretation results generated by the hybrid intelligent agent decision engine through an output mechanism adapted to different user needs and application scenarios, ensuring that the interpretation results can be delivered to the target objects (such as users, downstream systems, etc.) in a clear, understandable, and usable form. The output process needs to combine the content attributes of the interpretation results (such as structured data, analysis conclusions, and visual information) with the information acquisition habits of the target objects, selecting the appropriate output format. This requires meeting the data accuracy requirements in professional scenarios while also considering the intuitive information needs of ordinary users, ultimately achieving effective delivery and application of the insurance interpretation results. As one implementation method, the interpretation results can be provided to other systems in the form of a structured report, visualized in charts to show family protection gaps, or presented through natural language summaries to meet the different user needs for the interpretation results.

[0038] By adopting an adaptive output mechanism, the inefficiency of information transmission caused by a single output format is avoided. It can meet the precise data needs of professional scenarios and enable ordinary users to quickly understand and interpret the core information, effectively improving the practicality and acceptability of insurance interpretation results and ensuring the closed-loop value of the entire family insurance review report intelligent interpretation process.

[0039] The insurance review report interpretation method disclosed herein collects multi-source data related to family insurance, preprocesses and structures this data to generate a structured family risk and protection dataset, and then uses a hybrid agent decision engine, which includes a coordinator agent and where agents communicate asynchronously and share knowledge via message middleware, to collaboratively generate and output insurance interpretation results based on the structured dataset and user requests. Therefore, it solves the problems of existing technologies that rely solely on manual data collection and processing, resulting in high workload and error rates, or that rely on a single big data automatic parsing architecture lacking dynamic task matching mechanisms, intelligent decision-making and real-time response capabilities, and poor module collaboration. This method achieves efficient and accurate processing of family insurance data, precise adaptation of tasks and agents, and efficient collaboration among agents, thereby quickly generating insurance interpretation results that meet user needs and improving the efficiency, accuracy, real-time response capability, and system scalability of insurance interpretation.

[0040] As a specific implementation of this disclosure, based on the basic solution, the collection of multi-source data related to family insurance is further defined as follows: extracting key information from policy images through image recognition technology; connecting with the insurance business system through an application programming interface to obtain policy status and historical records in real time; and receiving family information and insurance demand data input by users through a user interaction interface.

[0041] Specifically, this implementation further refines and limits the collection methods on top of the basic multi-source data collection scheme for family insurance, in order to achieve multi-dimensional and full-scenario data coverage. Specifically, when collecting policy information using image recognition technology, for image materials such as scanned electronic copies of policies uploaded by users and photos of physical policies, a pre-trained visual recognition model (or OCR technology combined with insurance-specific key information extraction algorithms) is used to accurately identify and parse the text content in the images, extracting key insurance elements such as the policyholder's name, the insured's identity information, the sum insured, the scope of insurance liability, the coverage period, and the core content of exclusion clauses. Simultaneously, an image quality verification mechanism (such as clarity judgment and tilt correction) ensures the completeness and accuracy of the extracted information. When connecting with the insurance business system through an application programming interface (API), a standardized interface protocol (such as RESTful API) is used. The API establishes a stable data link with the insurance company's core business system and policy management platform, retrieving dynamic data in real time, such as the current validity status of target policies, historical payment records, past claims details, and policy amendment information. During data transmission, the API adheres to insurance industry data interaction standards, supporting common data formats such as JSON or XML to ensure real-time data acquisition and format compatibility. When receiving data through the user interaction interface, the API provides two interaction methods: structured forms and natural language input. The structured form includes basic family information fields such as family composition, age and health status of each member, annual family income, and existing types of insurance products, as well as the user's specific insurance needs options (such as focusing on critical illness protection, children's education fund planning, and retirement protection). If the user chooses natural language input (e.g., "I want to supplement my child's critical illness insurance"), the backend processing unit parses the input intent and transforms it into standardized demand data, ensuring the complete collection of personalized user information that the system cannot automatically obtain.

