Claim knowledge base construction method and device, equipment and storage medium

Abstract

The invention relates to the technical field of big data and the technical field of financial science and technology, and discloses a claim knowledge base construction method, which constructs a claim knowledge extraction sub-base by obtaining insurance claim data, analyzing the data by using a big language model, generating a claim dictionary and clause paraphrasing. Analyzing adjustment standards of different insurance types, constructing an adjustment standard sub-library in combination with claim settlement standards and claim payment conditions, obtaining historical claim settlement cases, performing tagging management, constructing a claim settlement case sub-library, and pushing similar cases to form a similar case recommendation sub-library. The expert opinions are used for generating a standard operation process and establishing a standard operation process sub-library. And all the sub-libraries are integrated to form a complete claim knowledge base, and new insurance claim data are regularly acquired to realize dynamic updating of the sub-libraries. Through structured and modularized knowledge integration, intelligence and standardization of the claim settlement process are realized, the claim settlement processing efficiency is remarkably improved, the operation complexity is reduced, the fraud risk is reduced, and the customer experience is improved.

Description

Technical Field

[0001] The present invention relates to the fields of big data technology and financial technology, and in particular to a method, device, equipment and storage medium for constructing a claims knowledge base. Background Art

[0002] With the rapid development of the insurance industry, claims settlement has become a key link for insurance companies to improve customer service quality and operational efficiency. However, the following problems are common in the existing claims settlement process:

[0003] First, information fragmentation is serious. The claims process involves multiple data sources, including insurance terms, historical claims cases, customer information, compensation conditions, etc., but these data are usually scattered in different systems and platforms, making it difficult to integrate and use them efficiently. This causes claims personnel to search for information across platforms when handling cases, greatly increasing the complexity of operations.

[0004] Secondly, the claims processing cycle is long. Due to the cumbersome claims process, it often takes a long time to manually review a large amount of documents, verify information, and calculate the amount of compensation. This not only reduces the efficiency of claims, but also affects customer satisfaction and experience.

[0005] The existing claims process lacks a systematic fraud risk identification mechanism and cannot effectively screen and prevent fraudulent behavior. Especially when facing complex claims cases, relying on manual identification cannot fully detect fraud risks.

[0006] In addition, inconsistent interpretation of clauses also often occurs. Insurance clauses are usually professional and complex, and different claims adjusters are prone to deviations when understanding and applying the clauses, resulting in inconsistent claims results, further increasing the risk of claims disputes.

[0007] The cumbersome operating system and many breakpoints in the claims process also limit the smoothness of the claims process. The existing claims system often involves multiple manual operation steps and decentralized subsystems, which are complex to operate and lack standardized processes, resulting in many operational obstacles for both customers and claims personnel in the process of handling claims cases.

[0008] Although there are some solutions on the market that attempt to optimize the claims process through information technology, these technical solutions lack a systematic and standardized knowledge base design paradigm. Most solutions are often designed in a one-sided way, making it difficult to integrate comprehensive expert experience and provide structured, modular and systematic claims knowledge.

[0009] Based on the above-mentioned deficiencies in the existing technologies, there is an urgent need for a systematic and standardized method for constructing a claims knowledge base that can effectively integrate knowledge and experience in the claims field, improve the intelligence level of the claims process, and provide instant, comprehensive and accurate information support for claims decisions. Summary of the invention

[0010] The main purpose of the present invention is to provide a method, device, equipment and storage medium for constructing a claims knowledge base, aiming to solve the technical problems that the prior art lacks a systematic and standardized method for constructing a claims knowledge base, resulting in information fragmentation, low processing efficiency and lack of effective knowledge integration in the claims process.

[0011] To achieve the above object, the present invention provides a method for constructing a claims knowledge base, comprising:

[0012] Obtaining insurance claim data, wherein the insurance claim data includes insurance terms and claim records;

[0013] Use a large language model to parse insurance claims data, extract key information including claim standards, liability exemptions, and compensation conditions, generate a claims dictionary and clause interpretations, and build a claims knowledge extraction sub-library based on the claims dictionary and the clause interpretations;

[0014] Analyzing settlement standards of different types of insurance based on the insurance claim data, and building a settlement standard sub-library based on the claim standards of different types of insurance and the compensation conditions;

[0015] Obtain historical claims cases and manage them by labeling according to case type, loss type and customer type, and build a claims case sub-library based on the historical claims cases managed by labeling;

[0016] Push similar cases based on the historical claims cases managed by labeling and similarity thresholds, and build a similar case recommendation sub-library based on the similar cases;

[0017] Obtain expert opinions on insurance operating procedures, generate standard operating procedures based on the expert opinions, and build a standard operating procedure sub-library based on the standard operating procedures;

[0018] A complete claims knowledge base is formed based on the integration of all sub-databases, and the data in the sub-databases are updated by regularly acquiring new insurance claims data.

[0019] Furthermore, to achieve the above-mentioned purpose, the present invention provides a claim knowledge base construction device, comprising:

[0020] A data collection module, which obtains insurance claim data, including insurance terms and claim records;

[0021] The knowledge extraction module uses a large language model to parse insurance claims data, extract key information including claim standards, liability exemptions, and compensation conditions, generate a claims dictionary and clause interpretations, and build a claims knowledge extraction sub-library based on the claims dictionary and the clause interpretations;

[0022] A settlement standard analysis module, which analyzes settlement standards of different types of insurance based on the insurance claim data, and constructs a settlement standard sub-library based on the claim standards of different types of insurance and the payment conditions;

[0023] The case labeling module obtains historical claims cases and manages them by labeling according to case type, loss type and customer type, and builds a claims case sub-library based on the historical claims cases managed by labeling;

[0024] A similar case recommendation module pushes similar cases based on the historical claim cases managed by labeling and the similarity threshold, and builds a similar case recommendation sub-library based on the similar cases;

[0025] A standard operating process management module obtains expert opinions on insurance operating processes, generates standard operating processes based on the expert opinions, and constructs a standard operating process sub-library based on the standard operating processes;

[0026] The knowledge base integration and update module forms a complete claims knowledge base based on the integration of all sub-bases, and updates the data in the sub-bases by regularly acquiring new insurance claims data.

[0027] Furthermore, to achieve the above-mentioned purpose, the present invention also provides a computer device, which includes a memory, a processor, and a claims knowledge base construction program stored in the memory and executable on the processor, and the claims knowledge base construction program, when executed by the processor, implements the steps of the claims knowledge base construction method as described above.

[0028] Furthermore, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, on which a claims knowledge base construction program is stored, and when the claims knowledge base construction program is executed by a processor, the steps of the claims knowledge base construction method as described above are implemented.

[0029] Beneficial effects: The present invention relates to the fields of big data technology and financial technology, and discloses a method for constructing a claims knowledge base, which obtains insurance claims data, including insurance clauses and claims records, and uses a large language model to parse the data, extracts key information such as claims standards, liability exemptions and payment conditions, generates a claims dictionary and clause interpretations, and constructs a claims knowledge extraction sub-base. The method includes analyzing the settlement standards of different types of insurance, constructing a settlement standard sub-base in combination with the claims standards and payment conditions, obtaining historical claims cases, and labeling and managing them according to case type, loss type and customer type, constructing a claims case sub-base, pushing similar cases, and forming a similar case recommendation sub-base. Expert opinions are used to generate standard operating procedures and establish a standard operating procedure sub-base. By integrating all sub-bases, a complete claims knowledge base is formed, and new insurance claims data is regularly obtained to achieve dynamic updates of the sub-base. The present invention realizes the intelligentization and standardization of the claims process through structured and modular knowledge integration; through structured analysis of insurance claims data, a claims dictionary and settlement standard sub-library are constructed to solve the problems of information fragmentation and inconsistent interpretation of terms; through labeling management of historical cases and pushing similar cases, the processing efficiency and accuracy of complex cases are improved; by combining expert opinions to generate standard operating procedures, the standardization of claims operations is realized; by integrating various sub-libraries and regularly updating the knowledge base, the timeliness and availability of knowledge are ensured. Overall, the present invention effectively simplifies the claims process, significantly improves the efficiency of claims processing, reduces operational complexity, reduces fraud risks, and improves customer experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0031] Figure 1 A schematic diagram of an application environment of a method for constructing a claims knowledge base in an embodiment of the present invention;

[0032] Figure 2 It is a flowchart of an embodiment of a method for constructing a claims knowledge base of the present invention;

[0033] Figure 3 A schematic diagram of functional modules of a preferred embodiment of the apparatus for constructing a claims knowledge base of the present invention;

[0034] Figure 4 A schematic diagram of the structure of a computer device in one embodiment of the present invention;

