Intelligent claim processing method and system, electronic equipment and storage medium

Abstract

The invention provides an intelligent claim processing method and system, electronic equipment and a storage medium. The intelligent claim processing method comprises the steps that a claim file is received, and the claim file comprises a plurality of claim sub-files of different file types; performing format conversion on each claim subfile to obtain a plurality of first files; each first file is processed to obtain a plurality of data matrixes, and the data matrixes comprise spatial semantic relations; determining a business scene corresponding to the claim file according to the plurality of data matrixes and a visual model; processing the plurality of data matrixes according to the service scene to obtain a plurality of pieces of first data information; according to the multiple pieces of first data information, constructing an evidence chain knowledge graph taking time and logic relations as edges; determining a claim rule database according to the business scene; and determining a target processing suggestion according to the evidence chain knowledge graph and the claim settlement rule database. The problems of low auditing efficiency, high cost, different standards and the like in the traditional claim process are solved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to an intelligent claims processing method and system, electronic device, and storage medium. Background Technology

[0002] The traditional claims processing process relies on manual processing of heterogeneous documents (such as invoices, photos, and chat logs) by human auditors. This results in a process that is extremely inefficient, costly, inconsistent in standards, and prone to errors. Summary of the Invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art or related art.

[0004] Therefore, the first aspect of the present invention proposes an intelligent claims processing method.

[0005] A second aspect of the present invention proposes an intelligent claims processing system.

[0006] A third aspect of the present invention provides an electronic device.

[0007] A fourth aspect of the present invention provides a storage medium.

[0008] In view of this, according to a first aspect of the present invention, an intelligent claims processing method is proposed, comprising: receiving a claims document, wherein the claims document includes multiple claims sub-files of different file types; converting the format of each claims sub-file to obtain multiple first files; processing each first file to obtain multiple data matrices, wherein the data matrices contain spatial semantic relationships; determining the business scenario corresponding to the claims document based on the multiple data matrices and a visual model; processing the multiple data matrices based on the business scenario to obtain multiple first data information; constructing an evidence chain knowledge graph with time and logical relationships as edges based on the multiple first data information; determining a claims rule database based on the business scenario; and determining a target processing suggestion based on the evidence chain knowledge graph and the claims rule database.

[0009] The intelligent claim processing method provided by this invention specifically includes: firstly, receiving claim documents, which can be uploaded by users or by the manufacturers involved in the claim. Claim documents refer to various supporting materials used to prove the legality and authenticity of the claim request during the claim process. Claim documents can include multiple claim sub-files of different file types. For example, claim sub-files can be accident scene photos, repair expense invoices, and customer service communication texts, etc. These claim sub-files have different file types; for example, accident scene photos are images, repair expense invoices are invoices, and customer service communication texts are chat logs. After receiving the claim documents, because the different claim sub-files have different file types, their file formats are also different, such as JPG (Joint Photographic Experts Group), PNG (Portable Network Graphics), PDF (Portable Document Format), and TXT (Plain Text), etc. Therefore, it is necessary to convert each claim sub-file into multiple first files with a unified format. Subsequently, multiple first files are processed separately to obtain multiple standard data matrices that retain the original spatial semantic relationships, facilitating subsequent model processing. After obtaining multiple data matrices, they are input into a pre-trained Large Vision Model (LVM). The visual model performs an overall analysis of the multiple data matrices to obtain the business scenario corresponding to the claim document. This business scenario could include car repair claims, product warranty identification, and car warranty claims, etc. After determining the business scenario corresponding to the claim document, a corresponding processing flow can be determined based on the business scenario. Then, the multiple data matrices are processed according to the processing flow to extract the key information, i.e., the first data information, from each data matrix. The multiple first data information are then fused to construct an evidence chain knowledge graph with time and logical relationships as edges. Simultaneously, a corresponding claims rule database is determined based on the determined business scenario. Different claims rule databases store different claims rules; for example, claims rules could be insurance claim terms and product warranty policies. Therefore, it is necessary to determine the corresponding claims rule database based on the determined business scenario. Finally, the evidence chain knowledge graph is matched and calculated with the claims rule database to obtain the final target processing suggestion. The target processing suggestion includes whether to process the claim and how to process it. This invention automates the entire process from document reception to decision suggestion, freeing reviewers from tedious repetitive work, greatly improving processing efficiency, and reducing labor costs.Furthermore, since the same claims rule database is used to process the same business scenarios in this invention, the problem of inconsistent claims standards in related technologies is solved.

[0010] In some technical solutions, optionally, the step of processing multiple data matrices according to a business scenario to obtain multiple first data information includes: determining a first data model set according to the business scenario, the first data model set including multiple second data models; determining the content information in each data matrix; determining the second data model corresponding to each data matrix according to the content information; and obtaining multiple first data information based on the multiple data matrices and the second data model corresponding to each data matrix.