[0042] By combining three differentiated data collection methods—image recognition, API integration, and user interaction interfaces—this approach not only covers unstructured data from paper / electronic insurance policies but also ensures the real-time nature and accuracy of dynamic policy data. Furthermore, it supplements personalized information related to family scenarios and user needs, effectively avoiding data loss or lag issues caused by a single collection method. This provides more comprehensive and accurate raw data support for subsequent preprocessing and structuring of multi-source data.

[0043] As a specific implementation of this disclosure, based on the basic scheme, the preprocessing and structuring of the multi-source data are further defined, including: cleaning the data, handling missing and outlier values, and standardizing the data format; extracting insurance-related entities and relationships from the text and data, and constructing a structured risk and protection data model; and conducting a preliminary assessment of family risk based on an actuarial model and calculating risk gap indicators.

[0044] Specifically, the processing flow was further refined to improve data quality and usability. In the data cleaning stage, for the collected multi-source data (such as text extracted from image recognition, business data from API integration, and user input information), a combination of rule engines and statistical methods was used to handle missing values—numerical data such as age and sum assured were filled with the mean or median, while categorical data such as policyholder occupation and insurance type were filled with the mode or marked "pending confirmation." Outliers (such as sum assured values ​​far exceeding the normal range or logically contradictory payment records) were identified using the IQR (interquartile range) method and corrected or removed in accordance with insurance industry data standards. Simultaneously, data format standardization was implemented, unifying the premium unit to "yuan / year," the date format to "YYYY-MM-DD," and the policy status to standardized expressions such as "valid / invalid / suspended," eliminating processing obstacles caused by format differences. In the entity and relation extraction stage, a Named Entity Recognition (NER) model finely tuned for the insurance field is adopted to extract core entities such as policyholder, insured, insurance type name, sum insured, coverage period, and insurance liability from text data (such as policy terms descriptions and user natural language input). Through a relation extraction algorithm based on dependency parsing, the relationships between entities are determined (such as "insured - corresponding - policy" and "policy - includes - insurance liability"). Based on this, a structured risk and protection data model is constructed. This model includes three core modules: family dimension (member composition, income and liabilities), protection dimension (key information of each policy), and risk dimension (potential risk types). The data in each module are linked through entity IDs to form a complete data link. In the initial risk assessment stage, common actuarial models in the insurance industry (such as life tables and critical illness incidence tables) are introduced. These are combined with family structure (such as the age of members and the number of dependents), economic status (such as annual income and mortgage balance), and existing protection (such as the sum insured of each type of insurance) to calculate core risk gap indicators. For example, the death risk gap = necessary expenses of the family in the next 10 years (including living expenses and debt repayment) - the total sum insured of existing life insurance, and the critical illness risk gap = industry-recommended critical illness treatment costs - the total sum insured of existing critical illness insurance. The calculation results are incorporated into the structured data model to form a complete family risk and protection dataset.

[0045] Refined data cleaning ensures the accuracy and standardization of raw data, preventing inferior data from interfering with subsequent decisions; structured risk and protection data models enable the orderly organization and correlation of data, facilitating rapid access by hybrid intelligent agents; preliminary risk assessment based on actuarial models quantifies core risk gaps in advance, providing a basis for subsequent collaborative decision-making and effectively improving the efficiency and accuracy of the overall insurance interpretation process.

[0046] As a specific implementation of this disclosure, based on the basic scheme, the multiple intelligent agents with different functions are further defined as follows: a reactive parsing intelligent agent, used to quickly respond to standardized queries; a goal-driven analysis intelligent agent, used to perform in-depth analysis and generate multiple optimization schemes; a utility optimization intelligent agent, used to evaluate and select schemes based on a preset utility function; and a learning and adaptation intelligent agent, used to dynamically adjust the decision model and parameters according to system operation data and user feedback.

