Claim settlement information analysis processing method and device, computer equipment and storage medium

Abstract

The invention relates to a claim settlement information analysis processing method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring standard insurance related information; performing knowledge extraction on the standard insurance related information to obtain claim settlement entity information and claim settlement relation information; performing knowledge fusion on the claim settlement entity information and the claim settlement relation information to construct a claim settlement knowledge graph; performing text coding on the claim knowledge graph to obtain a semantic representation vector, and performing comparative learning to obtain a semantic embedding vector; performing representation learning on the claim knowledge graph based on a GNN model to obtain a claim knowledge graph structure; and performing execution process judgment according to the semantic embedding vector and the claim knowledge graph structure based on the scene type information to obtain an execution process result, and performing execution operation of the business process according to the execution process result. The method and the device can be applied to financial service system application scenes, and can effectively realize rapid and accurate analysis of claim settlement information and corresponding service process execution.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, which can be applied in the fields of finance and medicine, and in particular relates to a claim settlement information analysis processing method and device, a computer device and a storage medium. BACKGROUND

[0002] With the rapid development of the insurance industry, the complexity and data volume of insurance claim settlement business are increasing, and the traditional claim settlement information processing method has been difficult to meet the needs of modern insurance business. The insurance claim settlement process involves a large amount of text data, image data and interaction of multi-party information, and how to efficiently and accurately process these information has become an important challenge faced by the insurance industry.

[0003] However, the prior art uses simple keyword matching or basic semantic analysis method when processing insurance claim settlement text, and lacks the ability to understand the deep semantics of the text. This leads to misjudgment when processing claim settlement applications with similar descriptions but different actual meanings, affecting the accuracy of claim settlement decisions.

[0004] At the same time, the prior art has difficulty in accurately identifying the relationship between different entities and potential risk factors when dealing with complex claim settlement scenarios. In particular, in complex claim settlement cases involving multiple insurance products, customers and risk events, existing methods have difficulty in effectively analyzing the relevance between them, for example, in the financial technology field, a claim settlement application mentions "investment loss due to market crash", the existing method may not be able to accurately identify the causal relationship between market fluctuations and investment loss, thereby affecting the accuracy of claim settlement decisions. Or in the field of digital medicine, the medical record mentions "patient suffered from cerebral hemorrhage due to hypertension", the existing method may not be able to accurately identify the causal relationship between hypertension and cerebral hemorrhage, thereby affecting the claim settlement decision. Therefore, it is unable to automatically adjust and optimize the business process according to the characteristics of different claim settlement scenarios, resulting in low efficiency of claim settlement processing. SUMMARY

[0005] The purpose of the embodiments of the present application is to provide a claim settlement information analysis processing method, device, computer device and storage medium to solve the problem of being unable to quickly and accurately analyze claim settlement information and execute corresponding business processes.

[0006] In a first aspect, the embodiments of the present application provide a claim settlement information analysis processing method, which adopts the technical solution as follows:

[0007] Obtain insurance-related information, preprocess the insurance-related information to obtain standard insurance-related information;

[0008] Perform knowledge extraction on the standard insurance-related information to obtain claim entity information and claim relationship information;

[0009] knowledge fusion is performed on the claim entity information and the claim relationship information to construct a claim knowledge graph;

[0010] text encoding is performed on the claim knowledge graph to obtain a semantic representation vector, and a semantic embedding vector is obtained by performing contrastive learning on the semantic representation vector based on a SimCSE algorithm;

[0011] representation learning is performed on the claim knowledge graph based on a GNN model to obtain a claim knowledge graph structure;

[0012] scene type information is obtained, execution flow judgment is performed based on the scene type information according to the semantic embedding vector and the claim knowledge graph structure to obtain an execution flow result, and an execution operation of a business flow is performed according to the execution flow result.

[0013] In a second aspect, the embodiments of the present application further provide a claim information analysis processing apparatus, which adopts the technical solutions as follows:

[0014] an information acquisition module configured to acquire insurance-related information, and perform preprocessing on the insurance-related information to obtain standard insurance-related information;

[0015] a knowledge extraction module configured to perform knowledge extraction on the standard insurance-related information to obtain claim entity information and claim relationship information;

[0016] a knowledge fusion module configured to perform knowledge fusion on the claim entity information and the claim relationship information to construct a claim knowledge graph;

[0017] a contrastive learning module configured to perform text encoding on the claim knowledge graph to obtain a semantic representation vector, and perform contrastive learning on the semantic representation vector based on a SimCSE algorithm to obtain a semantic embedding vector;

[0018] a representation learning module configured to perform representation learning on the claim knowledge graph based on a GNN model to obtain a claim knowledge graph structure;

[0019] an execution flow module configured to obtain scene type information, perform execution flow judgment based on the scene type information according to the semantic embedding vector and the claim knowledge graph structure to obtain an execution flow result, and perform an execution operation of a business flow according to the execution flow result.

[0020] In a third aspect, the embodiments of the present application further provide a computer device, which adopts the technical solutions as follows:

[0021] A computer device comprises a memory and a processor, the memory stores computer readable instructions, and the processor implements the steps of the claim settlement information analysis processing method according to any one of the above when executing the computer readable instructions.