[0011] In this technical solution, the steps of processing multiple data matrices according to a business scenario to obtain multiple first data information include: First, determining a first data model set based on the business scenario. This first data model set corresponds to the business processing flow; that is, after determining the business scenario, the required business processing flow is determined, and then the corresponding first data model set is determined based on the business processing flow. The first data model set includes multiple second data models. This is because, during the claims process, multiple aspects of data need to be comprehensively processed, so each second data model corresponds to a specific data processing step. Next, the content information in each data matrix is ​​determined, and then the corresponding second data model is determined based on this content information. In other words, the data processing flow for each data matrix is ​​determined based on its content information. Finally, the multiple data matrices are input into their respective second data models, which process the data matrices, extract key information, and output the corresponding first data information, which is standardized structured data. By determining a first set of data models based on business scenarios, and then determining a corresponding second data model from the first set of data models based on the content of each data matrix, the second data model is then used to extract key information from the data matrix. This achieves in-depth optimization for different business scenarios and ensures the accuracy and professionalism of information extraction.

[0012] In some technical solutions, optionally, after determining the target processing recommendation based on the evidence chain knowledge graph and the claims rule database, the solution further includes: determining a first confidence level corresponding to each data matrix based on multiple data matrices and a visual model; determining a second confidence level based on multiple first confidence levels; and adding the second confidence level to the target processing recommendation.

[0013] In this technical solution, after determining the target processing recommendation based on the evidence chain knowledge graph and the claims rule database, the solution further includes: firstly, inputting multiple data matrices into a visual model, and using the visual model to analyze the content of each data matrix to determine the first confidence level corresponding to each data matrix, i.e., determining the credibility of each claim sub-document. Then, fusing multiple first confidence levels to determine a second confidence level; that is, fusing the first confidence levels corresponding to each claim sub-document to determine the overall confidence level of the entire claim document, i.e., the second confidence level. Finally, adding the second confidence level to the target processing recommendation, so that the target processing recommendation includes the overall confidence level of the claim document, enabling reviewers to determine whether to adopt the target processing recommendation based on the overall confidence level.

[0014] In some technical solutions, optionally, the step of determining a second confidence level based on multiple first confidence levels includes: determining the weight corresponding to each data matrix based on the content information of each data matrix; and determining the second confidence level based on multiple weights and multiple first confidence levels.

[0015] In this technical solution, the step of determining the second confidence level based on multiple first confidence levels includes: First, determining the weight corresponding to each data matrix based on the content information of each data matrix. It can be understood that in the claims process, multiple data points need to be considered comprehensively; therefore, each data matrix corresponds to one data point, and the data point corresponding to that data matrix can be clearly determined based on its content information. Furthermore, different data points have different weights when forming the processing result. Therefore, the weight corresponding to each data matrix can be determined based on its content information. Further, the second confidence level is determined based on multiple weights and multiple first confidence levels; that is, a weighted average algorithm can be used to determine the final comprehensive confidence level.

[0016] In some technical solutions, optionally, after the step of determining the target processing recommendation based on the evidence chain knowledge graph and the claims rule database, the following steps are included: displaying the target processing recommendation, the claims documents, and the evidence chain knowledge graph.

[0017] In this technical solution, after determining the target processing suggestion based on the evidence chain knowledge graph and the claims rule database, the following steps are included: displaying the obtained target processing suggestion, the user-uploaded claim documents, and the constructed evidence chain knowledge graph, so that the reviewers can see the target processing suggestion, the claim documents, and the evidence chain knowledge graph. This allows the reviewers to complete the claims according to the target processing suggestion, and also allows them to re-verify the claims based on the claim documents and the evidence chain knowledge graph, thereby ensuring the accuracy of the claims.

[0018] In some technical solutions, the file types may optionally include one or more combinations of images, chat logs, invoices, and emails; the business scenarios may include one or more combinations of car repair scenarios, product warranty scenarios, and car warranty scenarios.

[0019] In this technical solution, file types can include images, chat logs, invoices, and emails. By converting various heterogeneous claim files, including images, chat logs, invoices, and emails, standardized data that preserves spatial semantic relationships is obtained, facilitating subsequent data model processing. Furthermore, the business scenarios can include car repair scenarios, product warranty scenarios, and vehicle warranty scenarios. Through business scenario identification, in-depth optimization is performed for different businesses, ensuring the accuracy and professionalism of information extraction.

[0020] In some technical solutions, the target processing recommendations may optionally include: judgment conclusion, basis of terms, recommended amount, and overall confidence level.

[0021] In this technical solution, the target processing recommendation can include information such as: judgment conclusion, clause basis, recommended amount, and overall confidence level. By setting the target processing recommendation to include information such as judgment conclusion, clause basis, recommended amount, and overall confidence level, reviewers can easily complete the claims review according to the target processing recommendation.

[0022] According to a second aspect of the present invention, an intelligent claims processing system is proposed, comprising: a first receiving module for receiving claims documents, wherein the claims documents include multiple claim sub-files of different file types; a format conversion module for converting the format of each claim sub-file to obtain multiple first files; a first processing module for processing each first file to obtain multiple data matrices, wherein the data matrices contain spatial semantic relationships; a second processing module for determining the business scenario corresponding to the claims document based on the multiple data matrices and a visual model; a third processing module for processing the multiple data matrices based on the business scenario to obtain multiple first data information; a fourth processing module for constructing an evidence chain knowledge graph with time and logical relationships as edges based on the multiple first data information; a first determining module for determining a claims rule database based on the business scenario; and a second determining module for determining a target processing suggestion based on the evidence chain knowledge graph and the claims rule database.