[0047] Specifically, the roles and implementation logic of the various intelligent agents participating in the collaboration are further refined. The reactive parsing agent pre-stores a standardized query rule base, which contains mappings between basic policy information (such as coverage amount, coverage period, and insurance liability) and structured dataset fields. When it receives a standardized query task assigned by the coordinator agent (such as "querying the coverage amount of a critical illness insurance policy" or "confirming the coverage period of a policy"), it directly locates the corresponding data fields and extracts information through keyword matching rules without complex calculations, with a response time controlled in milliseconds, ensuring efficient feedback for simple queries. The goal-driven analysis agent is equipped with a deep analysis model of the protection gap, an insurance product matching engine, and a plan planner. After receiving a deep analysis task (such as "optimizing the family protection gap"), it first calls the protection gap analysis model to calculate the gap based on family structure (member ages, income and liabilities), existing protection, and industry benchmark data. Then, it uses the product matching engine to filter suitable products based on semantic similarity of terms and user risk preferences. Finally, it generates multiple optimization plans (such as "purchasing additional child-specific insurance" or "replacing the main medical insurance") based on family budget constraints and protection priorities. Each plan includes specific product recommendations and coverage supplements. The process includes cost calculation; the utility optimization agent pre-defines a utility function based on insurance industry standards and user historical preferences, with coverage rate (weight 0.6), total premium cost (weight 0.3), and user acceptance (weight 0.1) as core indicators. After receiving multiple solutions generated by the goal-driven analysis agent, it substitutes the actual data of each solution to calculate the comprehensive utility score, sorts them from high to low, and marks the optimal solution and the reasons for its selection (e.g., "Solution B has a utility score of 89.2, which is optimal in terms of coverage and cost balance"); the learning adaptation agent collects two types of data in real time: system operation data (task execution time and decision accuracy of each agent) and user feedback data (explicit ratings of interpretation results and adoption of optimized solutions). The data is analyzed through an online incremental learning algorithm. If it is found that the accuracy of the reactive parsing agent in matching a certain type of new query decreases, its rule base field mapping is updated; if it is found that most users prefer to choose the lower-cost solution, the weight of "total premium cost" in the utility function of the utility optimization agent is adjusted to dynamically optimize the decision model and parameters of each agent.

[0048] By differentiating the functions of the intelligent agents, we can achieve rapid response to standardized queries while ensuring the professionalism of in-depth analysis and the richness of multiple solutions. The quantitative evaluation of the utility optimization agents avoids the subjectivity of solution selection, and the dynamic adjustment of the learning and adaptation agents allows the system to continuously adapt to user needs and changes in scenarios, which significantly improves the response efficiency, decision accuracy and long-term adaptability of the hybrid intelligent agent decision engine.

[0049] As a specific implementation of this disclosure, based on the basic solution, the output of the insurance interpretation results is further defined as follows: generating a structured insurance interpretation report; rendering visual charts to display coverage gaps and product comparison information; generating natural language summaries to concisely explain the interpretation results; and providing an interactive Q&A interface for subsequent user queries.

[0050] Specifically, the output format and implementation logic are further refined to adapt to different user needs and usage scenarios. When generating structured insurance interpretation reports, standardized data formats such as JSON or XML are used. The report is divided into three core modules: basic family protection information (policyholder, insured, existing policy list), risk assessment results (risk gap values ​​and risk levels for each dimension), and optimization plan details (recommended product name, coverage amount, premium, and supplementary protection direction). Each module's data is associated with a unique identifier ID, supporting direct access and data integration by downstream systems (such as insurance business platforms and user management systems). In the visualization chart rendering stage, data visualization is achieved using a front-end chart library (such as ECharts). For family protection gaps, bar charts are used to visually compare the difference between "existing coverage amount" and "recommended coverage amount." For comparisons of multiple optimization plans, radar charts are used to display the differences between plans from dimensions such as "coverage rate," "premium cost," and "tax benefits." The charts support zooming and data hovering for detailed viewing, allowing users to quickly capture key differences. When generating natural language summaries, a finely tuned language model adapted for the insurance field is used to extract core information from the interpretation results (such as "a family critical illness protection gap of 200,000 yuan" and "recommended to replace with medical insurance products that comply with the new policy"). The summary content is organized into short, conversational, and logically coherent sentences, with a length of no more than 300 words, ensuring that ordinary users can quickly understand the core conclusions. When providing an interactive question-and-answer interface, the interface is equipped with a natural language understanding (NLU) module, which supports users to input subsequent queries (such as "premium payment method for the recommended plan" and "details of the exclusion clauses of a certain insurance type"). After the NLU module analyzes the user's query intent, it connects with the system's built-in insurance knowledge base and the data of this interpretation result to generate accurate answers. The interface response time is controlled within 1 second to ensure smooth interaction.