[0035] Figure 5 FIG. 4 is another schematic diagram of the structure of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0036] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0037] The method for constructing a claims knowledge base provided by the embodiment of the present invention can be applied in Figure 1 In the application environment, the client communicates with the server through the network. The server can obtain insurance claim data, including insurance terms and claim records, through the client, parse the data using a large language model, extract key information such as claim standards, liability exemptions and payment conditions, generate a claim dictionary and terms interpretation, and build a claim knowledge extraction sub-library. The method includes analyzing the settlement standards of different types of insurance, building a settlement standard sub-library in combination with the claim standards and payment conditions, obtaining historical claim cases, and labeling them according to case type, loss type and customer type. The sub-library of claims cases is constructed, similar cases are pushed, and a similar case recommendation sub-library is formed. Expert opinions are used to generate standard operating procedures and establish a standard operating procedure sub-library. By integrating all sub-libraries, a complete claim knowledge base is formed, and new insurance claim data is obtained regularly to achieve dynamic updates of the sub-libraries. The present invention realizes the intelligence and standardization of the claims process through structured and modular knowledge integration; through structured analysis of insurance claims data, a claims dictionary and a settlement standard sub-library are constructed to solve the problems of information fragmentation and inconsistent interpretation of terms; by labeling historical cases and pushing similar cases, the processing efficiency and accuracy of complex cases are improved; by combining expert opinions to generate standard operating procedures, the standardization of claims operations is realized; by integrating various sub-libraries and regularly updating the knowledge base, the timeliness and availability of knowledge are ensured. Overall, the present invention effectively simplifies the claims process, significantly improves the efficiency of claims processing, reduces operational complexity, reduces fraud risks, and improves customer experience. Among them, the client can be but is not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices. The server can be implemented with an independent server or a server cluster consisting of multiple servers. The present invention is described in detail below through specific embodiments.

[0038] See also Figure 2 , Figure 2 This is a flowchart of an embodiment of a method for constructing a claims knowledge base provided by the present invention. It should be noted that although a logical sequence is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0039] like Figure 2 As shown, the method for constructing a claims knowledge base proposed by the present invention includes the following steps:

[0040] S10, obtaining insurance claim data, wherein the insurance claim data includes insurance terms and claim records;

[0041] In this embodiment, core data related to claims are obtained from different sources. Data sources may include the core business system of the insurance company, historical data repositories, policy information submitted by customers, etc. Usually, these data are scattered in different systems, and the data formats may be different. Through interfaces (such as APIs) or data import technology, these scattered data sources can be effectively integrated and unified into a processable structured data format.

[0042] Insurance claims data is a general term for various types of information involved in the claims process, usually including policy-related terms, customer information, accident or loss conditions, compensation conditions, historical claims records, etc. The following are possible types and descriptions of insurance claims data:

[0043] Policy Information:

[0044] Insurance terms: including all terms specified in the insurance policy, such as scope of liability, compensation conditions, exemption clauses, insured amount, premium payment plan, etc. This is the basic basis for claims.

[0045] Policy Number: A unique identifier for a policy that is used to associate it with a customer and their claims record.

[0046] Effective date and expiration date: Displays the validity period of the insurance contract to determine whether a claim occurs during the insurance liability period.

[0047] Insured amount and premium: the upper limit of the compensation amount stipulated in the policy and the insurance premium paid by the customer.

[0048] Insured person information: including the customer’s personal information, such as name, ID number, contact information, etc.

[0049] Claims record:

[0050] Historical claims events: records the claims requests that the insured has submitted, including the details of each case, such as claim type, loss amount, compensation amount, review status, etc.

[0051] Claim status: The processing status of the current claim case, such as under review, paid, rejected, etc.

[0052] Claims timeline: includes key time points such as the claim application date, review date, and payment date, used to track the claims process.

[0053] Claims decision and notes: The claims adjuster's review opinions and decision records, which may include additional instructions, special handling circumstances, etc.

[0054] Accident / Loss Information:

[0055] Accident details: Description of the accident in the claim case, including the type of accident (such as car accident, fire, medical incident, etc.), location and time of occurrence, etc.

[0056] Loss Amount: An estimate of the amount of property damaged or personal injury incurred, usually provided by a claims adjuster or outside loss assessor.

[0057] Loss type: such as personal injury, property damage, liability loss, etc., used to classify claims cases.

[0058] Loss vouchers: such as medical receipts, repair invoices, loss lists, etc., as evidence of the loss amount.

[0059] Responsibility identification information:

[0060] Responsible party information: If the accident involves third-party liability, record the responsible party's information, including detailed description of the liability determination.

[0061] Responsibility ratio: Based on the liability determination, the liability sharing ratio between the responsible party and the insured is clarified.

[0062] External appraisal report: such as an assessment report from a legal institution or a third-party expert, which serves as the basis for claims decisions.

[0063] Compensation related information:

[0064] Compensation Amount: The final compensation amount calculated based on the insurance terms and loss amount, which may include deductible, compensation ratio and other information.

[0065] Payment method: such as bank transfer, check payment, etc.

[0066] Payment recipient: The recipient of the compensation, usually the insured, but in some cases it may be a third party.

[0067] Obtain various types of insurance claims data including insurance clauses and claims records through the insurance company's internal management system (such as the insurance core system). Use the data interface (API) to connect the system and data source to regularly obtain the latest insurance clause updates and claims processing data. For different insurance products (such as auto insurance, health insurance, etc.), extract the corresponding clauses and claims records from the system to ensure the integrity and accuracy of the data.

[0068] Through the automated data collection of insurance clauses and claims records, the workload of manual data acquisition is reduced, effectively solving the problem of information fragmentation. Through standardized processing, data from different sources can be managed in a unified manner, laying a data foundation for subsequent intelligent analysis and knowledge extraction, and improving the efficiency and accuracy of data processing.

[0069] S20, using a large language model to parse insurance claims data, extract key information including claims standards, liability exemptions and compensation conditions, generate a claims dictionary and clause interpretations, and build a claims knowledge extraction sub-library based on the claims dictionary and the clause interpretations;

[0070] In this embodiment, a large language model is used to parse unstructured text data in insurance claims data. The model needs to have the ability to process natural language, understand the complex semantics of insurance clauses, and extract key information. The model needs to be able to accurately identify and process a large number of insurance clauses and claims records, and separate key information that is useful for claims. The large language model parses insurance claims data, mainly identifying and extracting key information such as claim standards, liability exemption clauses, and compensation conditions in the clauses. These data often exist in the form of unstructured text, and the model needs to understand and parse out useful information, such as specific compensation ratios, exemption conditions, etc.

[0071] The large language model processes the insurance clauses and automatically identifies the clauses involving the compensation standards. For example, the model can extract detailed information about the compensation ratio of medical expenses, the maximum compensation limit, etc. from the clauses. Liability exemption is a situation listed in the insurance clauses where the insurance company does not assume liability. Through the analysis of the large language model, the liability exemption scenarios in the clauses (such as drunk driving, intentional behavior, etc.) are automatically identified, and these scenarios are extracted and stored as structured information. The compensation conditions determine under what circumstances the insurance company will pay compensation. The large language model is responsible for parsing the compensation conditions in the clauses, such as whether it meets the insurance liability, whether the loss is within the insurance coverage, etc., and converting these conditions into information that can be processed automatically.

[0072] The extracted key information will be organized into a standardized dictionary. The purpose of the claims dictionary is to standardize the professional terms, clause interpretations and specific claims standards in the insurance field. The terms and terms in the dictionary will be standardized to ensure that claims personnel can quickly understand and reference them when using them. The large language model will generate corresponding interpretations for complex insurance clauses to ensure that claims personnel can quickly understand the content of the clauses. Interpretations can include explanations of certain professional terms or simplified explanations of complex clauses.

[0073] The extracted information such as claim standards, liability exemptions, compensation conditions, claim dictionary and clause interpretation will eventually be stored in the claim knowledge extraction sub-database. This sub-database will become a structured knowledge base, supporting claims personnel to quickly retrieve relevant clauses and rules when handling cases and help make claims decisions.

[0074] First, select a suitable large language model and perform fine-tuning training for the insurance field. The training data can include insurance clause text, historical claims records, and standard clause sets in the industry. The fine-tuning process focuses on improving the model's understanding of professional terms, clause structure, and compensation conditions, ensuring that the model can identify key information in insurance clauses.

[0075] During the parsing process, the large language model will parse the content of the insurance clauses sentence by sentence, and automatically extract information such as claim standards, liability exemption clauses, and compensation conditions. Based on the learned rules and contextual semantics, the model identifies which content belongs to the claim standards, which content belongs to the exemption conditions, and which conditions involve the compensation amount. The parsed information will be structured to form a clause database.

[0076] The parsed key information will be organized into a claims dictionary, which contains standardized terms and explanations. The large language model will further generate clause interpretations, simplifying and making complex clauses easier to understand, and forming clear interpretations. The dictionary and interpretations can be generated by automated tools so that claims personnel can directly quote them.

