Claim settlement information auditing method and system, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a claim settlement information auditing method, a claim settlement information auditing system, electronic equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring multi-modal and multi-dimensional evidence information associated with an insurance policy to be subjected to claim settlement; adopting a plurality of information processing agents to execute information processing operation of multiple dimensions and multiple modes on the evidence information to obtain processing results generated by the information processing agents respectively; and performing reasoning judgment on the processing result generated by each information processing agent by adopting a reasoning agent based on a preset auditing rule to obtain a responsibility judgment result of the insurance policy to be subjected to claim settlement. The method effectively improves the auditing efficiency and accuracy of the insurance policy to be subjected to claim settlement.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a claims information verification method, a claims information verification system, an electronic device, a storage medium, and a computer program product. Background Technology

[0002] In existing technologies, most e-commerce platforms offer users protection-type insurance, such as return shipping insurance. This type of insurance is purchased by the e-commerce platform from an insurance company. When a buyer applies for a return, it triggers the platform's after-sales return and refund process, which includes initiating the insurance company's claims process. In the claims process, the platform needs to provide the insurance company with a manual review interface to complete the claims review. However, during peak periods with high claims volumes, manual review is not only inefficient but also suffers from inconsistent standards and results. This is especially true in cross-border e-commerce, where cross-border logistics and multilingual documents further complicate manual review.

[0003] It is evident that existing methods for verifying claims information still require improvement. Summary of the Invention

[0004] This application provides a method for reviewing claims information, which can effectively improve the accuracy and efficiency of claims information review.

[0005] Accordingly, embodiments of this application also provide a claims information verification system, an electronic device, a storage medium, and a computer program product to ensure the implementation and application of the aforementioned claims information verification method.

[0006] To solve the above-mentioned technical problems, this application is implemented as follows: In a first aspect, embodiments of this application provide a method for verifying claims information, applied on a server side, the method comprising: Obtain multimodal and multidimensional evidentiary information related to the insurance policy pending claim; Multiple information processing agents are employed to perform multi-dimensional and multi-modal information processing operations on the evidence information, resulting in processing results generated by each of the information processing agents. An inference agent, based on preset review rules, infers and judges the processing results generated by each information processing agent to obtain the liability determination result of the insurance policy pending claim.

[0007] Secondly, this application provides a method for verifying claims information, applied to a client-side application, the method comprising: In response to the manual review of the audit report, obtain the data of the evidence chain diagram corresponding to the audit report; The evidence chain diagram is based on the data shown. The marked processing results and / or liability determination results are highlighted in the displayed evidence chain graph; wherein, the evidence chain graph is established by the pre-set server using the following method: performing unified timeline mapping processing on the evidence information and processing results of the audit report, mapping the evidence information and processing results to a unified timeline; associating the liability determination results with the corresponding evidence information on the unified timeline to obtain the evidence chain graph; and marking the processing results and / or liability determination results in the evidence chain graph that indicate an audit anomaly.

[0008] Thirdly, this application provides a claims information verification system, which includes a client and a server, wherein... The server is used to execute the steps of the method described in the first aspect; The client is used to perform the steps described in the second aspect.

[0009] Fourthly, embodiments of this application provide an electronic device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the method as described in the first or second aspect.

[0010] Fifthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first or second aspect.

[0011] Sixthly, embodiments of this application provide a computer program product, including a computer program / computer executable instructions, which, when executed by a processor in an electronic device, implement the method described in the first or second aspect.

[0012] Compared with the prior art, the embodiments of this application have the following advantages: By acquiring multimodal and multidimensional evidence information associated with the pending insurance claims from the server, multiple information processing agents are then employed to perform multimodal and multidimensional information processing operations on the evidence information. Each agent generates its own processing result, and these agents can process different modalities and dimensions of evidence information in parallel, thus improving the efficiency of the pending insurance claims review. Furthermore, a reasoning agent, based on preset review rules, infers and judges the processing results generated by each agent to obtain the liability determination result for the pending insurance claims, effectively improving the accuracy of the review. On the other hand, this method also effectively improves the review efficiency of pending insurance claims by automatically generating review reports. Attached Figure Description

[0013] Figure 1 This is one of the steps in the claims information review method disclosed in the embodiments of this application; Figure 2 This is a schematic diagram illustrating the principle of the claims information review method disclosed in the embodiments of this application; Figure 3 This is the second step in the flowchart of the claims information review method disclosed in the embodiments of this application; Figure 4 This is the third step in the flowchart of the claims information review method disclosed in the embodiments of this application; Figure 5 This is a schematic diagram of the client-server interaction in the claims information review system disclosed in the embodiments of this application; Figure 6 This is a schematic diagram of the structure of an exemplary device provided in one embodiment of this application. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] The claims information review method disclosed in this application improves the accuracy and efficiency of claims information review by automatically analyzing evidence information and combining it with review rules to obtain review results through multimodal information pre-review and material standardization, multi-agent division of labor and decision-making arrangement.

[0016] like Figure 1 As shown in the embodiment of this application, the claims information review method is applied to the server. The method includes steps 102 to 106.

[0017] The following is combined Figure 2 The schematic diagram illustrating the implementation principle of the claims information verification method explains the specific implementation methods of each step.

[0018] Step 102: Obtain multimodal and multidimensional evidence information related to the policy to be claimed.

[0019] In the cross-border e-commerce sector, the pending claims policies include, but are not limited to, return and exchange shipping insurance policies. The multimodal data includes, but is not limited to, one or more of the following modalities: images, videos, audio, and text. The multimodal, multi-dimensional evidentiary information associated with the pending claims policies includes, but is not limited to, one or more of the following: product name, product image, product video, logistics information, communication records, product order information, product order status, and policy information.

[0020] In practice, the aforementioned evidentiary information can be obtained through the interfaces of e-commerce platforms and insurance platforms. For example, users can communicate online with merchants on e-commerce platforms regarding returns and exchanges, and submit a return application through the e-commerce platform's client (hereinafter referred to as the "second preset client"), uploading photos and / or videos and / or audio of the returned goods, as well as information such as the reason for the return or exchange. The e-commerce platform obtains the photos and / or videos and / or audio uploaded by the user, as well as the text entered by the user, as part of the evidentiary information. On the other hand, the e-commerce platform obtains the conversation records of the online communication between the user and the merchant regarding returns and exchanges, as part of the evidentiary information. At the same time, the e-commerce platform will use the product order information, logistics information, product order status, and shipping insurance policy information associated with the current return or exchange operation as part of the evidentiary information, thereby obtaining evidence information of the insurance policy pending claim.

[0021] Optionally, obtaining multimodal and multidimensional evidence information associated with the policy pending claim includes: de-identifying the multimodal and multidimensional evidence information associated with the policy pending claim to obtain de-identified evidence information; standardizing the de-identified evidence information to obtain structured evidence information; obtaining the evidence requirement matrix matched with the policy pending claim and the evidence quality index matched by each information processing agent; performing a quality assessment on the structured evidence information based on the evidence requirement matrix and the evidence quality index to obtain an assessment result; if the assessment result indicates that the structured evidence information does not meet the evidence requirement matrix and the evidence quality index, outputting a prompt message carrying the assessment result to a second preset client; if the assessment result indicates that the structured evidence information meets the evidence requirement matrix and the evidence quality index, using the structured evidence information as multimodal and multidimensional evidence information associated with the policy pending claim.

[0022] Optionally, regional residency or encryption methods can be used to de-identify the multimodal and multi-dimensional evidence information associated with the policy to be claimed, resulting in de-identified evidence information. In the embodiments of this application, the specific implementation method used for de-identification is not limited.

[0023] The evidence requirement matrix includes, but is not limited to, a list of mandatory or optional evidence by category, channel, or region, such as a list of product images or evidence information fields; the evidence quality indicators include, but are not limited to, image resolution, image clarity, and logistics trajectory timestamps.