[0022] In a fourth aspect, the embodiments of the present application further provide a computer readable storage medium, which adopts the technical solutions as described below.

[0023] A computer readable storage medium stores computer readable instructions, and the computer readable instructions are executed by a processor to implement the steps of the claim settlement information analysis processing method according to any one of the above.

[0024] Compared with the prior art, the embodiments of the present application have the following beneficial effects: in the embodiments, insurance related information is obtained, the insurance related information is preprocessed to obtain standard insurance related information, knowledge extraction is performed on the standard insurance related information to obtain claim entity information and claim relationship information, knowledge fusion is performed on the claim entity information and the claim relationship information to construct a claim knowledge graph, text coding is performed on the claim knowledge graph to obtain a semantic representation vector, and based on a SimCSE algorithm, comparative learning is performed on the semantic representation vector to obtain a semantic embedding vector; based on a GNN model, representation learning is performed on the claim knowledge graph to obtain a claim knowledge graph structure; scene type information is obtained, and based on the scene type information, execution flow judgment is performed according to the semantic embedding vector and the claim knowledge graph structure to obtain an execution flow result, and execution operation of a business process is performed according to the execution flow result. Thus, the fast and accurate analysis of claim settlement information and the corresponding business process execution are effectively realized. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the schemes in the present application, the drawings needed in the description of the embodiments of the present application will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0026] Figure 1 is an exemplary system architecture diagram to which the present application can be applied;

[0027] Figure 2 Flowchart of an embodiment of the claim settlement information analysis processing method according to the present application;

[0028] Figure 3 is a structural schematic diagram of an embodiment of the claim settlement information analysis processing device according to the present application;

[0029] Figure 4is a structural schematic diagram of one embodiment of the computer device according to the present application. DETAILED DESCRIPTION

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used in the description herein is for describing particular embodiments only and is not intended to be limiting of the application; the description herein and the claims of the application and the above description of the drawings herein are not to be construed as limiting upon the scope of the application; the description herein and the claims of the application and the above description of the drawings herein use the term "comprising" and "having" and any variations thereof to mean "including but not limited to".

[0031] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily all directed to the same application. It is explicitly contemplated that embodiments described herein can be combined with each other.

[0032] In order to make the technical personnel in the art better understand the scheme of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings.

[0033] As shown in Figure 1 The system architecture 100 can include a terminal device 101, a network 102 and a server 103, the terminal device 101 can be a notebook computer 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is a medium for providing a communication link between the terminal device 101 and the server 103. The network 102 can include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0034] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0035] The terminal device 101 can be various electronic devices with a display screen and supporting web browsing, in addition to the notebook computer 1011, the tablet computer 1012 or the mobile phone 1013, the terminal device 101 can also be an electronic book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer and a desktop computer, etc.

[0036] The server 103 can be a server providing various services, for example, a background server providing support for the page displayed on the terminal device 101.

[0037] It should be noted that the claim information analysis processing method provided in the embodiments of the present application is generally executed by a server / terminal device, and correspondingly, the claim information analysis processing apparatus is generally arranged in a server / terminal device.

[0038] It should be understood that Figure 1 The number of terminal devices, networks and servers in

[0039] With reference to Figure 2 , a flow chart of one embodiment of the claim information analysis processing method according to the present application is shown. The claim information analysis processing method includes the following steps:

[0040] Step S10, obtaining insurance related information, preprocessing the insurance related information to obtain standard insurance related information;

[0041] In the present embodiment, the insurance related information can be related information in the field of financial technology insurance, for example, including financial insurance contract information, claim application information, accident report information, etc., or can be related information in the field of digital medical insurance, for example, including medical insurance contract information, medical record information, examination report information, etc.

[0042] Step S20, performing knowledge extraction on the standard insurance related information to obtain claim entity information and claim relationship information;

[0043] In this embodiment, the knowledge extraction of the standard insurance-related information includes entity recognition and entity relationship extraction. For example, in the field of financial technology, entity recognition and entity relationship extraction are performed on car insurance-related information to obtain car insurance claim entity information composed of entities such as car insurance policyholder, car insurance insured, car insurance period, and car insurance accident record, and car insurance claim relationship information composed of entity relationships such as car insurance policyholder-car insurance insured and car insurance insured-car insurance accident record. In the field of digital medicine, entity recognition and entity relationship extraction are performed on medical insurance-related information to obtain health insurance claim entity information composed of entities such as health insurance policyholder, health insurance insured, health insurance period, and health insurance medical record, and health insurance claim relationship information composed of entity relationships such as health insurance policyholder-health insurance insured and health insurance insured-health insurance medical record.

[0044] Step S30, knowledge fusion is performed on the claim entity information and the claim relationship information to construct a claim knowledge graph.