[0077] By storing the parsed and extracted claims dictionary and definitions in the claims knowledge extraction sub-base, a complete claims knowledge base is constructed. This sub-base can be queried through search tools, supporting claims personnel to quickly find the required clause information when handling cases.

[0078] By using a large language model to automatically parse insurance clauses, the parsing efficiency of insurance clauses can be significantly improved, avoiding the tedious process of manually processing clauses. The systematically extracted claims standards, liability exemptions, and payment conditions can ensure the integrity and consistency of information during the claims process, reducing the possibility of manual misunderstandings. The generated claims dictionary and clause interpretations can help claims personnel quickly understand complex clause contents, improving the accuracy and efficiency of claims decisions.

[0079] S30, analyzing settlement standards of different types of insurance based on the insurance claim data, and constructing a settlement standard sub-library based on the claim standards of different types of insurance and the compensation conditions;

[0080] In this embodiment, the insurance claims data of different types of insurance are analyzed to extract the settlement standards. Different types of insurance, such as auto insurance, health insurance, property insurance, etc., have large differences in compensation standards and processing procedures. Therefore, it is necessary to perform data analysis based on the terms, compensation conditions, loss types, etc. of each type of insurance to extract the settlement standards applicable to the type of insurance. These standards usually include compensation limits, deductibles, loss assessment methods, compensation ratios, etc. Through a comprehensive analysis of insurance terms, claims records, and historical cases, the settlement standards of different types of insurance in actual operations are extracted. The core of the analysis is to find the corresponding compensation conditions and claims standards for each type of insurance, so that the settlement standards are operational and accurate.

[0081] The claim standard is a standard set by the insurance company according to the terms of the policy to determine the amount of compensation, compensation conditions and scope of liability. Different types of insurance have different claim standards, such as the medical expense compensation ratio in health insurance and the vehicle repair cost compensation limit in auto insurance. The claim standard directly affects the accuracy of the settlement and the rationality of the compensation. The compensation conditions determine under what circumstances the insurance company will bear the compensation liability. The compensation conditions of each insurance type are different. For example, auto insurance may include accident liability division and deductible clauses, while health insurance may depend on the specific type of treatment. According to the compensation conditions of different insurance types, identify eligible claim records and clauses.

[0082] The settlement standard sub-library is a structured data storage system that summarizes the claim standards and payment conditions for each type of insurance. It ensures that the settlement rules for each type of insurance can be managed in a standardized manner and can be quickly called in actual claim operations. The design of the settlement standard sub-library can include the settlement model for each type of insurance, such as payment conditions, payment ratios, deductibles, and settlement logic. The purpose of the settlement standard sub-library is to provide accurate settlement basis for each claim.

[0083] By classifying the insurance company's claims data by type, the insurance clauses and historical claims records corresponding to different types of insurance products (such as auto insurance, health insurance, property insurance, etc.) are sorted out. The data of these insurance types are comprehensively analyzed using data analysis tools to extract the settlement standards unique to each insurance type. These standards include data fields such as payment conditions, payment limits, and deductibles. Based on the payment status in historical claims records, the application effect of the settlement standards in actual operations is analyzed to ensure that the extracted standards can reflect the actual claims processing situation.

[0084] The system matches the claim standards and compensation conditions of different insurance types, and identifies the claim standards applicable to each insurance type by parsing the insurance terms, such as the medical expense reimbursement ratio of health insurance and the vehicle repair expense compensation limit of auto insurance. For the compensation conditions of each insurance type (such as hospitalization requirements for health insurance or accident liability ratio for auto insurance), the system automatically extracts the corresponding compensation conditions and stores this information in the settlement standard sub-library.

[0085] According to the extracted settlement standards, a settlement standard model for each type of insurance is constructed. The settlement model contains information such as compensation conditions, compensation ratio, compensation limit and deductible. These models will be stored in the settlement standard sub-library so that the claims adjuster can quickly call the corresponding settlement standard when handling the case and calculate the compensation amount according to the specific circumstances of the case.

[0086] Through the above steps, the settlement standards of different types of insurance can be effectively managed in a systematic manner, reducing the manual calculation of compensation standards and deductibles during the claims process. The construction of the settlement standard sub-library enables claims personnel to quickly obtain settlement basis when handling cases, improving the accuracy and efficiency of settlement. After analysis, the claims standards and payment conditions of different types of insurance are stored in the sub-library, which facilitates the subsequent automated settlement process and reduces human errors and misunderstandings.

[0087] S40, obtaining historical claim cases, and managing them by labeling according to case type, loss type and customer type, and building a claim case sub-library based on the historical claim cases managed by labeling;

[0088] In this embodiment, a large number of historical claims cases are obtained from the insurance company's claims system. These case data may contain many years of claims records, covering information such as different types of insurance, compensation conditions, and liability determination. Data sources include the historical record database of the insurance claims system, external data sources (such as the industry claims sharing platform), and internal case management systems. Historical claims case data typically includes policy information, accident occurrence, compensation amount, review process, and other content. The data needs to be converted from unstructured or semi-structured text to a structured data form for subsequent processing and analysis.

[0089] Case type classification refers to dividing different claims cases into categories. For example, case types can include auto insurance, health insurance, property insurance, etc. Each claim case needs to be assigned a corresponding type label based on its case attributes.

[0090] Loss type classification: Loss type is the classification of cases according to the specific loss circumstances involved in the claim. For example, a car insurance case may involve vehicle loss, third-party property loss, personal injury, etc. The loss type of each claim case needs to be classified in detail in order to manage the claims data more accurately.

[0091] Customer type classification: Customer type labels are managed based on the type of the insured (customer). For example, customers can be divided into individual customers, corporate customers, etc. Through customer type labeling management, the claims behavior and characteristics of different customer groups can be better analyzed.

[0092] The purpose of label management is to classify historical claims cases for easy subsequent retrieval and analysis. The core of building a claims case sub-library is to classify, store and index cases according to the assigned labels. The case sub-library will become a structured database that supports claims personnel to quickly find similar historical cases when handling new cases and assist in decision-making.

[0093] First, the system imports claims case data in batches from the insurance company's historical claims database. The imported data may include policy information, payment records, accident details, review process, etc. After the data is imported, it needs to be cleaned to remove invalid, duplicate or erroneous data. The historical case data from different sources is standardized to ensure that the fields of each case are consistent, such as case ID, policy number, case type, payment amount, etc., to facilitate subsequent label management.

[0094] The system automatically assigns case type labels to each historical case (such as auto insurance, health insurance, property insurance, etc.). Automatic label assignment can be achieved through a rule engine or natural language processing technology based on case text.

[0095] Loss type label assignment: The system analyzes the loss details in each case and assigns labels based on the loss type (such as property damage, personal injury, etc.). The model can determine the category of the loss by analyzing the descriptive text in the claim record.

[0096] Customer type label assignment: Assign customer type labels to each case based on customer information (such as individual, enterprise, etc.). The system extracts customer type from policy information and automatically categorizes it.

[0097] Store labeled case data in a dedicated claims case sub-library. The sub-library needs to support the search function by tag to ensure that claims personnel can quickly find historical cases of the corresponding type. To improve search efficiency, the system will index case data and support multi-dimensional fast search. For example, claims personnel can search by case type, loss type, or customer type at the same time to find the most relevant historical cases.

[0098] Through the above steps, historical claims cases are managed in a systematic and structured manner, and label management can simplify complex case retrieval and decision support. Claims personnel can quickly find relevant historical cases and make reference decisions by searching for cases with the same label, greatly shortening the processing time of complex cases. At the same time, label management improves the efficiency of data organization and provides a solid foundation for subsequent intelligent analysis and prediction.

[0099] S50, pushing similar cases based on the historical claim cases managed by labeling and the similarity threshold, and building a similar case recommendation sub-library based on the similar cases;

[0100] In this embodiment, similarity analysis is performed based on historical claims cases that have been tagged and managed through labels (such as case type, loss type, customer type, etc.). By calculating the similarity between new claims cases and historical cases, cases with high similarity are identified. Here, the similarity threshold plays a decisive role. It defines the degree of similarity that must be achieved between new cases and historical cases. Only historical cases that exceed this threshold will be pushed to claims personnel for reference. Similarity can be calculated based on multiple dimensions, such as case type, loss type, claim amount, customer type, etc. The system calculates the similarity value between each historical case and the current case through a comprehensive comparison of different dimensions, and historical cases that meet the threshold conditions will be pushed out for reference by claims personnel.

[0101] When the system identifies historical claims cases that meet the similarity threshold, these cases will be stored in a special sub-library to form a similar case recommendation sub-library. This sub-library is mainly used to classify and store the pushed similar cases for subsequent rapid retrieval and decision-making reference. The content of the sub-library will be dynamically updated. Every time a new case is processed, the system will recommend and store relevant historical cases of the new case based on the similarity.