[0024] In practice, by comparing structured evidence information with corresponding lists of mandatory and optional evidence, the completeness of the evidence information can be assessed. This allows for verification of whether the evidence information covers the mandatory evidence list. If the evidence information does not cover the mandatory evidence list, the missing evidence is identified, prompting the user to supplement the missing evidence. By judging whether the evidence information in each field of the structured evidence information meets the corresponding evidence quality indicators, the quality of the evidence information can be assessed (such as assessing the clarity of images or the validity of logistics tracing). Evidence information that does not meet the quality requirements is identified, prompting the user to resubmit the relevant evidence.

[0025] Step 104: Multiple information processing agents are used to perform multi-dimensional and multi-modal information processing operations on the evidence information to obtain the processing results generated by each of the information processing agents.

[0026] In the specific implementation process, multiple information processing agents are set up in advance according to the types of evidence required for policy claims and the content that needs to be reviewed. Each agent is used to process a specified type of evidence or to process a specified content that needs to be reviewed.

[0027] Optionally, the information processing intelligent agent includes one or more of the following: image information processing intelligent agent, logistics information processing intelligent agent, order information processing intelligent agent, chat history processing intelligent agent, and exception handling intelligent agent.

[0028] Optionally, multiple information processing agents are employed to perform multi-dimensional and multi-modal information processing operations on the evidence information, obtaining processing results generated by each of the information processing agents. These operations include one or more of the following: using an image information processing agent to extract information and pre-verify the content of image-type evidence in the evidence information, obtaining the product elements and image evidence pre-verification results associated with the policy pending claim, which serve as the processing results generated by the image information processing agent; using a logistics information processing agent to extract information and pre-verify the content of logistics information-type evidence in the evidence information, obtaining the logistics trajectory elements and logistics information pre-verification results associated with the policy pending claim, which serve as the processing results generated by the image information processing agent; and using a logistics information processing agent to extract information and pre-verify the content of logistics information-type evidence in the evidence information, obtaining the logistics trajectory elements and logistics information pre-verification results associated with the policy pending claim, which serve as the processing results generated by the logistics information processing agent. The following steps are described: The logistics information processing agent generates the following processing results: An order information processing agent aligns the evidence information with order elements and policies to obtain the alignment results of the order elements and policy terms associated with the policy pending claim, which is then used as the processing result generated by the order information processing agent; a chat history processing agent extracts information from the chat history in the evidence information to obtain the communication elements associated with the policy pending claim, which is then used as the processing result generated by the chat history processing agent; and an anomaly handling agent performs anomaly identification processing on the evidence information to obtain a pre-audit result of the claim risk indicating whether there is an abnormal claim for the policy pending claim, which is then used as the processing result generated by the anomaly handling agent.

[0029] The specific implementation methods and applications of each information processing intelligent agent are described below.

[0030] I. Intelligent Agent for Image Information Processing The image information processing intelligent agent is used to extract information and pre-examine the content of the image evidence in the evidence information, and obtain the product elements and image evidence pre-examination results associated with the insurance policy to be claimed, which are used as the processing results generated by the image information processing intelligent agent.

[0031] For example, the image information processing agent is used to review whether product images or videos used as evidence meet the preset requirements for product image evidence, and extracts product elements as evidence from the product images or videos. These product elements include, but are not limited to, one or more of the following: product brand, model, name, product serial number, main product image, appearance image, and damaged location image. The pre-review results of the image evidence include, but are not limited to, one or more of the following: authenticity of the image evidence, damage assessment results, and consistency score of the image evidence. The damage assessment results include, but are not limited to, one or more of the following: damage type, damage location, and severity. The damage type includes, but is not limited to, any of the following: scratches, breakage, deformation, missing parts, cracks, and leakage. The severity includes, but is not limited to, damage proportion and damage quantity.

[0032] Optionally, the commodity elements and the damage type are pre-configured according to the specific application scenario.

[0033] Taking the return and exchange scenario on a cross-border e-commerce platform as an example, after entering the return and exchange process, buyers can upload one or more types of visual evidence through the e-commerce platform's client, such as photos of the main body of the product, close-up photos, product videos, product packaging photos, product order photos, and screenshots of the product description page. After obtaining the image evidence uploaded by the user, the client sends the image evidence to the server, which then calls the image information processing intelligent agent to extract information and pre-verify the content of the image evidence.

[0034] Optionally, an image information processing intelligent agent is used to extract information and pre-examine the image evidence in the evidence information to obtain the product elements and image evidence pre-examination results of the insurance policy to be claimed, which are used as the processing results generated by the image information processing intelligent agent, including one or more of the following operations.

[0035] The first operation involves performing anti-tampering detection on the videos and / or images in the video evidence within the evidence information to obtain a first processing result indicating the authenticity of the video evidence.

[0036] Optionally, the authenticity score of the image or video evidence can be obtained by analyzing the EXIF ​​(Exchangeable image file format) metadata (such as time zone and consistency of shooting software), detecting image recompression traces, verifying watermark consistency, etc., and using this authenticity score as part of the first processing result.

[0037] The second operation involves identifying the type of image damage in one or more of the following types of video evidence: main product photos, partial detail photos, and product videos. It also involves identifying the damaged parts and severity to obtain the first processing result indicating the damage assessment.

[0038] Optionally, the location of damage can be identified based on image segmentation technology, and a classification model can be used to determine the type of damage (such as scratches, breakage, deformation, etc.) and the severity of the damage.

[0039] The third operation involves using computer vision recognition technology to identify product elements in the videos and / or images of the image evidence. These product elements include one or more of the following: product brand, model, name, product serial number, and order description. Visual feature extraction processing is performed on the videos and / or images of the image evidence to obtain product visual features. Similarity calculation is performed between the product visual features and the order description to obtain a consistency score between the product image and the order description, which serves as the first processing result indicating the consistency score of the image evidence.

[0040] Optionally, computer vision recognition technology (such as using deep learning models) can be employed to identify product elements such as brand, model, and serial number from one or more types of image evidence uploaded by the user, including product photos, product videos, product packaging photos, product order photos, and screenshots of product description pages. Multimodal classification models or large language models can be used to infer the similarity between the product's visual features and the order description.

[0041] II. Intelligent Agent for Logistics Information Processing The logistics information processing intelligent agent is used to extract information and pre-examine the content of logistics information in the evidence information to obtain the logistics trajectory elements and logistics information pre-examination results associated with the insurance policy to be claimed, which are used as the processing results generated by the logistics information processing intelligent agent.

[0042] Optionally, a logistics information processing intelligent agent is used to extract information and pre-verify the content of logistics information-type evidence in the evidence information to obtain the logistics trajectory elements and logistics information pre-verification results associated with the policy to be claimed. This includes: using a logistics information processing intelligent agent to extract information from the logistics information-type evidence in the evidence information to obtain a sequence of logistics node information; performing unified spatiotemporal mapping processing on the logistics node information to obtain the mapped logistics node information; and based on the sequence of mapped logistics node information, performing logistics trajectory element extraction and logistics information pre-verification processing to obtain the logistics trajectory elements and logistics information pre-verification results corresponding to the logistics information-type evidence.

[0043] Optionally, the logistics node information includes: logistics node address and arrival time, departure time. The logistics node information is then subjected to unified spatiotemporal mapping processing to obtain the mapped logistics node information, including: mapping the logistics node addresses in the sequence to a unified geographic information system to obtain mapped logistics node addresses; mapping the arrival time and departure time in the sequence to a unified timeline to obtain mapped arrival and departure times; and obtaining the mapped logistics node information based on the mapped logistics node addresses, arrival times, and departure times.

[0044] The logistics trajectory elements include, but are not limited to, one or more of the following: a list of logistics node information and the time spent at each stage; the pre-audit results of the logistics information include, but are not limited to, one or more of the following: timeliness compliance assessment indicators, trajectory integrity score, trajectory rationality score, and list of abnormal events.

[0045] The logistics information evidence includes, but is not limited to, one or more of the following data: waybill number, original logistics information obtained from the carrier's application interface or page, logistics screenshots / PDF files uploaded by the user, shipping address and delivery address, logistics timeliness requirements, region and holidays, etc.