[0051] Structured reports meet system integration needs, visual charts lower the barrier to information comprehension, natural language summaries adapt to the reading habits of ordinary users, interactive Q&A interfaces resolve follow-up questions, and multi-format outputs collaboratively cover different usage scenarios and user groups, effectively improving the practicality, readability, and user interaction experience of insurance interpretation results, and ensuring that the interpretation value is fully delivered.

[0052] As a specific implementation of this disclosure, based on the basic scheme, the embodiments of this disclosure further include: dynamically optimizing the rule base, model parameters and utility function weights of each agent in the hybrid agent decision engine through an online learning algorithm based on user feedback data on the interpretation results and system decision logs.

[0053] Specifically, the system's self-optimization logic needs to be further supplemented to ensure that the hybrid agent decision engine can continuously adapt to changing needs. In the data collection phase, two types of core information need to be acquired simultaneously: First, user feedback data on the interpretation results, including explicit feedback (such as 1-5 star ratings for the interpretation report, explicit choices of "adopt / disadvantage" for optimization solutions, and pain points mentioned in written evaluations) and implicit feedback (such as the duration of viewing the interpretation report, the number of clicks on a certain type of solution, and whether the gap in coverage chart was fully viewed). Feedback data is transmitted to the system in real time through the user interaction interface and stored in the feedback database. Each data entry is associated with a corresponding interpretation task ID to trace its source. Second, the system decision log, covering task allocation records of the coordinator agents (such as "task type - assigned agent - time consumed"), intermediate calculation results of each agent (such as the risk gap calculation process of the goal-driven analysis agent, and the score details of the utility optimization agent), knowledge graph call records, etc. The log is stored in a structured format (such as JSON) and retained for no less than 6 months.

[0054] The application of online learning algorithms needs to be differentiated for different optimization objectives: For the rule base optimization of reactive analytical agents, an incremental rule learning algorithm is adopted. If multiple groups of users report that "the results are inaccurate when querying 'waiting period'", the algorithm analyzes the text features of the corresponding query in the logs and the rule matching deviation, and automatically supplements or corrects the matching logic of the "waiting period" field in the rule base. For the model parameter optimization of objective-driven analytical agents (such as the "family income weight" in the risk gap analysis model), a mini-batch gradient descent algorithm is adopted, and the parameters are adjusted based on user adoption. If most high-income family users do not adopt the "high coverage plan", the influence weight of "income" on the recommended coverage in the model is reduced. For the utility function weight optimization of utility optimization agents, a multi-objective online learning algorithm is adopted. If the proportion of negative evaluations related to "premium cost" in user feedback increases, the weight of "total premium cost" in the utility function is dynamically increased (e.g., from 0.3 to 0.4). All optimization processes are equipped with a threshold verification mechanism. The optimization results are only formally applied to the agent when the amount of feedback data of a certain type reaches a preset threshold (such as more than 50 feedbacks for the same problem) or when the accuracy of the optimized model improves by more than 5%, thus avoiding frequent adjustments caused by random feedback.

[0055] By transforming user feedback and system logs into optimization momentum through online learning algorithms, the rule base, model parameters, and utility function weights of the hybrid agent decision engine can dynamically adapt to changes in user needs and actual application scenarios. This effectively avoids the problems of "fixed rules and rigid decisions" in traditional systems, continuously improves the accuracy of agent decision-making and user satisfaction, and ensures the long-term service capability of the system.

[0056] It should be noted that the embodiments of this disclosure may include multiple steps. For ease of description, these steps are numbered, but these numbers are not a limitation on the execution time slots or execution order between the steps; these steps can be implemented in any order, and the embodiments of this disclosure do not limit this.

[0057] Corresponding to the above-described method for interpreting insurance review reports, this disclosure also proposes an insurance review report interpretation device. Since the device embodiments of this disclosure correspond to the method embodiments described above, details not disclosed in the device embodiments can be referred to the method embodiments described above, and will not be repeated here.