[0102] First, the system compares historical claims case data that is managed in a labeled manner. Each historical case has a clear label (such as case type, loss type, etc.). Through these labels, the system compares new cases in multiple dimensions. The system calculates similarity scores based on different label dimensions. For example, if the loss type of a case is "property loss", the system will prioritize comparing it with historical cases that also have "property loss". Similarity can be calculated using algorithms such as cosine similarity and Euclidean distance.

[0103] After calculating the similarity, the system compares the result with the set similarity threshold. Only when the similarity of the historical case reaches or exceeds this threshold will the system push the case to the claims adjuster for reference. When the similarity calculation of the new case is completed, the system will automatically extract the cases whose similarity reaches or exceeds the threshold from the historical case library, and push these cases to the claims adjuster in the form of a list or summary. The push content can include summary information of the historical case, such as case type, processing results, compensation amount, etc. The pushed similar cases will be dynamically stored in a separate sub-library. This sub-library is specifically used to save historical cases that meet the similarity threshold to ensure that these similar cases can be referenced again in the processing of subsequent cases. The content of the similar case recommendation sub-library will be continuously updated as new cases are processed and pushed. Whenever a new case meets the similar case push conditions, the content in the sub-library will be automatically updated and new similar case records will be added.

[0104] Through the above steps, historical cases similar to the current claim case can be automatically identified and pushed to the claims personnel in a timely manner. This not only speeds up the processing of complex cases, but also provides historical experience as a basis for decision-making, reducing the risk of claims. The construction of the similar case recommendation sub-library enables similar cases to be systematically stored and managed, facilitating subsequent rapid retrieval and reference, greatly improving the efficiency of claims and the accuracy of decision-making.

[0105] S60, obtaining expert opinions on insurance operation processes, generating standard operation processes based on the expert opinions, and building a standard operation process sub-library based on the standard operation processes;

[0106] In this embodiment, professional opinions and suggestions on the claims process are collected from experts in the insurance field. Expert opinions can involve best practices in the insurance company's internal claims process, optimization suggestions for operating steps, solutions to common problems, etc. This process can be achieved through interviews, questionnaires, expert meetings or seminars, etc., to ensure that the opinions collected cover different types of insurance and claims scenarios. Expert opinions usually exist in unstructured forms, such as oral discussions, written reports, etc. These unstructured opinions need to be converted into structured process information so that the system can process and store them later.

[0107] Based on the collected expert opinions, a standardized claims process is developed. The standard operating procedure is an optimized and standardized claims process that ensures that different claims personnel can follow consistent operating steps and specifications when handling claims cases. The complex claims process is transformed into a set of standardized operating procedures through the experience of experts. The operating procedures for each type of insurance and claim case type may be different, so it is necessary to design multiple operating procedures for different scenarios. These processes can include the following steps: preliminary review, accident investigation, liability determination, loss assessment, compensation calculation, etc. Each step in the process will list the required operations and precautions in detail.

[0108] The generated standard operating procedures are stored in the standard operating procedure sub-library for subsequent retrieval and call. The standard operating procedure sub-library is designed to convert the opinions and experience of experts into standardized data that is easy to query and operate, and to support claims personnel to quickly call the corresponding operating procedures in different scenarios through the system. The standard operating procedure sub-library can be updated according to new expert opinions or industry changes to ensure that the processes in the sub-library can adapt to the latest needs and regulations of the insurance industry.

[0109] Through the above steps, the experts' experience and opinions are systematically integrated into standard operating procedures, ensuring that claims personnel have a set of reliable standard operating procedures when handling cases. This not only improves the efficiency and consistency of claims operations, but also avoids claims delays or disputes caused by inconsistent procedures or improper operations. At the same time, the standard operating procedure sub-library can be dynamically updated to ensure that the operating procedures always comply with the latest industry regulations and actual needs.

[0110] S70, forming a complete claims knowledge base based on the integration of all sub-databases, and updating the data in the sub-databases by regularly acquiring new insurance claims data.

[0111] In this embodiment, various sub-databases (such as claims knowledge extraction sub-database, settlement standard sub-database, claims case sub-database, similar case recommendation sub-database, standard operating process sub-database, etc.) are integrated into a complete claims knowledge base. Each sub-database is independently constructed, covering claims knowledge of different dimensions (such as standard terms, historical cases, settlement standards, expert opinions, etc.), and the purpose of integration is to centrally manage the information of these sub-databases so that claims personnel can query and use all relevant knowledge in one system.

[0112] During the integration process, it is necessary to ensure the consistency and relevance of data between sub-databases. For example, the compensation conditions in the settlement standard sub-database need to match the case compensation data in the historical claims case sub-database to ensure that when processing claims cases, the system can call relevant knowledge across databases and provide complete decision support.

[0113] Data in the insurance claims industry is constantly updated as new cases occur and clauses change. The system can regularly obtain new insurance claims data, such as new insurance clauses, claims records, expert opinions, etc., to ensure that the content in the knowledge base is always up to date. Updated data can come from internal systems (such as insurance companies' claims databases) or external data sources (such as industry data sharing platforms). In order to improve the efficiency and accuracy of data updates, the system should regularly obtain the latest insurance claims data from data sources through automated methods such as API interfaces and batch data imports. The frequency of data acquisition can be set according to business needs, such as daily, weekly or monthly updates.

[0114] When new claims data is acquired, the system needs to automatically synchronize this data to each sub-database. Update operations not only include the addition of new data, but may also involve the modification, deletion, or reconstruction of existing data. For example, new claims standards may affect the entries in the claims standards sub-database, and new claims cases need to be added to the claims case sub-database. In order to ensure the reliability of updates, the system needs to perform version control and data verification when updating sub-database data to prevent knowledge base inconsistencies caused by data conflicts or erroneous updates. The update process of the sub-database also needs to record the time and content of each update to ensure data traceability.

[0115] Through the above steps, each sub-database is integrated into a complete claims knowledge base. Claims personnel can obtain comprehensive information support from one system when handling cases, which improves the overall efficiency of the claims process. The system automatically obtains and updates the latest insurance claims data to ensure that the knowledge base content is always up to date, reducing the decision-making risks caused by lagging or incomplete data. The updated automated process reduces manual operations and ensures the real-time and consistency of data.

[0116] The present invention relates to the fields of big data technology and financial technology, and discloses a method for constructing a claims knowledge base, which obtains insurance claims data, including insurance clauses and claims records, and uses a large language model to parse the data, extracts key information such as claims standards, liability exemptions and payment conditions, generates a claims dictionary and clause interpretations, and constructs a claims knowledge extraction sub-base. The method includes analyzing the settlement standards of different types of insurance, constructing a settlement standard sub-base in combination with the claims standards and payment conditions, obtaining historical claims cases, and labeling and managing them according to case types, loss types and customer types, constructing a claims case sub-base, pushing similar cases, and forming a similar case recommendation sub-base. Expert opinions are used to generate standard operating procedures and establish a standard operating procedure sub-base. By integrating all sub-bases, a complete claims knowledge base is formed, and new insurance claims data is regularly obtained to achieve dynamic updates of the sub-bases. The present invention realizes the intelligentization and standardization of the claims process through structured and modular knowledge integration; through structured analysis of insurance claims data, a claims dictionary and settlement standard sub-library are constructed to solve the problems of information fragmentation and inconsistent interpretation of terms; through labeling management of historical cases and pushing similar cases, the processing efficiency and accuracy of complex cases are improved; by combining expert opinions to generate standard operating procedures, the standardization of claims operations is realized; by integrating various sub-libraries and regularly updating the knowledge base, the timeliness and availability of knowledge are ensured. Overall, the present invention effectively simplifies the claims process, significantly improves the efficiency of claims processing, reduces operational complexity, reduces fraud risks, and improves customer experience.

[0117] In one embodiment, the above S50 includes:

[0118] S501, assigning corresponding label information to the current claim case based on the case type, loss type and customer type of the current claim case;

[0119] S502, analyzing the similarity between the label information of the current claim case and the label information of the historical claim cases, and determining similar historical cases whose similarity exceeds a similarity threshold;

[0120] S503, pushing the similar historical cases to provide reference for claim decision-making;

[0121] S504, recording the matching information between the current claim case and the pushed similar historical cases in the similar case recommendation sub-library, wherein the matching information includes case identification, label information and similarity score.

[0122] In this embodiment, label information is assigned to newly entered claims cases. These labels include case type (such as auto insurance, health insurance, property insurance, etc.), loss type (such as vehicle damage, personal injury, etc.), and customer type (such as individual customers, corporate customers, etc.). Label management enables the system to handle claims cases in a structured manner. The system can automatically assign labels to new cases through a rule engine or natural language processing technology. For example, based on the accident description, customer information, and claim application content, the system automatically identifies and assigns appropriate labels to the case.