[0046] The logistics information processing agent analyzes the logistics trajectory data of one or more carriers for the goods in the pending insurance claim, extracting sequences of logistics node information, such as transit addresses, key transit addresses, and arrival and departure times for each address. It then performs cross-timezone standardization on the arrival and departure times of the addresses in the sequences, mapping the arrival and departure times of logistics nodes from different carriers to a unified timeline, and converting each address into a geographical location in a unified geographic information system. Subsequently, based on the extracted sequences of logistics node information and the logistics timeliness requirements, regions, and holidays, it performs anomaly detection and trajectory aggregation processing on the logistics trajectory to obtain a pre-approval result for the logistics information. Anomaly detection of the logistics node information sequence includes one or more of the following operations: breakpoint detection (no arrival or departure event within the time limit), reverse flow detection (distance and direction conflict), and abnormal receipt detection (abnormal location / time / recipient). The trajectory aggregation processing includes, but is not limited to, one or more of the following operations: total duration aggregation, mileage estimation, carrier switching list, and coverage.

[0047] In some optional embodiments, time series analysis and node state transition identification methods can be used to analyze the logistics trajectory and extract logistics node information. Based on the logistics node information after cross-time zone standardization, statistical and machine learning methods are used to detect logistics trajectory anomalies (such as failure to update logistics information for a specified period, inconsistencies between logistics node order and arrival time). If anomalies are detected, an anomaly event list is generated based on the detected anomaly events and their severity; if no anomalies are detected, the anomaly event list is cleared. For example, by planning the path between the shipping address and the receiving address, a planned logistics trajectory is obtained. Then, the logistics node information and the planned logistics trajectory are compared to obtain a trajectory integrity score and a trajectory rationality score. As another example, by summarizing the time between logistics nodes, a summary time is obtained, and based on the summary time and logistics timeliness requirements, a timeliness compliance assessment index is derived.

[0048] In some alternative embodiments, a large language model can be used to infer the pre-approval result of logistics information based on the sequence of the mapped logistics node information.

[0049] III. Intelligent Agent for Order Information Processing The order information processing intelligent agent is used to perform order element and policy alignment processing on the evidence information to obtain the alignment result of the order elements and policy terms associated with the policy to be claimed, which is used as the processing result generated by the order information processing intelligent agent.

[0050] Optionally, the order elements include, but are not limited to, one or more of the following information: information of the transacting parties, order summary, product details, price, and quantity; the alignment result is used to indicate whether the order information is subject to policy terms and / or to indicate the policy terms applicable to the order information.

[0051] The types of order elements are determined according to application requirements.

[0052] Optionally, the input data to the order information processing agent includes, but is not limited to: order files (e.g., JSON / XML format order files) and policy terms (e.g., PDF / text format policy terms).

[0053] Optionally, an order information processing intelligent agent is used to perform order element and policy alignment processing on the evidence information to obtain the alignment result of the order elements and policy terms associated with the policy to be claimed, which is used as the processing result generated by the order information processing intelligent agent. This includes: scanning and text recognition of the policy terms to obtain the policy term text; parsing the order file in the evidence information to obtain the order information description text; performing entity recognition on the order information description text to obtain the entity associated with the order product in the order file; performing matching processing on the entity and the policy term text to obtain the matching result; and generating a policy term alignment result based on the matching result. For example, if the matching result indicates that a policy term matching the entity is found, the policy term matching the entity is used as the policy term alignment result; if the matching result indicates that no policy term matching the entity is found, a policy term alignment result indicating that the order elements and policy terms failed to match is generated.

[0054] In some optional embodiments, the order information processing agent uses OCR technology to scan the policy terms document to obtain the policy terms text. Then, multilingual natural language understanding technology is used to perform text understanding on the scanned policy terms text, extracting information such as product type and policy terms. Alternatively, the order document can be parsed according to its format to obtain order information. Furthermore, existing technologies can be used to identify order entities such as product category, name, price, date, and address included in the order information. Then, based on the extracted entity information, it is determined whether the order products are covered by the policy terms. If the order products are covered by the policy terms, the applicable policy terms are used as the policy terms alignment result. Furthermore, the order information processing agent can use existing text summarization models to extract preset order elements based on the scanned text content and the content in the order document.

[0055] IV. Chat Log Processing Intelligent Agent The chat log processing agent is used to extract information from the chat logs in the evidence information to obtain the communication elements associated with the insurance policy pending claim, which serve as the processing result generated by the chat log processing agent. The communication elements include one or more of the following: points of contention, statements of responsibility, and emotional intensity.

[0056] Optionally, the chat history includes, but is not limited to, chat text and screenshots of communication records. When the chat history includes screenshots of communication records, image recognition technology can be used first to scan and recognize the screenshots, converting the image-based chat history into text-based chat history.

[0057] Optionally, natural language understanding technology or existing text summarization algorithms can be used to perform semantic understanding and information extraction on the chat logs in text form, and summarize the core disputed issues to obtain the key points of the dispute and the statement of liability associated with the policy pending claim. Alternatively, a large language model can be used to infer the key points of the dispute, the statement of liability, and the intensity of emotion based on the chat logs, as communication elements associated with the policy pending claim.

[0058] Optionally, the chat history processing agent has the ability to generate multilingual communication elements.

[0059] Optionally, the points of contention include, but are not limited to, one or more of the following: subject matter, demands, and commitment time window.

[0060] Optionally, the statement of liability may include, but is not limited to, any of the following types: promise, denial, explanation, claim.

[0061] Optionally, the chat log processing agent is also used to generate evidence location information (such as message sequence number, page / box position, etc. in the chat log) corresponding to the statement of responsibility.

[0062] V. Exception Handling Intelligent Agent The anomaly handling agent is used to perform fraud and anomaly identification processing on the evidence information to obtain a pre-audit result of the claim risk indicating whether there is an abnormal claim in the policy pending claim.

[0063] For example, the anomaly handling agent determines whether there is abuse of return / exchange authority by classifying the account and order number that triggered the return / exchange operation. Another example is determining whether there is duplicate claiming by searching the claim records for the order number targeted by the return / exchange operation. Yet another example is determining whether there is duplicate claiming by comparing the evidence information with the evidence information in the claim records. Still another example is determining whether there is discrepancy between the goods and the description by matching the evidence information, such as images and / or videos, logistics information, and order information, with the order goods associated with the policy to be claimed. If any of the above situations exist, it can be considered fraudulent behavior and identified as an abnormal claim.

[0064] Each of the aforementioned information processing agents processes the corresponding evidence information and generates its own processing results, which serve as input data for the reasoning agent.

[0065] Step 106: Using a reasoning agent based on preset review rules, the processing results generated by each information processing agent are reasoned and judged to obtain the liability determination result of the policy pending claim.

[0066] As mentioned above, in some optional embodiments, the input of the inference agent includes at least the following information: preset review rules and processing results generated by each of the information processing agents. In other optional embodiments, the input of the inference agent includes at least the following information: preset review rules, processing results generated by each of the information processing agents, and historical precedents. The review rules are dynamically configured according to the application scenario requirements.

[0067] In practice, multimodal and multidimensional evidence information is input into information processing agents for processing different types of evidence, and each agent generates its own processing result. The processing result includes, but is not limited to, one or more of the following data: extracted relevant elements, and pre-audit results output for the corresponding type of evidence. The reasoning agent then performs responsibility determination reasoning on the various elements and pre-audit results generated by each information processing agent according to preset audit rules, obtaining responsibility reasoning results based on different types of data, or simply obtaining a responsibility reasoning result. The audit rules describe the reasoning logic for the audit results, the comprehensive confidence calculation logic, and the dynamic threshold.

[0068] Optionally, an inference agent, based on preset review rules, performs inference and judgment on the processing results generated by each information processing agent to obtain the liability determination result of the policy pending claim. This includes: matching the processing results generated by each information processing agent with preset review rules to obtain the review rules that each processing result matches; performing liability inference on the matched processing results based on the matched review rules to obtain the liability inference result and confidence level corresponding to each review rule; and fusing and judging the liability inference results based on the preset weight and confidence level threshold of the review rule matching, as well as the confidence level, to obtain the liability determination result of the policy pending claim. The liability determination result includes: the liability inference result, the matching rule corresponding to the liability inference result, the review conclusion, and the confidence level of the review conclusion.