[0023] The intelligent claim processing system provided by this invention specifically includes: a first receiving module, a format conversion module, a first processing module, a second processing module, a third processing module, a fourth processing module, a first determining module, and a second determining module. The first receiving module first receives claim documents, which can be uploaded by the user or by the manufacturer involved in the claim. Claim documents refer to various supporting materials used to prove the legality and authenticity of the claim request during the claim process. Claim documents can include multiple claim sub-files of different file types. For example, claim sub-files can be accident scene photos, repair expense invoices, and customer service communication texts, etc. These claim sub-files have different file types; for example, accident scene photos are images, repair expense invoices are invoices, and customer service communication texts are chat logs. Upon receiving the claim documents, the different claim sub-files have different file types and therefore different file formats, such as JPG (Joint Photographic Experts Group), PNG (Portable Network Graphics), PDF (Portable Document Format), and TXT (Plain Text). Therefore, the format conversion module needs to convert each claim sub-file into multiple first files with a unified format. The first processing module then processes these first files separately to obtain multiple standard data matrices that preserve the original spatial semantic relationships, facilitating subsequent model processing. After obtaining the multiple data matrices, the second processing module inputs them into a pre-trained Large Vision Model (LVM). The LVM performs an overall analysis of the multiple data matrices to determine the business scenario corresponding to the claim document. This business scenario could include automotive repair claims, product warranty identification, and automotive warranty claims, among others. After determining the business scenario corresponding to the claim documents, the third processing module determines the corresponding processing flow based on the business scenario. Then, it processes multiple data matrices according to the processing flow, extracting key information, i.e., the first data information, from each data matrix. Subsequently, the fourth processing module merges the multiple first data information sets to construct an evidence chain knowledge graph with time and logical relationships as edges. Simultaneously, the first determining module determines the corresponding claims rule database based on the determined business scenario. Different claims rule databases store different claims rules; for example, claims rules can be insurance claim clauses and product warranty policies. Therefore, it is necessary to determine the corresponding claims rule database based on the determined business scenario.Finally, the second determination module matches and calculates the evidence chain knowledge graph with the claims rule database to obtain the final target processing suggestion. The target processing suggestion includes whether to process the claim and how to process it. This invention automates the entire process from document receipt to decision suggestion, freeing reviewers from tedious repetitive work, greatly improving processing efficiency, and reducing labor costs. Furthermore, because this invention uses the same claims rule database for the same business scenarios, it solves the problem of inconsistent claims standards in related technologies.

[0024] According to a third aspect of the present invention, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the smart claim processing method as described above.

[0025] The electronic device provided by this invention, when the processor executes the computer program, implements the steps of the above-mentioned intelligent claim processing method, and can achieve the technical effects of any of the above technical solutions, which will not be elaborated here.

[0026] According to a fourth aspect of the invention, a storage medium is provided on which a computer program is stored, which, when executed by a processor, implements the steps of the intelligent claims processing method as described above.

[0027] The storage medium provided by this invention, when the computer program is executed by the processor, implements the steps of the above-described intelligent claim processing method, and can achieve the technical effects of any of the above technical solutions, which will not be elaborated here.

[0028] Additional aspects and advantages of the invention will become apparent in the following description or may be learned by practice of the invention. Attached Figure Description

[0029] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 A flowchart illustrating an embodiment of the intelligent claims processing method of the present invention is shown; Figure 2 The diagram illustrates a step in an intelligent claims processing method according to an embodiment of the present invention: processing multiple data matrices according to a business scenario to obtain multiple first data information. Figure 3 This diagram illustrates a flow chart of an intelligent claims processing method according to an embodiment of the present invention, following the step of determining a target processing suggestion based on an evidence chain knowledge graph and a claims rule database. Figure 4The diagram illustrates a step in determining a second confidence level based on multiple first confidence levels in an intelligent claims processing method according to an embodiment of the present invention. Figure 5 One of the schematic block diagrams of an embodiment of the intelligent claims processing system of the present invention is shown; Figure 6 A second schematic block diagram of an intelligent claims processing system according to an embodiment of the present invention is shown; Figure 7 A schematic diagram of the principle of an intelligent claims processing system according to an embodiment of the present invention is shown; Figure 8 A schematic block diagram of an electronic device according to an embodiment of the present invention is shown. Detailed Implementation

[0030] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other.

[0031] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0032] Figure 1 A flowchart illustrating an embodiment of the intelligent claims processing method of the present invention is shown. Specifically, it includes: Step 102: Receive the claim documents, which include multiple claim sub-files of different document types; Step 104: Convert the format of each claim sub-file to obtain multiple first files; Step 106: Process each first file to obtain multiple data matrices, where each data matrix contains spatial semantic relationships; Step 108: Determine the business scenario corresponding to the claim document based on multiple data matrices and visual models; Step 110: Process multiple data matrices according to the business scenario to obtain multiple first data information; Step 112: Construct an evidence chain knowledge graph with time and logical relationships as edges based on multiple primary data information; Step 114: Determine the claims rules database based on the business scenario; Step 116: Determine the target processing recommendation based on the evidence chain knowledge graph and the claims rules database.