[0123] The system uses label information to calculate the similarity between the current claim case and historical cases. The system compares the degree of match between the current case and historical cases based on multiple label dimensions (such as case type, loss type, customer type, etc.). The similarity threshold determines the standard for the system to consider the two cases to be sufficiently similar. By setting the similarity threshold, the system can automatically filter out similar historical cases that meet the conditions. Similarity calculations can use algorithms such as weighted similarity or cosine similarity.

[0124] After the system identifies historical cases whose similarity exceeds the set threshold, it will push these cases to the claims adjusters for decision-making reference. The pushed content includes the case type, loss details, and compensation results of similar historical cases. Similar cases can be automatically presented to claims adjusters through the system interface or generated through automated reports to help claims adjusters make the best decision based on historical processing situations.

[0125] The system stores the matching information between the current claim case and the pushed similar historical cases in a dedicated similar case recommendation sub-library. The matching information includes the case's identification, tag information, and similarity score, ensuring that the established similarities can be referenced in future case processing. The similar case recommendation sub-library is dynamically updated with each new case processing and similar case push, keeping the latest matching records, ensuring that the most relevant cases can be referenced in subsequent searches.

[0126] Through the above steps, the system of this embodiment can provide intelligent decision support for claims personnel, push historical cases similar to the current case, reduce the time for claims personnel to manually search for cases, and improve the efficiency and accuracy of case processing. The construction of the similar case recommendation sub-library ensures the systematic management and storage of similar cases, facilitates the rapid retrieval and call of future cases, and further improves the quality of claims services and decision-making level.

[0127] In one embodiment, in the above S40, after obtaining the historical claim cases, the following step is further included:

[0128] S401, extracting fraud cases identified as fraud from the historical claim cases;

[0129] S402, analyzing the fraud case to identify fraud patterns, where the fraud patterns include fabricating accidents, exaggerating losses, and forging documents;

[0130] S403, analyzing the characteristics of the fraud case and extracting risk tags, wherein the risk tags are used to identify the case type and the customer type;

[0131] S404, storing the fraud case, the fraud pattern and the risk label in a fraud risk identification sub-library;

[0132] S405, matching the current claim case with known fraud patterns through the fraud risk identification sub-database to generate a fraud warning.

[0133] In this embodiment, based on the existing historical claims case data, cases marked as "fraud" are automatically extracted. Fraud cases are usually identified during the claims review process, including cases with forged materials, false declarations, etc. The system uses specific rules or abnormal behaviors based on historical claims records to screen out cases identified as fraud. These cases will become the basis for subsequent analysis and fraud pattern extraction.

[0134] By analyzing historical fraud cases, common fraud patterns are identified. Fraud patterns may include fictitious accidents (such as reporting accidents that did not occur), exaggerating losses (deliberately increasing the amount of compensation), and forging documents (providing false materials). The system analyzes the behavioral patterns in fraud cases and summarizes different types of fraud methods. The analysis of fraud patterns can be achieved through rule engines or machine learning technology. The system can identify common fraud behavior patterns and then classify and summarize them.

[0135] The system will conduct a detailed analysis of each case identified as fraud and extract characteristic information of the case, such as case type (auto insurance, health insurance, property insurance, etc.) and customer type (individual customer, corporate customer, etc.). These characteristics will be converted into "risk tags" to identify which cases or customer types are more likely to commit fraud. These risk tags can help the system identify potential high-risk cases more quickly in subsequent claims processing, thereby providing claims personnel with early fraud warnings.

[0136] The analyzed fraud cases, fraud patterns and risk tags will be stored in the fraud risk identification sub-database. This sub-database is a database dedicated to managing and storing all fraud-related case information and pattern analysis results. The fraud risk identification sub-database will provide data support for future claims processing. Through the information in the sub-database, new fraud cases can be identified more accurately and early warnings can be issued when similar behaviors occur.

[0137] When processing a new claim, the system will match the feature information of the new case with the known fraud patterns in the fraud risk identification sub-database. If the new case is similar to certain known fraud patterns, the system will automatically generate a fraud warning prompt to remind the claims personnel that the case may have fraud risks. The warning mechanism determines whether a warning is needed by calculating the similarity between the case features and the fraud patterns in the sub-database. Cases with similarity exceeding the set threshold will trigger a warning prompt for further review by the claims personnel.

[0138] Through the above steps, the system of this embodiment can effectively identify and prevent fraud in the claims process. Through the analysis and pattern extraction of historical fraud cases, the system can summarize common fraud behavior patterns and provide early warning support for future claims cases to reduce the occurrence of fraud cases. The construction of the fraud risk identification sub-library provides a data foundation for the system, helping claims personnel to better identify high-risk cases when handling cases, and improving the safety and accuracy of claims decisions.

[0139] In one embodiment, the above S60 includes:

[0140] S601, based on different insurance types and corresponding operational links, collect expert opinions in the fields of loss assessment, settlement review, fraud risk management and complaint handling;

[0141] S602, based on the expert opinions, formulate standard operating procedures for different types of insurance, wherein the standard operating procedures include the operation steps and key links of investigation, loss assessment, review and complaint handling;

[0142] S603, modularize the standard operating process and store it in a standard operating process sub-library.

[0143] In this embodiment, opinions and suggestions are collected from experts in the insurance field according to different types of insurance products (such as auto insurance, health insurance, property insurance, etc.) and different operational links (such as damage assessment, settlement review, fraud risk management, complaint handling, etc.). Experts may come from different claims operation links, such as damage assessment experts, risk management experts, and complaint handling experts, to ensure that the standard operating procedures of each link are based on professional knowledge. Experts' operational suggestions on the claims process of each type of insurance are collected through meetings, interviews, surveys or questionnaires to ensure that the processing of each link meets best practices and industry standards.

[0144] Based on the collected expert opinions, standardized operating procedures are systematically formulated for different insurance types and operational links. Each process should include specific steps and key operating points for investigation, damage assessment, review, fraud risk management and complaint handling to ensure that each link has clear guidelines and standards. The operational steps may cover accident investigation, damage assessment, settlement review, fraud risk analysis and handling of customer complaints. The key links include risk points that need to be paid attention to and necessary review mechanisms to ensure that each step in the process can be executed according to standards.

[0145] Modularize the standard operating procedures for different types of insurance so that the procedures for each type of insurance can be called independently or combined, which is convenient for customizing the calling of process modules according to specific cases. Modularization ensures the flexibility of the process in different situations. Store the modularized standard operating procedures in the standard operating procedure sub-library to ensure that the sub-library can provide standardized operating procedure support for claims personnel. The sub-library needs to have a flexible search function to facilitate claims personnel to quickly obtain the corresponding standard procedures according to the insurance type and operation requirements.

[0146] The system collects best practice suggestions for each link in the claims process through surveys, expert interviews or expert seminars. Expert opinions include specific operational steps and key risk points for investigation, loss assessment, review, fraud management and complaint handling. The collected expert opinions are classified and sorted according to different insurance types and operational links to ensure that each link of each insurance type can cover comprehensive operational specifications.

[0147] The system converts expert opinions into operational standard operating procedures. For each type of insurance, develop corresponding detailed operating procedures for investigation, loss assessment, review and complaint handling, and mark the key operating points and control elements of each link. When formulating the process, set necessary risk control points and review mechanisms for high-risk points in each link (such as fraud identification, complaint escalation, etc.). Modularize the standard operating procedures for different types of insurance so that the procedures can be combined and called according to the steps. After modularization, each step can be applied independently or dynamically combined according to the needs of different cases. Store the modularized standard operating procedures in the standard operating procedure sub-library to facilitate claims personnel to call the appropriate standard procedures when handling cases. The sub-library needs to support query and dynamic update to adapt to industry changes.

[0148] Through the above steps, the opinions of experts in this embodiment are systematized and streamlined, making the claims processing procedures of different types of insurance standardized and modular. The formulation of standard operating procedures ensures the standardization and consistency of the claims processing process, and reduces the operational deviations and risks caused by inconsistent processes. The modular process design provides flexible operating tools for claims personnel to ensure that each case can be handled in accordance with best practices. At the same time, the construction of the standard operating process sub-library makes the retrieval and call of the process more convenient, improving the efficiency of claims processing and the standardization of operations.

[0149] In one embodiment, the above S30 includes:

[0150] S301, extracting settlement data of each insurance type from insurance clauses and claim records, wherein the settlement data includes payment conditions, payment limits and deductibles;

[0151] S302, generating settlement standards for each insurance type including loss assessment, compensation amount and compensation ratio according to the claim settlement standards and compensation conditions of each insurance type;

[0152] S303, based on the settlement standard of each insurance type, establishing a settlement standard model for each insurance type, wherein the settlement standard model includes compensation conditions, compensation limits and settlement logic;

[0153] S304, storing the generated settlement standard model for each insurance type in the settlement standard sub-library.