[0069] The reasoning agent is used to make a comprehensive judgment on the processing results generated by each information processing agent, combined with dynamically configurable review rules (i.e., application scenario rules).

[0070] Optionally, the processing results generated by each information processing agent can be semantically matched with preset review rules to obtain the review rules that each processing result corresponds to. For example, when configuring review rules, the system sets a rule name or rule tag for each review rule. By performing semantic similarity matching between the field name or result description corresponding to the processing result and the rule name or rule tag, the review rules that each processing result corresponds to can be obtained. For example, the damage assessment result in the processing result of the image information processing agent can be matched with the damage ratio review rule in the review rules.

[0071] The process of performing liability inference on the matched processing results based on the hit review rules to obtain the liability inference result and confidence level corresponding to each review rule includes: executing each hit review rule to perform liability inference on the matched processing results to obtain the liability inference result and confidence level corresponding to each review rule. The confidence level can be the degree of matching between the processing result and the review rule. For example, executing a hit damage ratio review rule yields the liability inference result and confidence level regarding whether the damage ratio generated by the image information processing agent corresponding to that review rule meets the claim conditions. As another example, executing a hit subject image consistency review rule yields the liability inference result and confidence level regarding whether the subject image and the product image included in the generated result of the image information processing agent corresponding to that review rule are consistent.

[0072] Optionally, review rules can be maintained through a rules engine, including but not limited to: adding rules, editing rules, deleting rules, setting the weight of rule matching, and setting the confidence level of rule matching. These review rules can be managed based on parameters such as category, channel, and region. For example, different categories of goods may have different preset review rules, and goods from different channels may have different preset review rules.

[0073] The claims information review method disclosed in this application effectively balances the timeliness and accuracy of claims review by dynamically setting review rules based on product category, channel, region, or claim request traffic time period. For example, during peak claim request traffic periods, the confidence threshold can be lowered, thereby diverting more pending claims to the automated claims process and improving the timeliness of claims review. Furthermore, by setting separate review rules for different channels, the accuracy of claims review results can be improved through refined review rules.

[0074] Optionally, the predefined reasoning logic of the hit audit rules can be executed through the rule engine, thereby executing each hit audit rule to perform responsibility reasoning on the matched processing results, and obtaining the responsibility reasoning result corresponding to each audit rule and the confidence level of the responsibility reasoning result.

[0075] In practice, weights can be dynamically set for each review rule, and a confidence threshold for the liability inference result can be set. Optionally, based on the preset weights and confidence thresholds matched by the review rules, and the confidence level, the liability inference result is fused and judged to obtain the liability determination result for the policy pending claim. This includes: performing a weighted calculation on the confidence level of the corresponding liability inference result based on the preset weights matched by each review rule to obtain a comprehensive confidence level; obtaining the review conclusion for the policy pending claim based on the liability inference result and the comprehensive confidence level; and obtaining the liability determination result for the policy pending claim based on the liability inference result, the review rule corresponding to the liability inference result, and the review conclusion. For example, the weight of the subject image consistency review rule (e.g., 80%) in image evidence can be set higher than the weight of the damage ratio review rule (e.g., 20%).

[0076] Optionally, the rule engine can execute the fusion processing logic of the matched review rules, and based on the preset weights matched by each review rule, perform a weighted summation of the confidence levels of the corresponding liability inference results to obtain a comprehensive confidence level; and, based on the liability inference results and the comprehensive confidence level, obtain the review conclusion of the policy pending claim. The review conclusion includes, but is not limited to, any of the following: review approved, claim denied, supplementary evidence required, or manual review.

[0077] The review conclusion for the pending insurance policy, based on the liability inference result and the overall confidence level, includes: obtaining a review conclusion indicating approval when the overall confidence level is greater than or equal to the confidence level threshold, the evidence chain data coverage in the liability inference result meets the preset integrity condition, and the liability inference result matches low risk; obtaining a review conclusion indicating rejection of compensation when the overall confidence level is greater than or equal to the confidence level threshold, the evidence chain data coverage in the liability inference result meets the preset integrity condition, and the liability inference result matches high risk; obtaining a review conclusion indicating supplementary evidence when the overall confidence level is greater than or equal to the confidence level threshold, the evidence chain data coverage in the liability inference result does not meet the preset integrity condition; and obtaining a review conclusion indicating manual review when the overall confidence level is less than the confidence level threshold, or when the liability inference result matches high risk.

[0078] For example, by executing the fusion processing logic of the hit conclusion reasoning rules through a preset rule engine, a comprehensive judgment is made on the overall confidence level, the completeness of the evidence chain data, the authenticity of the image evidence generated by the image information processing agent, the list of abnormal events generated by the logistics information processing agent, and the pre-audit result of the claim risk generated by the abnormality processing agent, to obtain the audit conclusion of the policy to be claimed. For example, if the overall confidence level is greater than or equal to the confidence level threshold and the image of the damaged location is missing, an audit conclusion instructing supplementary evidence is obtained. As another example, if an abnormal event exists in the list of abnormal events or duplicate claims are detected, an audit conclusion instructing manual review is obtained.

[0079] In some alternative embodiments, the review conclusion of the policy pending claim can also be obtained by using a large language model, based on the liability reasoning result and the overall confidence level.

[0080] Furthermore, the liability reasoning results corresponding to each audit rule are used as liability interpretations, and the corresponding audit rules are used as hit rules. The audit conclusions are then integrated to obtain the liability determination results for the policy pending claim.

[0081] Optionally, the audit rules matched in the liability inference result also include version information, used for subsequent traceability and auditing of liability determination results. By persistently storing the version information of the matched audit rules together with the liability inference result, it is possible to retrospectively review the audit results of claims policies across time dimensions. Even if the current audit rules have changed, the judgment logic of the liability inference result can still be restored.

[0082] Reference Figure 3 Optionally, after the reasoning agent uses a preset review rule to reason and judge the processing results generated by each information processing agent to obtain the liability determination result of the policy to be claimed, the method further includes: step 108.

[0083] Step 108: The decision orchestration agent is used to perform supplementary judgment processing on the liability determination result to obtain the review report of the policy pending claim.

[0084] Optionally, the liability determination result includes: liability interpretation. The supplementary judgment processing of the liability determination result using a decision orchestration intelligent agent to obtain the audit report of the policy pending claim includes: performing consistency verification on the liability interpretation to obtain a verification result; performing risk control management on the liability interpretation based on preset supplementary judgment conditions to obtain a risk control management result; and generating the audit report of the policy pending claim based on the liability determination result, the verification result, and the risk control management result.

[0085] The process involves a consistency check on the interpretation of responsibility, yielding a check result. This includes checking the consistency of core elements such as product elements generated by the image information processing agent, order elements generated by the order information processing agent, and logistics trajectory elements generated by the logistics information processing agent. The check identifies any conflicts between these core elements. If a conflict exists, a check result indicating inconsistency in the interpretation of responsibility is obtained; otherwise, a check result indicating inconsistency is obtained. By performing consistency checks on the interpretations of responsibility generated based on the results from different information processing agents, not only can conflicts in the evidence chain uploaded by the user be checked, but the adverse effects of illusions generated during the execution of various information processing agents and rule engines on the review results can also be avoided.

[0086] In the process of performing risk control management on the interpretation of responsibility based on preset supplementary judgment conditions and obtaining the risk control management results, external risk control capabilities can be introduced, such as calling external interfaces to obtain a blacklist of claims accounts as supplementary judgment conditions, so as to enhance the reliability of the audit system.

[0087] The claims information review method disclosed in this application identifies fraudulent behavior in the information processing intelligent agent and identifies blacklists in the decision-making and orchestration intelligent agent. The two intelligent agents work together to manage risk control, which can prevent reviewers from using sensitive information (such as account information and claims records used for fraudulent behavior identification) without authorization.