[0033] The intelligent claim processing method provided by this invention specifically includes: firstly, receiving claim documents, which can be uploaded by users or by the manufacturers involved in the claim. Claim documents refer to various supporting materials used to prove the legality and authenticity of the claim request during the claim process. Claim documents can include multiple claim sub-files of different file types. For example, claim sub-files can be accident scene photos, repair expense invoices, and customer service communication texts, etc. These claim sub-files have different file types; for example, accident scene photos are images, repair expense invoices are invoices, and customer service communication texts are chat logs. After receiving the claim documents, because the different claim sub-files have different file types, their file formats are also different, such as JPG (Joint Photographic Experts Group), PNG (Portable Network Graphics), PDF (Portable Document Format), and TXT (Plain Text), etc. Therefore, it is necessary to convert each claim sub-file into multiple first files with a unified format. Subsequently, multiple first files are processed separately to obtain multiple standard data matrices that retain the original spatial semantic relationships, facilitating subsequent model processing. After obtaining multiple data matrices, they are input into a pre-trained Large Vision Model (LVM). The visual model performs an overall analysis of the multiple data matrices to obtain the business scenario corresponding to the claim document. This business scenario could include car repair claims, product warranty identification, and car warranty claims, etc. After determining the business scenario corresponding to the claim document, a corresponding processing flow can be determined based on the business scenario. Then, the multiple data matrices are processed according to the processing flow to extract the key information, i.e., the first data information, from each data matrix. The multiple first data information are then fused to construct an evidence chain knowledge graph with time and logical relationships as edges. Simultaneously, a corresponding claims rule database is determined based on the determined business scenario. Different claims rule databases store different claims rules; for example, claims rules could be insurance claim terms and product warranty policies. Therefore, it is necessary to determine the corresponding claims rule database based on the determined business scenario. Finally, the evidence chain knowledge graph is matched and calculated with the claims rule database to obtain the final target processing suggestion. The target processing suggestion includes whether to process the claim and how to process it. This invention automates the entire process from document reception to decision suggestion, freeing reviewers from tedious repetitive work, greatly improving processing efficiency, and reducing labor costs.Furthermore, since the same claims rule database is used to process the same business scenarios in this invention, the problem of inconsistent claims standards in related technologies is solved.

[0034] Figure 2 The diagram illustrates a step in an intelligent claims processing method according to an embodiment of the present invention: processing multiple data matrices according to a business scenario to obtain multiple first data information. The step of processing multiple data matrices according to a business scenario to obtain multiple first data information includes: Step 202: Determine the first data model set based on the business scenario. The first data model set includes multiple second data models. Step 204: Determine the content information in each data matrix; Step 206: Determine the second data model corresponding to each data matrix based on the content information; Step 208: Obtain multiple first data information based on multiple data matrices and the second data model corresponding to each data matrix.

[0035] In this embodiment, the steps of processing multiple data matrices according to a business scenario to obtain multiple first data information include: firstly, determining a first data model set based on the business scenario, wherein the first data model set corresponds to the business processing flow; that is, after determining the business scenario, determining the business processing flow required for the business scenario, and then determining the corresponding first data model set based on the business processing flow. The first data model set includes multiple second data models. It can be understood that during the claims process, multiple aspects of data need to be comprehensively processed, therefore each second data model corresponds to a specific aspect of the data processing process. Next, determining the content information in each data matrix; for example, the content information could be photos of an accident or invoices for incurred expenses. Then, determining the second data model corresponding to each data matrix based on the content information; that is, determining the data processing flow required for the data matrix based on the content information of the data matrix. Finally, inputting the multiple data matrices into their corresponding second data models, processing the data matrices using the second data models, extracting key information, and outputting the corresponding first data information, wherein the first data information is standardized structured data. By determining a first set of data models based on business scenarios, and then determining a corresponding second data model from the first set of data models based on the content of each data matrix, the second data model is then used to extract key information from the data matrix. This achieves in-depth optimization for different business scenarios and ensures the accuracy and professionalism of information extraction.

[0036] Figure 3The diagram illustrates a flowchart of an intelligent claims processing method according to an embodiment of the present invention, following the step of determining a target processing suggestion based on an evidence chain knowledge graph and a claims rule database; wherein, following the step of determining the target processing suggestion based on the evidence chain knowledge graph and the claims rule database, the method includes: Step 302: Determine the first confidence level corresponding to each data matrix based on multiple data matrices and the visual model; Step 304: Determine the second confidence level based on multiple first confidence levels; Step 306: Add the second confidence level to the target processing suggestion.