[0154] In this embodiment, settlement data related to each type of insurance is extracted from the insurance terms and historical claims records. The extracted settlement data mainly includes information such as compensation conditions, compensation limits and deductibles. These data are the core basis for settlement and are used to calculate the specific compensation amount in each claim case. The data can come from the insurance company's insurance terms database and historical claims records in the claims system. The system automatically parses the compensation conditions (such as whether it meets the compensation standards), compensation limits (such as the maximum compensation amount) and deductibles (such as the amount of the deductible part) in the terms.

[0155] The system generates detailed settlement standards based on the claim standards and payment conditions of different types of insurance. These standards cover loss assessment (such as vehicle loss, medical expenses, etc.), compensation amount (calculated based on actual losses) and compensation ratio (such as the ratio borne by the insurance company). The settlement standards for different types of insurance may be different. For example, the settlement standards for auto insurance may involve the compensation limit for vehicle repair costs, while health insurance may involve the reimbursement ratio for hospitalization expenses. The system calculates according to the terms of each insurance type and the actual loss situation.

[0156] By integrating the settlement standards of each type of insurance into a complete settlement standard model. The settlement standard model includes the compensation conditions (such as the scope of insurance liability), the compensation limit (such as the maximum compensation amount) and the settlement logic (such as how to calculate the compensation amount). The settlement standard model is the core data structure that provides automated support for subsequent claims calculations. Through the settlement standard model, the system can automatically calculate the corresponding compensation amount for each claim case and ensure that the settlement of each case complies with the provisions of the terms.

[0157] The generated settlement standard model will be stored in the settlement standard sub-library for subsequent use in the claims processing process. The role of the sub-library is to classify and store the settlement standard model for each insurance type, ensuring that claims personnel can quickly call the corresponding settlement standard when handling cases of different insurance types. The settlement standard sub-library needs to support regular updates to adapt to changes in insurance terms or updates to insurance types. After each update, the new settlement standard will be automatically synchronized to the sub-library to ensure the timeliness and accuracy of the data.

[0158] The system first extracts settlement data related to each type of insurance from the insurance company's insurance terms database and claims records. The extracted content includes compensation conditions (such as whether the compensation requirements are met), compensation limits (such as the maximum compensation amount for a certain type of insurance) and deductibles (the part that the insurance company does not pay). The extracted data is standardized to ensure that the data format of different types of insurance is consistent, which is convenient for the subsequent generation of settlement models.

[0159] The system calculates the settlement standards based on the claim standards for each type of insurance. This includes loss assessment (such as property damage or medical expenses), compensation amount (actual compensation amount) and compensation ratio (the portion borne by the insurance company). The system automatically calculates according to the rules in the terms. A standard template is generated for each type of insurance to standardize the settlement rules under that type of insurance. The template contains information such as compensation limits, deductibles and compensation conditions.

[0160] The system integrates the generated settlement standards into a settlement standard model. The model contains the settlement standards for each type of insurance and defines the logic for calculating the compensation amount. The settlement logic includes the specific steps for calculating the compensation amount based on the loss amount, as well as how to apply the calculation rules for the compensation limit and deductible.

[0161] The settlement standard model generated for each insurance type will be stored in the settlement standard sub-library. The sub-library is classified and managed by insurance type to ensure that the settlement models of different insurance types can be quickly retrieved and called. The system needs to support dynamic updates of the sub-library. When the insurance terms change, the system can automatically update the corresponding settlement standard model to ensure the timeliness of the settlement standard.

[0162] Through the above steps, the system can automatically extract settlement standards from insurance clauses and claim records and build a standardized settlement model. The construction of the settlement standard sub-library enables claims personnel to quickly call corresponding settlement standards when handling cases of different insurance types, thereby improving claims efficiency and accuracy. Through dynamic updates, the settlement standard sub-library is always kept up to date, ensuring that claims calculations comply with the latest clauses and rules.

[0163] In one embodiment, the above S70 includes:

[0164] S701, integrating the data in the claims knowledge extraction sub-database, the settlement standard sub-database, the claims case sub-database, the similar case recommendation sub-database and the standard operating procedure sub-database, and establishing associations between the sub-databases;

[0165] S702, processing the data in all sub-databases in a unified format and building a unified query interface for all sub-databases;

[0166] S703, regularly obtaining new insurance claims data from the insurance claims data source, and verifying, cleaning and formatting the newly obtained insurance claims data;

[0167] S704: Synchronize the new data to each corresponding sub-database through an incremental or full update mechanism, so that the data in each sub-database is in an updated state.

[0168] In this embodiment, the data of each independent sub-library (claims knowledge extraction sub-library, settlement standard sub-library, claims case sub-library, similar case recommendation sub-library and standard operating process sub-library) are integrated to form a complete claims knowledge base. The association relationship between sub-libraries is realized through the design of logic or data structure. For example, a case in the claims case sub-library can be associated with the claims calculation rules in the settlement standard sub-library to ensure the interaction and consistency between data. The association relationship between sub-libraries is established so that interrelated data can be extracted from different sub-libraries when processing claims cases. For example, when processing claims cases, the system can automatically call the corresponding payment standards from the settlement standard sub-library, retrieve similar historical cases from the claims case sub-library, and call the corresponding process operation guide from the standard operating process sub-library.

[0169] The data formats in different sub-databases may be different, especially in different insurance types or different claims operation links, the data structure may change. This step uses data standardization to unify the format of the data in all sub-databases to ensure that the data in each sub-database can interact and be used in the same system. Build a unified query interface to facilitate the system or users to quickly query and call the data in each sub-database. The query interface needs to be flexible and support functions such as multi-condition query and cross-sub-database query, so that claims personnel can query all data related to the claim case on the same platform.

[0170] New claims data is regularly obtained from insurance claims data sources (such as internal claims systems or external data sharing platforms). To ensure the quality of new data, the system needs to verify and clean the data before importing it to remove invalid, duplicate or erroneous data. The data formats of different data sources may be different, so the system needs to convert the format of the newly acquired data to ensure that the new data is consistent with the existing data in the sub-database. For example, the system may convert the claims records in the external data source into a structured data format that matches the internal sub-database.

[0171] For data that changes frequently, the system uses an incremental update mechanism to update only newly added or modified data to avoid burdening the entire database. Incremental updates are suitable for updating claims records or case data. For data that needs to be reinitialized or adjusted on a large scale, the system uses a full update mechanism to reconstruct the data of the entire sub-database. Full updates are suitable for data synchronization when major terms are changed or the system is upgraded.

[0172] This embodiment integrates the data of each sub-database into a complete claims knowledge base, and the system realizes centralized data management, enabling claims personnel to query and call all relevant data from one platform, greatly improving work efficiency. The regular data update mechanism ensures that the data in the sub-database is always up to date, avoiding decision-making risks caused by data lag or inconsistency. The unified data format and query interface provide a more convenient operating environment, support multi-dimensional query and cross-sub-database call, and improve the intelligence and automation level of the claims process.

[0173] In one embodiment, in the above S70, after all sub-databases are integrated to form a complete claims knowledge base, the following is further included:

[0174] S801, extracting document data from pending claims, comparing the document data with the insurance clause information in the claims knowledge extraction sub-database, and verifying the validity and compliance of the document data;

[0175] S802, generating a claim suggestion based on the settlement standards in the settlement standards sub-library, and verifying whether the claim suggestion meets the claim conditions in the settlement standards;

[0176] S803, according to the label information of the pending claim case, push similar historical cases whose similarity with the pending claim case exceeds a similarity threshold from the claim case sub-database;

[0177] S804, combining the risk labels of the fraud risk identification sub-database, performing risk analysis on the pending claim case to analyze whether there is potential fraud in the pending claim case;

[0178] S805, generating a compensation proposal, a case handling report and a document list based on the compensation proposal, similar historical cases and risk analysis results;

[0179] S806, synchronously updating the processed case data to the corresponding sub-database in the claims knowledge base.

[0180] In this embodiment, relevant document data (such as insurance policies, bills, invoices, repair orders, etc.) are extracted from the pending claims, and these documents are compared with the insurance terms information in the claims knowledge extraction sub-database. The purpose of the comparison is to verify whether the document data complies with the provisions of the insurance policy and to ensure that the submitted materials are authentic, valid, and compliant. The system will check whether the information in the document data is consistent with the provisions of the terms according to the requirements of the insurance terms. For example, whether the amount of compensation, time of accident, division of responsibilities, etc. in the document comply with the provisions of the insurance terms.