[0088] Optionally, the liability determination result further includes: an audit conclusion. The step of generating an audit report for the policy pending claim based on the liability determination result, the verification result, and the risk control management result includes: determining the risk level of the liability determination result based on the verification result and the risk control management result; updating the audit conclusion to indicate manual review if the verification result indicates an inconsistency in the interpretation of liability, or if the risk control management result indicates the existence of risk control risk, and generating an audit report for the policy pending claim based on the updated liability determination result; and generating an audit report for the policy pending claim based on the liability determination result if the verification result indicates a consistent interpretation of liability and the risk control management result indicates the absence of risk control risk.

[0089] The generated audit report includes: an explanation of responsibility, the audit rules that were met, and the audit conclusion. In the embodiments of this application, by presenting the audit conclusion and explanation of responsibility based on the met audit rules in the audit report, a complete chain of evidence is demonstrated, making the audit conclusion transparent and credible.

[0090] Optionally, after using a decision orchestration agent to perform supplementary judgment processing on the liability determination result to obtain the review report of the policy pending claim, the process further includes: performing unified timeline mapping processing on the evidence information and the processing result, mapping the evidence information and the processing result to a unified timeline; associating the liability determination result with the corresponding evidence information on the unified timeline to obtain an evidence chain graph; marking the processing result and / or the liability determination result that indicate anomalies in the evidence chain graph, so that the first preset client highlights the marked processing result and / or liability determination result when displaying the evidence chain spatiotemporal graph.

[0091] For example, first, a standard timeline is established. Then, the upload time and / or generation time of the evidence information are converted into times on the timeline. Based on the converted times, nodes for the evidence information are created on the timeline, thus mapping the evidence information and the processing results to the timeline. For example, a node corresponding to the uploaded image evidence is created at the timeline position. Another example is a node for the policy taking effect at the timeline position corresponding to the shipping time. Yet another example is creating nodes corresponding to the arrival and departure times of each logistics node in the processing results at the corresponding time points on the timeline. Furthermore, the liability determination result corresponding to the evidence information can be used as attribute information of the nodes of the evidence information on the timeline, associating the liability determination result with the corresponding evidence information on the unified timeline, thereby obtaining an evidence chain graph.

[0092] The server then associates and stores the audit results with the evidence chain graph.

[0093] The processing results indicating anomalies may include, for example, identifying tampered image evidence, indicating duplicate claims, or indicating incomplete logistics tracking. The liability determination result indicating anomalies may be a liability determination result with a confidence level lower than a preset confidence threshold.

[0094] By establishing a spatiotemporal graph of the evidence chain and marking the processing results and / or responsibility determination results of the abnormal audits, and then when the spatiotemporal graph of the evidence chain is displayed on the first preset client, the processing results and / or responsibility determination results of the abnormal audits are highlighted (e.g., when the main image of the product is found to be inconsistent with the product, the node corresponding to the upload time of the image-type evidence in the evidence information, as well as the responsibility determination result and / or processing result as the node attribute are highlighted), which makes it easier for relevant personnel to quickly locate the evidence information of the abnormal audits (such as inconsistent evidence, fraud evidence, etc.), and helps to improve the efficiency of manual review.

[0095] Reference Figure 3 The claims information review method disclosed in this application, after the decision orchestration intelligent agent performs supplementary judgment processing on the liability determination result to obtain the review report of the policy to be claimed, further includes: step 110.

[0096] Step 110: Perform hierarchical routing processing on the pending claims policy based on the audit report.

[0097] Optionally, the liability determination result in the audit report may also include: an audit conclusion. The step of performing hierarchical routing processing on the pending claim policy based on the audit report includes: sending the associated information of the audit report to a first preset client when the audit conclusion indicates manual review; outputting supplementary evidence prompt information to a second preset client when the audit conclusion indicates supplementary evidence is required; and performing automatic claims processing on the pending claim policy based on the audit report when the audit conclusion indicates approval or rejection of compensation.

[0098] The first preset client is a client for a claims information review system, such as the client logged in by a manual reviewer. The associated information of the review report includes, but is not limited to, any one or more of the following: review report number, name, claims policy information, and review report summary. Reviewers can view the review report or its corresponding evidence chain diagram based on the associated information.

[0099] The second preset client can be a client that provides evidence information for the pending insurance policy submitted by the user, such as the client of an e-commerce platform. The supplementary evidence prompts include, but are not limited to, audit reports. Users can view the audit reports and obtain the necessary supplementary evidence based on the prompts. After supplementing evidence through the second preset client, users can submit a claim application again, triggering the audit process for the pending insurance policy.

[0100] If the review conclusion indicates that the review is approved, the server automatically executes the claims process for the policy pending claim. If the review conclusion indicates that the claim is rejected, the server automatically executes the claim rejection process for the policy pending claim. Specific implementation methods for the server automatically executing the claims process and claim rejection process for the policy pending claim are found in existing technologies and will not be repeated in this application embodiment.

[0101] Optional, refer to Figure 3 After the decision orchestration agent performs supplementary judgment processing on the liability determination result to obtain the review report of the policy pending claim, the method further includes: step 112.

[0102] Step 112: The evidence information, the processing result, the liability determination result, and the audit report are stored together.

[0103] For example, the evidence information, the processing results, the liability determination results, and the audit report can be stored in a preset database to ensure that the entire claims process of the insurance policy is traceable, providing data support for subsequent random checks and replication of the claims process.

[0104] In summary, the claims information review method disclosed in this application involves obtaining multimodal and multi-dimensional evidence information associated with the policy to be claimed from the server; then, multiple information processing agents perform multi-dimensional and multimodal information processing operations on the evidence information to obtain processing results generated by each information processing agent; finally, a reasoning agent, based on preset review rules, infers and judges the processing results generated by each information processing agent to obtain the liability determination result of the policy to be claimed. This method effectively improves the review efficiency of policies to be claimed by automatically generating review results and using multiple information processing agents to process evidence information of different modalities and dimensions; furthermore, by using information processing agents with specific functions in conjunction with preset review rules to review the evidence information and supplementing the judgment with reinforcement rules, the accuracy of the review of policies to be claimed is effectively improved.

[0105] like Figure 4 As shown, based on the above embodiments, this embodiment also provides a method for reviewing claims information applied to a client, including steps 402 to 406.

[0106] Step 402: In response to the manual review of the audit report, obtain the data of the evidence chain diagram corresponding to the audit report.

[0107] Optionally, policy reviewers can log in to the policy review system's client (i.e., the first preset client mentioned above) to view policy review reports requiring manual review. In practice, after receiving the associated information of the review report sent by the server (such as the report number and summary information), the client displays the associated information and detects the user's manual review operation on the report. Upon detecting that the user has triggered a manual review of the report, the client can further retrieve the evidence chain graph data of the review report from the server.

[0108] Step 404: Based on the data displayed, the evidence chain diagram is presented.

[0109] Optionally, the evidence chain graph includes several nodes arranged chronologically. Each node is associated with evidence information or processing results of the insurance claim policy. The evidence chain graph also includes attribute information for each node, including the processing results and / or liability determination results of the evidence information mapped to the current node.

[0110] The evidence chain graph is established by the pre-defined server using the following method: performing unified timeline mapping on the evidence information and processing results of the audit report, mapping the evidence information and processing results to a unified timeline; associating the responsibility determination result with the corresponding evidence information on the unified timeline to obtain the evidence chain graph; and marking the processing results and / or responsibility determination results in the evidence chain graph that indicate an audit anomaly.

[0111] The method for establishing the evidence chain map is described in the previous embodiments and will not be repeated here.

[0112] Optionally, a timeline can be displayed, and then, according to the time points corresponding to the nodes in the evidence chain diagram, the nodes corresponding to the evidence information and the attribute information of the nodes can be displayed on the timeline, such as the evidence information associated with the node, the processing result, the responsibility determination result, etc.

[0113] Step 406: Highlight the marked processing results and / or liability determination results in the displayed evidence chain diagram.

[0114] For example, the pre-marked processing results and / or liability determination results in the evidence chain graph displayed on the client can be highlighted.