[0037] In this embodiment, after determining the target processing recommendation based on the evidence chain knowledge graph and the claims rule database, the method further includes: firstly, inputting multiple data matrices into a visual model, and using the visual model to analyze the content of each data matrix to determine the first confidence level corresponding to each data matrix, i.e., determining the credibility of each claim sub-document. The visual model used in this embodiment can be the same as the one used to determine the business scenario; that is, while using the visual model to determine the business scenario corresponding to multiple data matrices, the confidence level corresponding to each data matrix is ​​also determined. Then, the multiple first confidence levels are fused to determine the second confidence level; that is, the first confidence levels corresponding to each claim sub-document are fused to determine the overall confidence level of the entire claim document, i.e., the second confidence level. Finally, the second confidence level is added to the target processing recommendation, so that the target processing recommendation includes the overall confidence level of the claim document. This allows reviewers to determine whether to adopt the target processing recommendation based on the overall confidence level.

[0038] Figure 4 The diagram illustrates a step in a smart claim processing method according to an embodiment of the present invention to determine a second confidence level based on multiple first confidence levels; wherein, the step of determining the second confidence level based on multiple first confidence levels includes: Step 402: Determine the weight of each data matrix based on the content information of each data matrix; Step 404: Determine the second confidence level based on multiple weights and multiple first confidence levels.

[0039] In this embodiment, the step of determining the second confidence level based on multiple first confidence levels includes: firstly, determining the weight corresponding to each data matrix based on the content information of each data matrix. It is understood that in the claims process, multiple data points need to be considered comprehensively; therefore, each data matrix corresponds to one data point, and the data point corresponding to that data matrix can be clearly determined based on its content information. Furthermore, different data points have different weights when forming the processing result. Therefore, the weight corresponding to each data matrix can be determined based on its content information. Further, determining the second confidence level based on multiple weights and multiple first confidence levels, i.e., using a weighted average algorithm to determine the final comprehensive confidence level.

[0040] In some embodiments, optionally after the step of determining the target processing recommendation based on the evidence chain knowledge graph and the claims rule database, the method includes: displaying the target processing recommendation, the claims documents, and the evidence chain knowledge graph.

[0041] In this embodiment, after determining the target processing suggestion based on the evidence chain knowledge graph and the claims rule database, the process includes: displaying the obtained target processing suggestion, the user-uploaded claim documents, and the constructed evidence chain knowledge graph, so that the reviewers can see the target processing suggestion, the claim documents, and the evidence chain knowledge graph. This allows the reviewers to complete the claims process according to the target processing suggestion, and also allows them to re-verify the claims based on the claim documents and the evidence chain knowledge graph, thereby ensuring the accuracy of the claims process.

[0042] In some embodiments, the file type may optionally include one or more combinations of images, chat logs, invoices, and emails; the business scenario may include one or more combinations of car repair scenarios, product warranty scenarios, and car warranty scenarios.

[0043] In this embodiment, file types can include images, chat logs, invoices, and emails. By converting various heterogeneous claim files, including images, chat logs, invoices, and emails, standardized data that retains spatial semantic relationships is obtained, facilitating subsequent data model processing. Furthermore, business scenarios can include car repair scenarios, product warranty scenarios, and vehicle warranty scenarios. By identifying business scenarios and performing in-depth optimization for different businesses, the accuracy and professionalism of information extraction are ensured.

[0044] In some embodiments, the target processing recommendation may optionally include: a judgment conclusion, the basis of the terms, the recommended amount, and the overall confidence level.

[0045] In this embodiment, the target processing recommendation may include information such as: judgment conclusion, clause basis, recommended amount, and overall confidence level. By setting the target processing recommendation to include information such as judgment conclusion, clause basis, recommended amount, and overall confidence level, reviewers can easily complete the claims review according to the target processing recommendation.

[0046] Figure 5 This diagram illustrates one embodiment of an intelligent claim processing system according to the present invention. The intelligent claim processing system includes: a file input module, a format unification and conversion module, a multimodal classification module, a dynamic routing processing module, a scene adaptation processor, a rule decision engine, and a human-machine collaborative review module. Specifically, firstly, the file input module receives various heterogeneous claim files submitted by users, including images, chat logs, invoices, emails, etc. Then, the format unification and conversion module uses a layout-aware format conversion engine to unify different file formats (such as JPG, PNG, PDF, and TXT) into a standard data matrix that preserves the original spatial semantic relationships, facilitating subsequent model processing. Next, the multimodal classification module inputs the standard data matrix into a pre-trained Large Vision Model (LVM). The visual model performs an overall analysis of the file content and outputs the business scenario to which the batch of files belongs (e.g., car repair claim, product warranty identification, and car warranty claim) and the corresponding confidence level. Then, the dynamic routing processing module dynamically distributes the data to the corresponding scene adaptation processor based on the previously output classification results, i.e., the business scenario. It is understood that the processes performed in different business scenarios are different, such as... Figure 7 As shown, the classification results are divided into three scenarios: automotive repair, product warranty, and medical insurance. The automotive repair scenario requires a dedicated automotive claims process, which includes damage analysis and component identification. The product warranty scenario requires a dedicated warranty identification process, which includes serial number verification and warranty period calculation. The medical insurance scenario requires a dedicated medical claims process, which includes medical record parsing and cost matching. Therefore, data from different business scenarios needs to be distributed to different scenario adaptation processors. The scenario adaptation processors include corresponding dedicated processing models, as shown in Table 1. Table 1

[0047] Table 1 shows that for car repair claims, the dedicated processing model is a vehicle damage analysis model, which extracts key identification elements such as the damaged area, repair items, and responsible party. For product warranty identification, the dedicated processing model is a serial number OCR (Optical Character Recognition) + warranty period calculation model, which extracts key identification elements such as product model, purchase date, and fault type. For maintenance service claims, the dedicated processing model is a maintenance record analysis model, which extracts key identification elements such as maintenance history, project compliance, and mileage.