[0181] The system automatically calculates the compensation proposal for pending claims by calling the settlement standards in the settlement standard sub-library. The compensation proposal includes the estimated compensation amount, compensation ratio, etc. After generating the compensation proposal, the system will verify the compensation proposal according to the compensation conditions in the settlement standard sub-library. Ensure that the compensation amount, compensation ratio, etc. comply with the insurance terms and settlement standards to prevent overpayment or non-compliant compensation.

[0182] The label information of the pending claim case (such as case type, loss type, customer type, etc.) will be matched with the historical cases in the claim case sub-database. The system calculates the similarity and pushes the historical cases whose similarity with the label information of the current case exceeds the set threshold. The historical cases whose similarity exceeds the threshold will be automatically pushed to the claims personnel for reference to the handling results of the historical cases and to help optimize the handling decision of the current case.

[0183] The system combines the risk tags in the fraud risk identification sub-database (such as false declarations, exaggerated losses, etc.) to conduct risk analysis on pending claims and determine whether the case has fraud risks. Fraudulent behavior may include forging documents, exaggerating losses, and falsely reporting accidents. If the pending claim case has a high degree of match with the fraud pattern in the fraud risk identification sub-database, the system will automatically generate a fraud warning to prompt the claims personnel to conduct a more rigorous review of the case.

[0184] The system automatically generates compensation proposals, case handling reports and document lists based on compensation suggestions, similar historical cases and fraud risk analysis results. The compensation proposal includes compensation amount, compensation conditions, etc.; the case handling report summarizes the case handling process and risk analysis results; the document list lists the required documents and materials. Through automated document generation tools, the time and cost of manual report writing are reduced, ensuring the standardization and accuracy of each document.

[0185] The processed case data (including compensation suggestions, case processing reports, similar cases, risk analysis results, etc.) will be automatically updated to the corresponding sub-database in the claims knowledge base. The system stores the updated data according to the sub-database category (such as the claims case sub-database, the settlement standard sub-database, etc.). Through regular data synchronization and updating, the data in the claims knowledge base is always kept up to date, which is convenient for future case processing and decision-making reference.

[0186] Through the above steps, the system of this embodiment can automatically process the document verification, payment suggestion generation, similar case push, fraud risk analysis and other operations in the claim case, greatly improving the efficiency and accuracy of the claim process. The automatically generated documents and reports simplify the work of the claims personnel and ensure document standardization and compliance. By dynamically updating case data, the data in the claims knowledge base is always kept up to date, providing more accurate decision support and reference for future case processing.

[0187] In one embodiment, a claim knowledge base construction device is provided, and the claim knowledge base construction device corresponds one-to-one to the claim knowledge base construction method in the above embodiment. Figure 3 , Figure 3 This is a functional module diagram of a preferred embodiment of the claim knowledge base construction device of the present invention. Data collection module 10, knowledge extraction module 20, settlement standard analysis module 30, case labeling module 40, similar case recommendation module 50, standard operating process management module 60 and knowledge base integration and update module 70. The functional modules are described in detail as follows:

[0188] The data collection module 10 acquires insurance claim data, wherein the insurance claim data includes insurance terms and claim records;

[0189] The knowledge extraction module 20 uses a large language model to parse the insurance claim data, extract key information including claim standards, liability exemptions and compensation conditions, generate a claim dictionary and clause interpretations, and construct a claim knowledge extraction sub-library based on the claim dictionary and the clause interpretations;

[0190] A settlement standard analysis module 30, which analyzes settlement standards of different types of insurance based on the insurance claim data, and constructs a settlement standard sub-library based on the claim standards of different types of insurance and the compensation conditions;

[0191] The case labeling module 40 obtains historical claim cases, and manages them by labeling according to case type, loss type and customer type, and builds a claim case sub-library based on the historical claim cases managed by labeling;

[0192] A similar case recommendation module 50 pushes similar cases based on the historical claim cases managed by labeling and the similarity threshold, and builds a similar case recommendation sub-library based on the similar cases;

[0193] The standard operating process management module 60 obtains expert opinions on insurance operating processes, generates standard operating processes based on the expert opinions, and constructs a standard operating process sub-library based on the standard operating processes;

[0194] The knowledge base integration and updating module 70 integrates all sub-bases to form a complete claims knowledge base, and updates the data in the sub-bases by regularly acquiring new insurance claims data.

[0195] In one embodiment, the similar case recommendation module 50 is specifically used to:

[0196] Based on the case type, loss type and customer type of the current claim case, the current claim case is assigned corresponding label information;

[0197] Analyze the similarity between the label information of the current claim case and the label information of the historical claim cases, and determine similar historical cases whose similarity exceeds the similarity threshold;

[0198] Push similar historical cases to provide reference for claims decision-making;

[0199] The matching information between the current claim case and the pushed similar historical cases is recorded in the similar case recommendation sub-library, wherein the matching information includes case identification, label information and similarity score.

[0200] In one embodiment, the case labeling module 40 is specifically used to:

[0201] extracting fraud cases identified as fraud from the historical claims cases;

[0202] Analyze the fraud cases to identify fraud patterns, including fabricating accidents, exaggerating losses, and forging documents;

[0203] Analyze the characteristics of the fraud case and extract risk labels, where the risk labels are used to identify the case type and customer type;

[0204] Storing the fraud cases, the fraud patterns and the risk labels in a fraud risk identification sub-database;

[0205] Through the fraud risk identification sub-database, the current claim cases are matched with known fraud patterns to generate fraud warnings.

[0206] In one embodiment, the standard operating process management module 60 is specifically used to:

[0207] According to different insurance types and corresponding operation links, collect expert opinions in the fields of loss assessment, settlement review, fraud risk management and complaint handling;

[0208] Based on the expert opinions, formulate standard operating procedures for different types of insurance, including the operating steps and key links of investigation, loss assessment, review and complaint handling;

[0209] The standard operating procedures are modularized and stored in a standard operating procedure sub-library.

[0210] In one embodiment, the adjustment standard analysis 30 is specifically used to:

[0211] Extracting settlement data for each type of insurance from insurance clauses and claim records, wherein the settlement data includes payment conditions, payment limits and deductibles;

[0212] Generate the settlement standards for each insurance type, including loss assessment, compensation amount and compensation ratio, based on the claims standards and compensation conditions of each insurance type;

[0213] Based on the settlement standards of each insurance type, a settlement standard model is established for each insurance type, wherein the settlement standard model includes compensation conditions, compensation limits and settlement logic;

[0214] The generated settlement standard model for each insurance type is stored in the settlement standard sub-library.

[0215] In one embodiment, the knowledge base integration and updating module 70 is specifically used to:

[0216] Integrate the data in the claims knowledge extraction sub-database, settlement standard sub-database, claims case sub-database, similar case recommendation sub-database and standard operating process sub-database, and establish associations between the sub-databases;

[0217] Process the data in all sub-databases in a unified format and build a unified query interface for all sub-databases;

[0218] Regularly obtain new insurance claims data from insurance claims data sources, and verify, clean and convert the newly acquired insurance claims data;

[0219] Through the incremental or full update mechanism, the new data is synchronized to each corresponding sub-database, so that the data in each sub-database is in an updated state.

[0220] In one embodiment, the knowledge base integration and updating module 70 is specifically used to:

[0221] Extract document data from pending claims, compare the document data with the insurance clause information in the claims knowledge extraction sub-database, and verify the validity and compliance of the document data;

[0222] Generate a claim proposal based on the settlement standards in the settlement standards sub-library, and verify whether the claim proposal meets the claim conditions in the settlement standards;

[0223] According to the label information of the pending claim case, similar historical cases whose similarity with the pending claim case exceeds a similarity threshold are pushed from the claim case sub-database;

[0224] Combined with the risk labels of the fraud risk identification sub-database, risk analysis is performed on pending claims to determine whether there is potential fraud in the pending claims;

[0225] Generate compensation proposals, case handling reports and document lists based on the compensation proposals, similar historical cases and risk analysis results;

[0226] The processed case data is synchronously updated to the corresponding sub-database in the claims knowledge base.

[0227] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client via a network connection. When the computer program is executed by the processor, it implements the functions or steps on the service side of a method for constructing a claims knowledge base.

[0228] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps of the client side of a method for constructing a claims knowledge base

[0229] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0230] Obtaining insurance claim data, wherein the insurance claim data includes insurance terms and claim records;

[0231] Use a large language model to parse insurance claims data, extract key information including claim standards, liability exemptions, and compensation conditions, generate a claims dictionary and clause interpretations, and build a claims knowledge extraction sub-library based on the claims dictionary and the clause interpretations;

[0232] Analyzing settlement standards of different types of insurance based on the insurance claim data, and building a settlement standard sub-library based on the claim standards of different types of insurance and the compensation conditions;

[0233] Obtain historical claims cases and manage them by labeling according to case type, loss type and customer type, and build a claims case sub-library based on the historical claims cases managed by labeling;

[0234] Push similar cases based on the historical claims cases managed by labeling and similarity thresholds, and build a similar case recommendation sub-library based on the similar cases;

[0235] Obtain expert opinions on insurance operating procedures, generate standard operating procedures based on the expert opinions, and build a standard operating procedure sub-library based on the standard operating procedures;

[0236] A complete claims knowledge base is formed based on the integration of all sub-databases, and the data in the sub-databases are updated by regularly acquiring new insurance claims data.