[0115] In summary, the claims information review method disclosed in this application establishes an evidence chain graph on the server side and marks the processing results and / or liability determination results that indicate anomalies in the review within the evidence chain graph. This allows the client to obtain the data of the evidence chain graph corresponding to the review report after detecting the manual review operation. When the evidence chain graph is displayed based on the data, the marked processing results and / or liability determination results are highlighted in the displayed evidence chain graph. This facilitates the review personnel in quickly locating the evidence information that indicates anomalies and helps improve the efficiency of manual review.

[0116] Based on the above embodiments, this embodiment also provides a claims information verification system, which includes a client and a server. The following describes the system in conjunction with... Figure 5 The interactive process shown illustrates the specific implementation of the claims information review system.

[0117] Step 502: The server obtains multimodal and multidimensional evidence information associated with the insurance policy to be claimed.

[0118] Step 504: The server uses multiple information processing agents to perform multi-dimensional and multi-modal information processing operations on the evidence information to obtain the processing results generated by each information processing agent.

[0119] Step 506: The server uses an inference agent based on preset review rules to infer and judge the processing results generated by each information processing agent to obtain the liability determination result of the policy pending claim.

[0120] Step 508: The server uses a decision orchestration agent to perform supplementary judgment processing on the liability determination result, and obtains the review report of the policy pending claim.

[0121] Step 510: The server performs unified timeline mapping on the evidence information and processing results of the audit report, mapping the evidence information and processing results to a unified timeline.

[0122] Step 512: The server associates the responsibility determination result in the audit report with the corresponding evidence information on the unified timeline to obtain an evidence chain graph.

[0123] Step 514: The server marks the processing results and / or responsibility determination results that indicate anomalies in the evidence chain graph.

[0124] Step 516: When the audit conclusion indicates that manual review is required, the server outputs the associated information of the audit report to the client.

[0125] Step 518: The client displays the associated information of the audit report.

[0126] Step 520: In response to the manual review of the audit report, the client obtains the data of the evidence chain graph.

[0127] Step 522: The client displays the evidence chain graph based on the data displayed, and highlights the marked processing results and / or liability determination results in the displayed evidence chain graph.

[0128] For the specific implementation methods of each step executed by the server and client, please refer to the relevant descriptions in the previous embodiments, which will not be repeated here.

[0129] In summary, the claims information review system disclosed in this application improves the review efficiency of policies pending claims by setting up multiple information processing agents on the server side to process evidence information of different modalities and dimensions, which can be executed in parallel. An inference agent, based on preset review rules, performs reasoning and judgment on the processing results generated by each information processing agent to obtain the liability determination result of the policy pending claims. Then, a decision orchestration agent performs supplementary judgment processing on the liability determination result to obtain the review report of the policy pending claims. Combined with reinforcement rules for supplementary judgment, this effectively improves the review accuracy of policies pending claims. Furthermore, this method effectively improves the review efficiency of policies pending claims by automatically generating review reports. Furthermore, for audit reports requiring manual review, an evidence chain diagram corresponding to the audit report is established, and the processing results and / or responsibility determination results indicating abnormalities in the evidence chain diagram are marked. When the client displays the evidence chain diagram based on the data, the marked processing results and / or responsibility determination results are highlighted in the displayed evidence chain diagram, which facilitates the reviewers to quickly locate the evidence information indicating abnormalities and helps improve the efficiency of manual review.

[0130] Based on the above embodiments, this embodiment also provides a claims information verification device, applied to a server, the device comprising: The evidence information acquisition module is used to acquire multimodal and multi-dimensional evidence information related to the insurance policy to be claimed; The information processing module is used to employ multiple information processing agents to perform multi-dimensional and multi-modal information processing operations on the evidence information, and obtain the processing results generated by each of the information processing agents. The rule review module is used to use a reasoning agent to reason and judge the processing results generated by each information processing agent based on preset review rules, so as to obtain the liability determination result of the insurance policy to be claimed.

[0131] Optionally, the step of employing multiple information processing agents to perform multi-dimensional and multi-modal information processing operations on the evidence information, and obtaining processing results generated by each of the information processing agents, includes one or more of the following operations: An image information processing intelligent agent is used to extract information and pre-examine the image evidence in the evidence information to obtain the product elements and image evidence pre-examinement results associated with the insurance policy to be claimed, which are used as the processing results generated by the image information processing intelligent agent. A logistics information processing intelligent agent is used to extract information and pre-examine the content of logistics information in the evidence information to obtain the logistics trajectory elements and logistics information pre-examination results associated with the insurance policy to be claimed, which are used as the processing results generated by the logistics information processing intelligent agent. An order information processing intelligent agent is used to align the evidence information with order elements and the policy to obtain the alignment result of the order elements and policy terms associated with the policy to be claimed, which is used as the processing result generated by the order information processing intelligent agent. A chat history processing AI extracts information from the chat history in the evidence information to obtain the communication elements associated with the insurance policy pending claim, which are then used as the processing result generated by the chat history processing AI. An anomaly processing agent is used to perform anomaly identification processing on the evidence information to obtain a pre-audit result of the claim risk indicating whether there is an abnormal claim in the policy pending claim, which is used as the processing result generated by the anomaly processing agent.

[0132] Optionally, the step of using a logistics information processing intelligent agent to extract information and pre-verify the content of logistics information-type evidence in the evidence information to obtain the logistics trajectory elements and logistics information pre-verification results associated with the policy to be claimed includes: A logistics information processing intelligent agent is used to extract information from the logistics information-type evidence in the evidence information to obtain a sequence of logistics node information. The logistics node information is subjected to a unified spatiotemporal mapping process to obtain the mapped logistics node information. Based on the sequence of logistics node information after mapping processing, logistics trajectory elements are extracted and logistics information is pre-verified to obtain the logistics trajectory elements and logistics information pre-verification results corresponding to the logistics information evidence.

[0133] Optionally, the step of employing an inference agent to infer and judge the processing results generated by each information processing agent based on preset review rules, and obtaining the liability determination result of the policy pending claim, includes: The reasoning agent matches the processing results generated by each of the information processing agents with preset review rules to obtain the review rules that each processing result matches. Based on the matched review rules, responsibility inference is performed on the matching processing results to obtain the responsibility inference results and confidence levels corresponding to each review rule; Based on the preset weights and confidence thresholds of the audit rules, and the confidence level, the liability inference results are fused and judged to obtain the liability determination results of the policy pending claim.

[0134] Optionally, the device further includes: The decision orchestration module is used to supplement the liability determination result with a decision orchestration intelligent agent to obtain the review report of the insurance policy pending claim.

[0135] Optionally, the liability determination result includes: liability interpretation; the supplementary judgment processing of the liability determination result using a decision orchestration intelligent agent to obtain the review report of the policy pending claim includes: The consistency of the interpretation of responsibility is verified, and the verification result is obtained. Risk control management is performed on the interpretation of responsibility based on preset supplementary judgment conditions to obtain risk control management results; Based on the liability determination result, the verification result, and the risk control management result, an audit report for the pending claim policy is generated.

[0136] Optionally, the liability determination result further includes: an audit conclusion. The step of generating an audit report for the pending claim policy based on the liability determination result, the verification result, and the risk control management result includes: Based on the verification results and the risk control management results, the risk level of the liability determination result is determined; If the verification result indicates that the interpretation of liability is inconsistent, or if the risk control management result indicates that there is a risk control risk, the review conclusion is updated to indicate manual review, and an review report for the policy pending claim is generated based on the updated liability determination result. If the verification result indicates that the interpretation of liability is consistent and the risk control management result indicates that there is no risk control risk, an audit report for the policy pending claim is generated based on the liability determination result.

[0137] Optionally, the liability determination result further includes: an audit conclusion. After the decision orchestration agent performs supplementary judgment processing on the liability determination result to obtain the audit report of the policy pending claim, the device further includes: The routing module is used to output the associated information of the audit report to a first preset client when the audit conclusion indicates that manual review is required; The routing module is further configured to output a supplementary evidence prompt message to a second preset client when the audit conclusion indicates that supplementary evidence is required; and, If the audit conclusion indicates that the audit is approved or the claim is rejected, an automatic claim processing operation will be performed on the policy pending claim based on the audit report.