[0048] Each dedicated processing model extracts key information and outputs it as standardized structured data (such as JSON format). Then, it automatically merges all sourced structured data to construct a contextualized evidence chain knowledge graph with time and logical relationships as edges. Subsequently, the evidence chain knowledge graph is input into the rule decision engine. This engine matches and calculates the evidence chain knowledge graph based on a pre-defined, configurable contextualized rule base (such as insurance claim clauses and product warranty policies), ultimately generating a processing suggestion with a judgment conclusion, clause basis, suggested amount, and overall confidence level. Finally, in the human-machine collaborative review module, the generated processing suggestion, key evidence chain, and original documents are presented together on the module's visual interface. Fields with low confidence levels are highlighted on the interface for reviewers to quickly locate, review, and correct, ultimately completing the review. The review results can be used for model retraining, forming a closed-loop optimization.

[0049] Compared with related technologies, the advantages of this invention are as follows: This invention automates the entire process from document reception to decision-making recommendations, freeing reviewers from tedious repetitive tasks and greatly improving processing efficiency. Simultaneously, this invention, through scene recognition and dedicated model invocation, performs in-depth optimization for different business needs, ensuring the accuracy and professionalism of information extraction. Furthermore, this invention adopts a modular design, allowing for easy configuration to add new claim scenarios and their corresponding processing flows and rules, flexibly adapting to business development needs. Moreover, by constructing an evidence chain knowledge graph, this invention can effectively identify contradictions or non-compliance points among multiple materials, enhancing the ability to identify potential fraudulent activities. Finally, this invention, through the establishment of a human-machine collaborative interface, clearly displays the system's decision-making process and key evidence, assisting humans in efficiently and accurately completing the final review, ensuring the transparency and explainability of the decision.

[0050] Figure 6A second schematic block diagram of an intelligent claims processing system according to an embodiment of the present invention is shown; wherein, the intelligent claims processing system 60 includes: The first receiving module 602 is used to receive claim documents, wherein the claim documents include multiple claim sub-files of different file types; The format conversion module 604 is used to convert the format of each claim sub-file separately to obtain multiple first files; The first processing module 606 is used to process each first file to obtain multiple data matrices, wherein the data matrices contain spatial semantic relationships; The second processing module 608 is used to determine the business scenario corresponding to the claim document based on multiple data matrices and visual models; The third processing module 610 is used to process multiple data matrices according to the business scenario to obtain multiple first data information. The fourth processing module 612 is used to construct an evidence chain knowledge graph with time and logical relationships as edges based on multiple first data information. The first determination module 614 is used to determine the claims rule database based on the business scenario; The second determination module 616 is used to determine the target processing recommendation based on the evidence chain knowledge graph and the claims rule database.

[0051] The intelligent claim processing system 60 provided by this invention specifically includes: a first receiving module 602, a format conversion module 604, a first processing module 606, a second processing module 608, a third processing module 610, a fourth processing module 612, a first determining module 614, and a second determining module 616. The first receiving module 602 first receives claim documents, which can be uploaded by users or by the manufacturers involved in the claim. Claim documents refer to various supporting materials used to prove the legality and authenticity of the claim request during the claim process. Claim documents can include multiple claim sub-files of different file types. For example, claim sub-files can be accident scene photos, repair expense invoices, and customer service communication texts, etc. These claim sub-files have different file types; for example, accident scene photos are images, repair expense invoices are invoices, and customer service communication texts are chat logs. Upon receiving the claim documents, the different claim sub-files have different file types and therefore different file formats, such as JPG (Joint Photographic Experts Group), PNG (Portable Network Graphics), PDF (Portable Document Format), and TXT (Plain Text). Therefore, the format conversion module 604 needs to convert each claim sub-file into multiple first files with a unified format. Subsequently, the first processing module 606 processes these first files separately to obtain multiple standard data matrices that retain the original spatial semantic relationships, facilitating subsequent model processing. After obtaining the multiple data matrices, the second processing module 608 inputs them into a pre-trained Large Vision Model (LVM). The LVM performs an overall analysis of the multiple data matrices to determine the business scenario corresponding to the claim documents. This business scenario could include automotive repair claims, product warranty identification, and automotive warranty claims, among others. After determining the business scenario corresponding to the claim documents, the third processing module 610 determines the corresponding processing flow based on the business scenario, and then processes multiple data matrices according to the processing flow to extract the key information, i.e., the first data information, from each data matrix. Subsequently, the fourth processing module 612 merges the multiple first data information to construct an evidence chain knowledge graph with time and logical relationships as edges. At the same time, the first determination module 614 determines the corresponding claims rule database based on the determined business scenario. Different claims rule databases store different claims rules; for example, claims rules can be insurance claim terms and product warranty policies, etc. Therefore, it is necessary to determine the corresponding claims rule database based on the determined business scenario.Finally, the second determining module 616 matches and calculates the evidence chain knowledge graph with the claims rule database to obtain the final target processing suggestion. The target processing suggestion includes whether to process the claim and how to process it. This invention automates the entire process from document reception to decision suggestion, freeing reviewers from tedious repetitive work, greatly improving processing efficiency, and reducing labor costs. Furthermore, because this invention uses the same claims rule database for processing the same business scenarios, it solves the problem of inconsistent claims standards in related technologies.