[0237] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0238] Obtaining insurance claim data, wherein the insurance claim data includes insurance terms and claim records;

[0239] Use a large language model to parse insurance claims data, extract key information including claim standards, liability exemptions, and compensation conditions, generate a claims dictionary and clause interpretations, and build a claims knowledge extraction sub-library based on the claims dictionary and the clause interpretations;

[0240] Analyzing settlement standards of different types of insurance based on the insurance claim data, and building a settlement standard sub-library based on the claim standards of different types of insurance and the compensation conditions;

[0241] Obtain historical claims cases and manage them by labeling according to case type, loss type and customer type, and build a claims case sub-library based on the historical claims cases managed by labeling;

[0242] Push similar cases based on the historical claims cases managed by labeling and similarity thresholds, and build a similar case recommendation sub-library based on the similar cases;

[0243] Obtain expert opinions on insurance operating procedures, generate standard operating procedures based on the expert opinions, and build a standard operating procedure sub-library based on the standard operating procedures;

[0244] A complete claims knowledge base is formed based on the integration of all sub-databases, and the data in the sub-databases are updated by regularly acquiring new insurance claims data.

[0245] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0246] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0247] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0248] It should be noted that if software tools or components other than those of the Company appear in the embodiments of the present application, they are only used for illustration and do not represent actual use. The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the above-mentioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the above-mentioned embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents; and these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for constructing a claims knowledge base, characterized in that: The following steps are involved: Obtaining insurance claim data, wherein the insurance claim data includes insurance terms and claim records; Use a large language model to parse insurance claims data, extract key information including claim standards, liability exemptions, and payment conditions, generate a claims dictionary and clause interpretations, and build a claims knowledge extraction sub-library based on the claims dictionary and the clause interpretations; Analyzing settlement standards of different types of insurance based on the insurance claim data, and building a settlement standard sub-library based on the claim standards of different types of insurance and the compensation conditions; Obtain historical claims cases and manage them by labeling according to case type, loss type and customer type, and build a claims case sub-library based on the historical claims cases managed by labeling; Push similar cases based on the historical claims cases managed by labeling and similarity thresholds, and build a similar case recommendation sub-library based on the similar cases; Obtain expert opinions on insurance operating procedures, generate standard operating procedures based on the expert opinions, and build a standard operating procedure sub-library based on the standard operating procedures; A complete claims knowledge base is formed based on the integration of all sub-databases, and the data in the sub-databases are updated by regularly acquiring new insurance claims data.

2. The method for constructing a claims knowledge base according to claim 1, characterized in that: Based on the historical claim cases managed by labeling and the similarity threshold, similar cases are pushed, and based on the similar cases, a similar case recommendation sub-library is constructed, including: Based on the case type, loss type and customer type of the current claim case, the current claim case is assigned corresponding label information; Analyze the similarity between the label information of the current claim case and the label information of the historical claim cases, and determine similar historical cases whose similarity exceeds the similarity threshold; Push similar historical cases to provide reference for claims decision-making; The matching information between the current claim case and the pushed similar historical cases is recorded in the similar case recommendation sub-library, wherein the matching information includes case identification, label information and similarity score.

3. The method for constructing a claims knowledge base according to claim 1, characterized in that: After obtaining historical claims cases, it also includes: extracting fraud cases identified as fraud from the historical claims cases; Analyze the fraud cases to identify fraud patterns, including fabricating accidents, exaggerating losses, and forging documents; Analyze the characteristics of the fraud case and extract risk labels, where the risk labels are used to identify the case type and customer type; Storing the fraud cases, the fraud patterns and the risk labels in a fraud risk identification sub-database; Through the fraud risk identification sub-database, the current claim cases are matched with known fraud patterns to generate fraud warnings.

4. The method for constructing a claims knowledge base according to claim 1, characterized in that: Obtain expert opinions on insurance operation processes, generate standard operation processes based on the expert opinions, and build a standard operation process sub-library based on the standard operation processes, including: According to different insurance types and corresponding operation links, collect expert opinions in the fields of loss assessment, settlement review, fraud risk management and complaint handling; Based on the expert opinions, formulate standard operating procedures for different types of insurance, including the operating steps and key links of investigation, loss assessment, review and complaint handling; The standard operating procedures are modularized and stored in a standard operating procedure sub-library.

5. The method for constructing a claims knowledge base as claimed in claim 1, characterized in that: Analyzing the settlement standards of different types of insurance based on the insurance claim data, and constructing a settlement standard sub-library based on the claim standards of different types of insurance and the compensation conditions, including: Extracting settlement data for each type of insurance from insurance clauses and claim records, wherein the settlement data includes payment conditions, payment limits and deductibles; Generate the settlement standards for each insurance type, including loss assessment, compensation amount and compensation ratio, based on the claims standards and compensation conditions of each insurance type; Based on the settlement standards of each insurance type, a settlement standard model is established for each insurance type, wherein the settlement standard model includes compensation conditions, compensation limits and settlement logic; The generated settlement standard model for each insurance type is stored in the settlement standard sub-library.

6. The method for constructing a claims knowledge base according to claim 1, characterized in that: A complete claims knowledge base is formed based on the integration of all sub-databases, and the data in the sub-databases are updated by regularly acquiring new insurance claims data, including: Integrate the data in the claims knowledge extraction sub-database, settlement standard sub-database, claims case sub-database, similar case recommendation sub-database and standard operating process sub-database, and establish associations between the sub-databases; Process the data in all sub-databases in a unified format and build a unified query interface for all sub-databases; Regularly obtain new insurance claims data from insurance claims data sources, and verify, clean and convert the newly acquired insurance claims data; Through the incremental or full update mechanism, the new data is synchronized to each corresponding sub-database, so that the data in each sub-database is in an updated state.

7. The method for constructing a claims knowledge base according to claim 1, characterized in that: After all sub-databases are integrated to form a complete claims knowledge base, it also includes: Extract document data from pending claims, compare the document data with the insurance clause information in the claims knowledge extraction sub-database, and verify the validity and compliance of the document data; Generate a claim proposal based on the settlement standards in the settlement standards sub-library, and verify whether the claim proposal meets the claim conditions in the settlement standards; According to the label information of the pending claim case, similar historical cases whose similarity with the pending claim case exceeds a similarity threshold are pushed from the claim case sub-database; Combined with the risk labels of the fraud risk identification sub-database, risk analysis is performed on pending claims to determine whether there is potential fraud in the pending claims; Generate compensation proposals, case handling reports and document lists based on the compensation proposals, similar historical cases and risk analysis results; The processed case data is synchronously updated to the corresponding sub-database in the claims knowledge base.

8. A claim knowledge base construction device, characterized in that: The claim knowledge base construction device comprises: A data collection module, which obtains insurance claim data, including insurance terms and claim records; The knowledge extraction module uses a large language model to parse insurance claims data, extract key information including claim standards, liability exemptions, and compensation conditions, generate a claims dictionary and clause interpretations, and build a claims knowledge extraction sub-library based on the claims dictionary and the clause interpretations; A settlement standard analysis module, which analyzes settlement standards of different types of insurance based on the insurance claim data, and constructs a settlement standard sub-library based on the claim standards of different types of insurance and the payment conditions; The case labeling module obtains historical claims cases and manages them by labeling according to case type, loss type and customer type, and builds a claims case sub-library based on the historical claims cases managed by labeling; A similar case recommendation module pushes similar cases based on the historical claim cases managed by labeling and the similarity threshold, and builds a similar case recommendation sub-library based on the similar cases; A standard operating process management module obtains expert opinions on insurance operating processes, generates standard operating processes based on the expert opinions, and constructs a standard operating process sub-library based on the standard operating processes; The knowledge base integration and update module forms a complete claims knowledge base based on the integration of all sub-databases, and updates the data in the sub-databases by regularly acquiring new insurance claims data.

9. A computer device, characterized in that: The computer device includes a memory, a processor, and a claims knowledge base construction program stored in the memory and executable on the processor. When the claims knowledge base construction program is executed by the processor, the steps of the claims knowledge base construction method as described in any one of claims 1-7 are implemented.

10. A computer-readable storage medium, characterized in that: The storage medium stores a claims knowledge base construction program, and when the claims knowledge base construction program is executed by the processor, the steps of the claims knowledge base construction method according to any one of claims 1 to 7 are implemented.

Images