[0058] Figure 2 This is a schematic diagram of the structure of an insurance inspection report interpretation device provided in an embodiment of this disclosure, as shown below. Figure 2 As shown, it includes: Data acquisition unit 21 is used to collect multi-source data related to family insurance; The generation unit 22 is used to preprocess and structure the multi-source data to generate a structured family risk and protection dataset. Decision unit 23 is used to generate insurance interpretation results by using multiple intelligent agents with different functions to make collaborative decisions based on the structured family risk and protection dataset and user requests through a hybrid intelligent agent decision engine. Output unit 24 is used to output the insurance interpretation result; The hybrid agent decision engine includes a coordinator agent, which dynamically allocates tasks to other agents based on task attributes. The agents communicate asynchronously and share knowledge through a message middleware.

[0059] The insurance review report interpretation device disclosed herein collects multi-source data related to family insurance, preprocesses and structures the multi-source data to generate a structured family risk and protection dataset, and then uses a hybrid agent decision engine that includes a coordinator agent and where agents communicate asynchronously and share knowledge through message middleware. Based on the structured dataset and user requests, it collaboratively makes decisions to generate and output insurance interpretation results. Therefore, it can solve the problems of existing technologies that rely solely on manual data collection and processing, resulting in high workload and error rates, or rely on a single big data automatic parsing architecture that lacks a dynamic task matching mechanism, insufficient intelligent decision-making and real-time response capabilities, and poor module collaboration. It achieves efficient and accurate processing of family insurance data, precise adaptation of tasks and agents, and efficient collaboration among agents, thereby quickly generating insurance interpretation results that meet user needs and improving the efficiency, accuracy, real-time response capability, and system scalability of insurance interpretation.

[0060] Furthermore, in one possible implementation of this embodiment, the acquisition unit 21 is also used for: Key information is extracted from policy images using image recognition technology; Data can be integrated with the insurance business system through the application programming interface to obtain policy status and historical records in real time. The system receives family information and insurance needs data input by users through a user interaction interface.

[0061] Furthermore, in one possible implementation of this embodiment, the generation unit 22 is also used for: The data is cleaned, missing and outlier values ​​are removed, and the data format is standardized. Extract insurance-related entities and relationships from text and data to construct a structured risk and protection data model; A preliminary assessment of household risk is conducted based on an actuarial model, and a risk gap indicator is calculated.

[0062] Furthermore, in one possible implementation of this embodiment, the plurality of intelligent agents with different functions include: Reactive parsing agents are used to quickly respond to standardized queries; A goal-driven analytical agent is used to perform deep analysis and generate multiple optimization schemes. A utility optimization agent is used to evaluate and select the best solution based on a preset utility function; Learning-adaptive agents are used to dynamically adjust decision-making models and parameters based on system operation data and user feedback.

[0063] Furthermore, in one possible implementation of this embodiment, the output unit 24 is also used for: Generate structured insurance interpretation reports; Render visual charts to illustrate coverage gaps and product comparison information; Generate a natural language summary to concisely explain the interpretation results; Provide an interactive Q&A interface for subsequent user queries.

[0064] Furthermore, in one possible implementation of this embodiment, such as Figure 2 As shown, it also includes: The optimization unit 25 is used to dynamically optimize the rule base, model parameters, and utility function weights of each agent in the hybrid agent decision engine based on user feedback data on the interpretation results and system decision logs through an online learning algorithm.

[0065] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of this embodiment, and the principle is the same, so it is not limited in this embodiment.

[0066] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0067] Figure 3 A schematic block diagram of an example electronic device 300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0068] like Figure 3As shown, the electronic device 300 includes a computing unit 301, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 302 or a computer program loaded from storage unit 308 into RAM (Random Access Memory) 303. The RAM 303 may also store various programs and data required for the operation of the electronic device 300. The computing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An I / O (Input / Output) interface 305 is also connected to the bus 304.