[0052] In some embodiments, optionally, the third processing module 610 is specifically used to determine a first data model set according to a business scenario, the first data model set including multiple second data models; determine the content information in each data matrix; determine the second data model corresponding to each data matrix according to the content information; and obtain multiple first data information based on the multiple data matrices and the second data model corresponding to each data matrix.

[0053] In this embodiment, the third processing module 610 is specifically used to first determine a first data model set based on the business scenario. This first data model set corresponds to the business processing flow; that is, after determining the business scenario, the required business processing flow is determined, and then the corresponding first data model set is determined based on the business processing flow. The first data model set includes multiple second data models. It can be understood that during the claims process, multiple aspects of data need to be comprehensively processed, so each second data model corresponds to a specific aspect of the data processing process. Next, the content information in each data matrix is ​​determined, and then the second data model corresponding to each data matrix is ​​determined based on the content information. In other words, the data processing process to be performed on the data matrix is ​​determined based on the content information of the data matrix. Finally, the multiple data matrices are input into their respective second data models, and the second data models are used to process the data matrices, extract key information, and output corresponding first data information, where the first data information is standardized structured data. By determining a first set of data models based on business scenarios, and then determining a corresponding second data model from the first set of data models based on the content of each data matrix, the second data model is then used to extract key information from the data matrix. This achieves in-depth optimization for different business scenarios and ensures the accuracy and professionalism of information extraction.

[0054] In some embodiments, the intelligent claims processing system 60 may optionally include: a third determining module for determining a first confidence level corresponding to each data matrix based on a plurality of data matrices and a visual model; a fourth determining module for determining a second confidence level based on a plurality of first confidence levels; and a fifth processing module for adding the second confidence level to a target processing recommendation.

[0055] In this embodiment, the intelligent claims processing system 60 further includes a third determining module, a fourth determining module, and a fifth processing module. The third determining module first inputs multiple data matrices into a visual model, using the visual model to analyze the content of each data matrix and determine the first confidence level corresponding to each data matrix, i.e., determining the credibility of each claim sub-file. Subsequently, the fourth determining module fuses the multiple first confidence levels to determine a second confidence level; that is, it fuses the first confidence levels corresponding to each claim sub-file to determine the overall confidence level of the entire claim file, i.e., the second confidence level. Finally, the fifth processing module adds the second confidence level to the target processing suggestion, so that the target processing suggestion includes the overall confidence level of the claim file. This allows the system to determine whether to adopt the target processing suggestion based on the overall confidence level.

[0056] In some embodiments, optionally, the fourth determining module is specifically configured to determine the weight corresponding to each data matrix based on the content information of each data matrix; and to determine a second confidence level based on multiple weights and multiple first confidence levels.

[0057] In this embodiment, the fourth determining module is specifically used to first determine the weight corresponding to each data matrix based on the content information of each data matrix. It can be understood that in the claims process, multiple data points need to be considered comprehensively. Therefore, each data matrix corresponds to one data point, and the data point corresponding to each data matrix can be clearly determined based on its content information. Furthermore, different data points have different weights when forming the processing result. Therefore, the weight corresponding to each data matrix can be determined based on its content information. Further, a second confidence level is determined based on multiple weights and multiple first confidence levels; that is, a weighted average algorithm can be used to determine the final comprehensive confidence level.

[0058] In some embodiments, the intelligent claims processing system 60 may optionally include a display module for displaying target processing suggestions, claims documents, and a knowledge graph of the chain of evidence.

[0059] In this embodiment, the intelligent claims processing system 60 further includes a display module. The display module can show the target processing suggestion, the user-uploaded claim documents, and the constructed evidence chain knowledge graph. This allows reviewers to see the target processing suggestion, claim documents, and evidence chain knowledge graph, enabling them to complete the claim processing according to the target processing suggestion and to re-verify the claim documents based on the evidence chain knowledge graph, thus ensuring the accuracy of the claim processing.

[0060] In some embodiments, the file type may optionally include one or more combinations of images, chat logs, invoices, and emails; the business scenario may include one or more combinations of car repair scenarios, product warranty scenarios, and car warranty scenarios.

[0061] In this embodiment, file types can include images, chat logs, invoices, and emails. By converting various heterogeneous claim files, including images, chat logs, invoices, and emails, standardized data that retains spatial semantic relationships is obtained, facilitating subsequent data model processing. Furthermore, business scenarios can include car repair scenarios, product warranty scenarios, and vehicle warranty scenarios. By identifying business scenarios and performing in-depth optimization for different businesses, the accuracy and professionalism of information extraction are ensured.