[0138] Optionally, after the decision orchestration agent performs supplementary judgment processing on the liability determination result to obtain the review report of the policy pending claim, the device further includes: The storage module is used to associate and store the evidence information, the processing results, the liability determination results, and the audit report.

[0139] Optionally, after the decision orchestration agent performs supplementary judgment processing on the liability determination result to obtain the review report of the policy pending claim, the device further includes: The evidence chain graph establishment module is used to perform unified timeline mapping processing on the evidence information and the processing results, mapping the evidence information and the processing results to a unified timeline; and to associate the responsibility determination result with the corresponding evidence information on the unified timeline to obtain the evidence chain graph. The evidence chain graph establishment module is also used to mark the processing results and / or responsibility determination results that indicate anomalies in the evidence chain graph, so that the first preset client highlights the marked processing results and / or responsibility determination results when displaying the evidence chain spatiotemporal graph.

[0140] The claims information verification device disclosed in this application is used to implement the above-mentioned claims information verification method. For the specific implementation of each module of the device, please refer to the specific implementation of the corresponding steps in the foregoing method embodiments, which will not be repeated here.

[0141] In summary, the claims information verification device disclosed in this application obtains multimodal and multi-dimensional evidence information associated with the policy to be claimed from the server; then, multiple information processing agents perform multi-dimensional and multimodal information processing operations on the evidence information to obtain processing results generated by each information processing agent; and finally, an inference agent, based on preset verification rules, infers and judges the processing results generated by each information processing agent to obtain the liability determination result of the policy to be claimed. This device effectively improves the verification efficiency of policies to be claimed by automatically generating verification results and using multiple information processing agents to process evidence information of different modalities and dimensions; and by using information processing agents with specific functions in conjunction with preset verification rules to verify the evidence information and supplementing the judgment with reinforcement rules, it effectively improves the verification accuracy of policies to be claimed.

[0142] Based on the above embodiments, this embodiment also provides a claims information verification device, applied to a client, the device comprising: The evidence chain graph data acquisition module is used to acquire the evidence chain graph data corresponding to the audit report in response to the manual review of the audit report. An evidence chain graph display module is used to display the evidence chain graph based on the data displayed. The evidence chain graph display module is further configured to highlight the marked processing results and / or responsibility determination results in the displayed evidence chain graph; wherein the evidence chain graph is established by a pre-defined server using the following method: performing unified timeline mapping processing on the evidence information and processing results of the audit report, mapping the evidence information and processing results to a unified timeline; associating the responsibility determination results with the corresponding evidence information on the unified timeline to obtain the evidence chain graph; and marking the processing results and / or responsibility determination results in the evidence chain graph that indicate an audit anomaly.

[0143] The claims information verification device disclosed in this application is used to implement the above-mentioned claims information verification method. For the specific implementation of each module of the device, please refer to the specific implementation of the corresponding steps in the foregoing method embodiments, which will not be repeated here.

[0144] In summary, the claims information review device disclosed in this application establishes an evidence chain graph on the server side and marks the processing results and / or liability determination results that indicate anomalies in the review within the evidence chain graph. This allows the client to obtain the data of the evidence chain graph corresponding to the review report after detecting the manual review operation. When displaying the evidence chain graph, the client highlights the marked processing results and / or liability determination results in the displayed evidence chain graph, making it easier for reviewers to quickly locate the evidence information that indicates anomalies and improving the efficiency of manual review.

[0145] This application also provides a non-volatile readable storage medium storing one or more modules (programs). When these modules are applied to a device, they enable the device to execute the instructions for the method steps in this application.

[0146] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the methods described in this application.

[0147] This application also provides an electronic device, including: a processor and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method described in this application embodiment. In this application embodiment, the electronic device includes devices such as servers and terminal devices.

[0148] This application also discloses a computer program product, including a computer program / computer executable instructions, which, when executed by a processor in an electronic device, implement the method described in this application.

[0149] Embodiments of this disclosure can be implemented as an apparatus with any suitable hardware, firmware, software, or any combination thereof, configured as desired, and the apparatus may include electronic devices such as servers (clusters) and terminals. Figure 6 An exemplary apparatus 600 is schematically shown that can be used to implement the various embodiments described in this application.

[0150] In one embodiment, Figure 6 An exemplary device 600 is shown, which includes one or more processors 602, a control module (chipset) 604 coupled to at least one of the processors 602, a memory 606 coupled to the control module 604, a non-volatile memory (NVM) / storage device 608 coupled to the control module 604, one or more input / output devices 610 coupled to the control module 604, and a network interface 612 coupled to the control module 604.

[0151] Processor 602 may include one or more single-core or multi-core processors, and processor 602 may include any combination of general-purpose processors or special-purpose processors (e.g., graphics processors, application processors, baseband processors, etc.). In some embodiments, device 600 can serve as a server, terminal, or other device as described in the embodiments of this application.

[0152] In some embodiments, the apparatus 600 may include one or more computer-readable media (e.g., memory 606 or NVM / storage device 608) having instructions 614 and one or more processors 602 that are combined with the one or more computer-readable media and configured to execute the instructions 614 to implement the module and thus perform the actions described in this disclosure.

[0153] In one embodiment, the control module 604 may include any suitable interface controller to provide any suitable interface to at least one of the processors 602 and / or any suitable device or component communicating with the control module 604.

[0154] The control module 604 may include a memory controller module to provide an interface to the memory 606. The memory controller module may be a hardware module, a software module, and / or a firmware module.

[0155] Memory 606 may be used, for example, to load and store data and / or instructions 614 for device 600. In one embodiment, memory 606 may include any suitable volatile memory, such as suitable DRAM. In some embodiments, memory 606 may include double data rate type quad synchronous dynamic random access memory (DDR4 SDRAM).

[0156] In one embodiment, the control module 604 may include one or more input / output controllers to provide an interface to the NVM / storage device 608 and (one or more) input / output devices 610.

[0157] For example, NVM / storage device 608 may be used to store data and / or instructions 614. NVM / storage device 608 may include any suitable non-volatile memory (e.g., flash memory) and / or may include any suitable (one or more) non-volatile storage devices (e.g., one or more hard disk drives (HDDs), one or more optical disc drives (CDs), and / or one or more digital universal optical disc (DVD) drives).

[0158] NVM / storage device 608 may include storage resources that are part of a device on which device 600 is mounted, or that are accessible to the device but do not necessarily have to be part of the device. For example, NVM / storage device 608 may be accessed via a network through one or more input / output devices 610.

[0159] One or more input / output devices 610 may provide an interface for device 600 to communicate with any other suitable device. Input / output devices 610 may include communication components, audio components, sensor components, etc. A network interface 612 may provide an interface for device 600 to communicate via one or more networks. Device 600 may wirelessly communicate with one or more components of a wireless network according to any of one or more wireless network standards and / or protocols, such as accessing wireless networks based on communication standards, such as Bluetooth, WiFi, 2G, 3G, 4G, 5G, etc., or combinations thereof.

[0160] In one embodiment, at least one of the processors 602 may be logically packaged with one or more controllers (e.g., memory controller modules) of the control module 604. In one embodiment, at least one of the processors 602 may be logically packaged with one or more controllers of the control module 604 to form a system-in-package (SiP). In one embodiment, at least one of the processors 602 may be integrated with the logic of one or more controllers of the control module 604 on the same die. In one embodiment, at least one of the processors 602 may be integrated with the logic of one or more controllers of the control module 604 on the same die to form a system-on-a-chip (SoC).

[0161] In various embodiments, device 600 may be, but is not limited to, a component, integrated circuit, or chip in a terminal. The device may be a mobile electronic device or a non-mobile electronic device. For example, mobile electronic devices may be mobile phones, tablets, laptops, PDAs, in-vehicle electronic devices, wearable devices, ultra-mobile personal computers (UMPCs), netbooks, or personal digital assistants (PDAs), etc., while non-mobile electronic devices may be servers, network-attached storage (NAS), personal computers (PCs), televisions (TVs), ATMs, or self-service machines, etc. The embodiments of this application do not impose specific limitations.