[0069] Multiple components in electronic device 300 are connected to I / O interface 305, including: input unit 306, such as keyboard, mouse, etc.; output unit 307, such as various types of displays, speakers, etc.; storage unit 308, such as disk, optical disk, etc.; and communication unit 309, such as network card, modem, wireless transceiver, etc. Communication unit 309 allows electronic device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0070] The computing unit 301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processor, controller, microcontroller, etc. The computing unit 301 performs the various methods and processes described above, such as the insurance review report interpretation method. For example, in some embodiments, the insurance review report interpretation method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 300 via ROM 302 and / or communication unit 309. When the computer program is loaded into RAM 303 and executed by the computing unit 301, one or more steps of the methods described above may be performed. Alternatively, in other embodiments, the computing unit 301 may be configured to perform the aforementioned insurance review report interpretation method by any other suitable means (e.g., by means of firmware).

[0071] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0072] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0073] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0074] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0075] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, and blockchain networks.

[0076] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0077] It's important to note that artificial intelligence (AI) is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. AI software technologies primarily include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.

[0078] The various numerical designations such as "first," "second," etc., used in this disclosure are merely for ease of description and are not intended to limit the scope of the embodiments of this disclosure, nor do they indicate a sequential order.

[0079] At least one of the features described in this disclosure can also be described as one or more, and multiple features can be two, three, four or more, and this disclosure does not impose any limitations. In the embodiments of this disclosure, for a technical feature, the technical features in that technical feature are distinguished by "first", "second", "third", "A", "B", "C" and "D", etc., and there is no sequential order or size order among the technical features described by "first", "second", "third", "A", "B", "C" and "D".

[0080] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0081] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for interpreting insurance inspection reports, characterized in that, include: Collect multi-source data related to family insurance; The multi-source data is preprocessed and structured to generate a structured family risk and protection dataset; Through a hybrid agent decision engine, based on the structured family risk and protection dataset and user requests, multiple agents with different functions make collaborative decisions to generate insurance interpretation results. Output the insurance interpretation results; The hybrid agent decision engine includes a coordinator agent, which dynamically allocates tasks to other agents based on task attributes. The agents communicate asynchronously and share knowledge through a message middleware.

2. The method according to claim 1, characterized in that, The collection of multi-source data related to family insurance includes: Key information is extracted from policy images using image recognition technology; Data can be integrated with the insurance business system through the application programming interface to obtain policy status and historical records in real time. The system receives family information and insurance needs data input by users through a user interaction interface.

3. The method according to claim 1, characterized in that, The preprocessing and structuring of the multi-source data includes: The data is cleaned, missing and outlier values ​​are removed, and the data format is standardized. Extract insurance-related entities and relationships from text and data to construct a structured risk and protection data model; A preliminary assessment of household risk is conducted based on an actuarial model, and a risk gap indicator is calculated.

4. The method according to claim 1, characterized in that, The multiple intelligent agents with different functions include: Reactive parsing agents are used to quickly respond to standardized queries; A goal-driven analytical agent is used to perform deep analysis and generate multiple optimization schemes. A utility optimization agent is used to evaluate and select the best solution based on a preset utility function; Learning-adaptive agents are used to dynamically adjust decision-making models and parameters based on system operation data and user feedback.

5. The method according to claim 1, characterized in that, The output of the insurance interpretation result includes: Generate structured insurance interpretation reports; Render visual charts to illustrate coverage gaps and product comparison information; Generate a natural language summary to concisely explain the interpretation results; Provide an interactive Q&A interface for subsequent user queries.

6. The method according to claim 1, characterized in that, Also includes: Based on user feedback on the interpretation results and system decision logs, the rule base, model parameters, and utility function weights of each agent in the hybrid agent decision engine are dynamically optimized through online learning algorithms.

7. An insurance inspection report interpretation device, characterized in that, include: The data collection unit is used to collect multi-source data related to family insurance. The generation unit is used to preprocess and structure the multi-source data to generate a structured family risk and protection dataset. The decision-making unit is used to generate insurance interpretation results by using multiple agents with different functions to make collaborative decisions based on the structured family risk and protection dataset and user requests through a hybrid agent decision-making engine. The output unit is used to output the insurance interpretation result; The hybrid agent decision engine includes a coordinator agent, which dynamically allocates tasks to other agents based on task attributes. The agents communicate asynchronously and share knowledge through a message middleware.

8. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-6.

Images