[0062] In some embodiments, the target processing recommendation may optionally include: a judgment conclusion, the basis of the terms, the recommended amount, and the overall confidence level.

[0063] In this embodiment, the target processing recommendation may include information such as: judgment conclusion, clause basis, recommended amount, and overall confidence level. By setting the target processing recommendation to include information such as judgment conclusion, clause basis, recommended amount, and overall confidence level, reviewers can easily complete the claims review according to the target processing recommendation.

[0064] Figure 8 A schematic block diagram of an electronic device according to an embodiment of the present invention is shown; wherein, the electronic device 80 includes a memory 802, a processor 804, and a computer program stored in the memory 802 and executable on the processor 804, wherein the processor 804 executes the computer program to implement the steps of the smart claim processing method as described above.

[0065] The electronic device 80 provided by the present invention, when the processor 804 executes the computer program, implements the steps of the above-described intelligent claim processing method, and can achieve the technical effects of any of the above embodiments, which will not be repeated here.

[0066] One embodiment of the present invention provides a storage medium on which a computer program is stored, which, when executed by a processor, implements the steps of the smart claims processing method as described above.

[0067] The storage medium provided by this invention enables the computer program to implement the steps of the above-described intelligent claim processing method when executed by a processor, achieving the technical effects of any of the above embodiments, which will not be elaborated further.

[0068] In the description of this specification, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance, unless otherwise expressly specified and limited. The terms "connection," "installation," and "fixing," etc., should be interpreted broadly. For example, "connection" can mean a fixed connection, a detachable connection, or an integral connection; it can mean a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0069] In the description of this specification, the terms "one embodiment," "some embodiments," "specific embodiment," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0070] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A smart claims processing method, characterized in that, include: Receive claim documents, wherein the claim documents include multiple claim sub-files of different file types; Each of the claim sub-files is converted to a different format to obtain multiple first files; Each of the first files is processed to obtain multiple data matrices, wherein the data matrices contain spatial semantic relationships; The business scenario corresponding to the claim document is determined based on multiple data matrices and visual models; The multiple data matrices are processed according to the business scenario to obtain multiple first data information; Construct an evidence chain knowledge graph with time and logical relationships as edges based on multiple first data information; The claims rules database is determined based on the aforementioned business scenario; Target processing recommendations are determined based on the evidence chain knowledge graph and the claims rules database.

2. The intelligent claim processing method according to claim 1, characterized in that, The step of processing multiple data matrices according to the business scenario to obtain multiple first data information includes: A first data model set is determined based on the business scenario, and the first data model set includes multiple second data models; Determine the content information in each of the data matrices; Determine the second data model corresponding to each data matrix based on the content information; Multiple sets of first data information are obtained based on multiple data matrices and the second data model corresponding to each data matrix.

3. The intelligent claim processing method according to claim 1, characterized in that, Following the step of determining the target processing recommendation based on the evidence chain knowledge graph and the claims rule database, the method further includes: A first confidence level is determined for each of the data matrices based on the multiple data matrices and the visual model; A second confidence level is determined based on multiple first confidence levels; Add the second confidence level to the target processing recommendation.

4. The intelligent claim processing method according to claim 3, characterized in that, The step of determining the second confidence level based on multiple first confidence levels includes: The weights corresponding to each data matrix are determined based on the content information of each data matrix. The second confidence level is determined based on the multiple weights and the multiple first confidence levels.

5. The intelligent claim processing method according to claim 1, characterized in that, Following the step of determining the target processing recommendation based on the evidence chain knowledge graph and the claims rule database, the following is included: The target processing suggestions, the claim documents, and the evidence chain knowledge graph are displayed.

6. The intelligent claims processing method according to any one of claims 1 to 5, characterized in that, The file types include one or more combinations of images, chat logs, invoices, and emails; The business scenarios include: car repair scenarios, product warranty scenarios, and one or more combinations of car warranty scenarios.

7. The intelligent claims processing method according to any one of claims 1 to 5, characterized in that, The proposed target processing recommendations include: the judgment conclusion, the basis of the terms, the recommended amount, and the overall confidence level.

8. An intelligent claims processing system, characterized in that, include: A first receiving module is configured to receive claim documents, wherein the claim documents include multiple claim sub-files of different file types; A format conversion module is used to convert the format of each claim sub-file to obtain multiple first files; A first processing module is configured to process each of the first files to obtain multiple data matrices, wherein the data matrices contain spatial semantic relationships; The second processing module is used to determine the business scenario corresponding to the claim document based on multiple data matrices and visual models. The third processing module is used to process the multiple data matrices according to the business scenario to obtain multiple first data information. The fourth processing module is used to construct an evidence chain knowledge graph with time and logical relationships as edges based on multiple first data information. The first determining module is used to determine the claims rule database according to the business scenario; The second determining module is used to determine the target processing suggestion based on the evidence chain knowledge graph and the claims rule database.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent claims processing method as described in any one of claims 1 to 7.

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

Images