[0162] In various embodiments, device 600 may have more or fewer components and / or different architectures. For example, in some embodiments, device 600 includes one or more cameras, a keyboard, a liquid crystal display (LCD) screen (including a touch screen display), a non-volatile memory port, multiple antennas, a graphics chip, an application-specific integrated circuit (ASIC), and a speaker.

[0163] The device can use a main control chip as a processor or control module, and sensor data, location information, etc. can be stored in a memory or NVM / storage device. The sensor group can be used as an input / output device, and the communication interface can include a network interface.

[0164] This application also provides an electronic device, including: a processor; and a memory storing executable code thereon. When the executable code is executed, the processor performs one or more methods as described in this application embodiment. In this application embodiment, the memory can store various types of data, such as target files, file-application association data, and user behavior data, thereby providing a data foundation for various processing operations.

[0165] This application also provides one or more machine-readable media having executable code stored thereon, which, when executed, causes a processor to perform one or more of the methods described in this application.

[0166] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0167] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0168] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0169] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0170] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0171] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0172] Finally, it should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one…" does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element. In the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following related objects are in an "or" relationship.

[0173] The foregoing has provided a detailed description of a claims information verification method, a claims information verification system, an electronic device, a storage medium, and a computer program product provided by this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for verifying claims information, applied on the server side, characterized in that, The method includes: Obtain multimodal and multidimensional evidentiary information related to the insurance policy pending claim; Multiple information processing agents are employed to perform multi-dimensional and multi-modal information processing operations on the evidence information, resulting in processing results generated by each of the information processing agents. An inference agent, based on preset review rules, infers and judges the processing results generated by each information processing agent to obtain the liability determination result of the insurance policy pending claim.

2. The method according to claim 1, characterized in that, The method employs multiple information processing agents to perform multi-dimensional and multi-modal information processing operations on the evidence information, obtaining processing results generated by each of the information processing agents, including one or more of the following operations: An image information processing intelligent agent is used to extract information and pre-examine the image evidence in the evidence information to obtain the product elements and image evidence pre-examinement results associated with the insurance policy to be claimed, which are used as the processing results generated by the image information processing intelligent agent. A logistics information processing intelligent agent is used to extract information and pre-examine the content of logistics information in the evidence information to obtain the logistics trajectory elements and logistics information pre-examination results associated with the insurance policy to be claimed, which are used as the processing results generated by the logistics information processing intelligent agent. An order information processing intelligent agent is used to align the evidence information with order elements and the policy to obtain the alignment result of the order elements and policy terms associated with the policy to be claimed, which is used as the processing result generated by the order information processing intelligent agent. A chat history processing AI extracts information from the chat history in the evidence information to obtain the communication elements associated with the insurance policy pending claim, which are then used as the processing result generated by the chat history processing AI. An anomaly processing agent is used to perform anomaly identification processing on the evidence information to obtain a pre-audit result of the claim risk indicating whether there is an abnormal claim in the policy pending claim, which is used as the processing result generated by the anomaly processing agent.

3. The method according to claim 2, characterized in that, The process employs a logistics information processing intelligent agent to extract information and pre-verify the content of logistics information-type evidence from the evidence information, thereby obtaining the logistics trajectory elements and pre-verification results of the logistics information associated with the policy pending claim, including: A logistics information processing intelligent agent is used to extract information from the logistics information-type evidence in the evidence information to obtain a sequence of logistics node information. The logistics node information is subjected to a unified spatiotemporal mapping process to obtain the mapped logistics node information. Based on the sequence of logistics node information after mapping processing, logistics trajectory elements are extracted and logistics information is pre-verified to obtain the logistics trajectory elements and logistics information pre-verification results corresponding to the logistics information evidence.

4. The method according to claim 1, characterized in that, The method employs an inference agent based on preset review rules to infer and judge the processing results generated by each information processing agent, thereby obtaining the liability determination result of the policy pending claim, including: The reasoning agent matches the processing results generated by each of the information processing agents with preset review rules to obtain the review rules that each processing result matches. Based on the matched review rules, responsibility inference is performed on the matching processing results to obtain the responsibility inference results and confidence levels corresponding to each review rule; Based on the preset weights and confidence thresholds of the audit rules, and the confidence level, the liability inference results are fused and judged to obtain the liability determination results of the policy pending claim.

5. The method according to claim 1, characterized in that, After the step of employing a reasoning agent to infer and judge the processing results generated by each information processing agent based on preset review rules to obtain the liability determination result of the policy pending claim, the method further includes: A decision orchestration agent is used to supplement the liability determination results to obtain an audit report for the policy pending claim.

6. The method according to claim 5, characterized in that, The liability determination result includes: liability explanation. The supplementary judgment processing of the liability determination result using a decision orchestration intelligent agent to obtain the review report of the policy pending claim includes: The consistency of the interpretation of responsibility is verified, and the verification result is obtained. Risk control management is performed on the interpretation of responsibility based on preset supplementary judgment conditions to obtain risk control management results; Based on the liability determination result, the verification result, and the risk control management result, an audit report for the pending claim policy is generated.

7. The method according to claim 6, characterized in that, The liability determination result also includes: an audit conclusion. The audit report generated based on the liability determination result, the verification result, and the risk control management result for the pending claim policy includes: Based on the verification results and the risk control management results, the risk level of the liability determination result is determined; If the verification result indicates that the interpretation of liability is inconsistent, or if the risk control management result indicates that there is a risk control risk, the review conclusion is updated to indicate manual review, and an review report for the policy pending claim is generated based on the updated liability determination result. If the verification result indicates that the interpretation of liability is consistent and the risk control management result indicates that there is no risk control risk, an audit report for the policy pending claim is generated based on the liability determination result.

8. The method according to claim 1, characterized in that, The liability determination result also includes: an audit conclusion. After the method uses a decision orchestration agent to perform supplementary judgment processing on the liability determination result to obtain the audit report of the policy pending claim, it further includes: If the audit conclusion indicates that manual review is required, the associated information of the audit report is output to the first preset client. If the audit conclusion indicates that supplementary evidence is required, a supplementary evidence prompt message will be output to the second preset client. If the audit conclusion indicates that the audit is approved or the claim is rejected, an automatic claim processing operation will be performed on the policy pending claim based on the audit report.

9. The method according to claim 1, characterized in that, After the method employs a decision orchestration agent to perform supplementary judgment processing on the liability determination result to obtain the review report of the policy pending claim, the method further includes: The evidence information, the processing results, the liability determination results, and the audit report are stored together.

10. The method according to claim 1, characterized in that, After the decision orchestration agent performs supplementary judgment processing on the liability determination result to obtain the review report of the policy pending claim, the process further includes: The evidence information and the processing result are mapped to a unified timeline. The responsibility determination result is associated with the corresponding evidence information on the unified timeline to obtain an evidence chain diagram; The processing results and / or responsibility determination results that indicate anomalies in the evidence chain graph are marked, so that the first preset client highlights the marked processing results and / or responsibility determination results when displaying the evidence chain spatiotemporal graph.

11. A method for verifying claims information, applied to a client-side application, characterized in that, The method includes: In response to the manual review of the audit report, obtain the data of the evidence chain diagram corresponding to the audit report; The evidence chain diagram is based on the data shown. The marked processing results and / or liability determination results are highlighted in the displayed evidence chain graph; wherein, the evidence chain graph is established by the pre-set server using the following method: performing unified timeline mapping processing on the evidence information and processing results of the audit report, mapping the evidence information and processing results to a unified timeline; associating the liability determination results with the corresponding evidence information on the unified timeline to obtain the evidence chain graph; and marking the processing results and / or liability determination results in the evidence chain graph that indicate an audit anomaly.

12. A claims information verification system, characterized in that, The system includes: a client and a server, wherein, The server is used to perform the steps of the method as described in any one of claims 1-10; The client is used to perform the steps of the method as described in claim 11.

13. An electronic device, characterized in that, include: The electronic device includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method as described in any one of claims 1-11.

14. A computer-readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the method as described in any one of claims 1-11.

15. A computer program product comprising a computer program / computer-executable instructions, characterized in that, When the computer program / computer executable instructions are executed by a processor in an electronic device, the method of any one of claims 1-11 is implemented.

Images