[0045] In this embodiment, knowledge fusion includes entity alignment, entity relationship fusion, knowledge graph construction, and the like. Among them, entity alignment refers to uniformly processing entity information of different sources or different formats to ensure consistency and uniqueness of the entity information in the knowledge graph; entity relationship fusion refers to integrating entity relationship information of different sources to ensure accuracy and integrity of the relationship; and knowledge graph construction refers to integrating entities and relationships into a graph structure to form a complete knowledge system. Specifically, in the field of financial technology, knowledge fusion is performed on car insurance claim entity information and car insurance claim relationship information to obtain a car insurance claim knowledge graph; and in the field of digital medicine, knowledge fusion is performed on health insurance claim entity information and health insurance claim relationship information to obtain a health insurance claim knowledge graph.

[0046] Step S40, text encoding is performed on the claim knowledge graph to obtain a semantic representation vector, and a semantic embedding vector is obtained by performing contrastive learning on the semantic representation vector based on a SimCSE algorithm.

[0047] In this embodiment, text encoding refers to the process of converting text data of the claim knowledge graph into a numerical vector. Text encoding can be performed by a pre-trained BERT model. For example, in the financial technology field, the vehicle insurance claim knowledge graph is input into the pre-trained BERT model to obtain a vehicle insurance claim semantic representation vector. In the digital medical field, the health insurance claim knowledge graph is input into the pre-trained BERT model to obtain a health insurance claim semantic representation vector. Contrastive learning is an unsupervised or self-supervised learning method based on the SimCSE algorithm. It learns the representation of data by contrasting positive and negative samples, so that the representation of positive sample pairs is closer, and the representation of negative sample pairs is farther apart. SimCSE is a text representation learning method based on contrastive learning, which learns the semantic embedding vector of the text through contrastive learning. The positive and negative samples come from the text data of the claim knowledge graph. Specifically, in the financial technology field, the vehicle insurance claim semantic representation vector is subjected to positive and negative sample contrastive learning according to the SimCSE algorithm to obtain a vehicle insurance semantic embedding vector. In the digital medical field, the health insurance claim semantic representation vector is subjected to positive and negative sample contrastive learning according to the SimCSE algorithm to obtain a vehicle insurance semantic embedding vector.

[0048] Step S50, performing representation learning on the claim knowledge graph based on the GNN model to obtain a claim knowledge graph structure;

[0049] In this embodiment, representation learning refers to a graph structure representation method that converts nodes (entities) and edges (relationships) in the claim knowledge graph into vector representations to better analyze and predict the graph structure. The GNN model contains multiple layers, and each layer updates the feature vector of the node. Through multi-layer propagation, the node can aggregate more distant neighbor information, thereby capturing the global structure information of the graph. Specifically, in the financial technology field, the vehicle insurance claim knowledge graph is input into the GNN model to obtain a vehicle insurance claim knowledge graph structure. In the digital medical field, the health insurance claim knowledge graph is input into the GNN model to obtain a health insurance claim knowledge graph structure.

[0050] Step S60, obtaining scenario type information, performing flow judgment based on the semantic embedding vector and the claim knowledge graph structure according to the scenario type information, obtaining an execution flow result, and performing a business process execution operation according to the execution flow result.

[0051] In this embodiment, scenario type information refers to specific scenario classification corresponding to different claim businesses, including but not limited to "claim risk judgment", "claim fraud detection", "claim process optimization", etc. The execution flow is the execution flow result obtained by combining the scenario type information, the semantic embedding vector, and the claim knowledge graph structure, including claim risk assessment, claim fraud detection, claim process optimization, etc.

[0052] The embodiment obtains insurance related information, pre-processes the insurance related information to obtain standard insurance related information, extracts knowledge from the standard insurance related information to obtain claim entity information and claim relationship information, fuses the claim entity information and the claim relationship information to construct a claim knowledge graph, encodes the claim knowledge graph to obtain a semantic representation vector, and performs contrast learning on the semantic representation vector based on a SimCSE algorithm to obtain a semantic embedding vector. The claim knowledge graph is represented and learned based on a GNN model to obtain a claim knowledge graph structure. Scene type information is obtained, and the semantic embedding vector and the claim knowledge graph structure are used to determine an execution process based on the scene type information to obtain an execution process result, and a business process is executed according to the execution process result. Thus, the claim information can be quickly and accurately analyzed and the corresponding business process can be executed.

[0053] The embodiment method can be applied to claim information analysis and processing in a financial business system. The car insurance related insurance contract information, claim application information, accident report information, etc. are uploaded to the system, pre-processed and knowledge extracted by the system to obtain car insurance policyholder, car insurance insured, car insurance period, car insurance accident record, etc. entity and car insurance policyholder-car insurance insured, car insurance insured-car insurance accident record, etc. entity relationship. Knowledge fusion is performed according to the entity and the entity relationship to obtain a car insurance claim knowledge graph. The car insurance claim knowledge graph is processed by text encoding, contrast learning and representation learning to obtain a car insurance semantic embedding vector representing text semantic understanding and a car insurance claim knowledge graph structure representing graph structure relationship. Finally, according to the preset "claim risk judgment", "claim fraud detection", "claim process optimization" and other scene type information and the car insurance semantic embedding vector and the car insurance claim knowledge graph structure, a comprehensive execution process is determined to determine the most suitable execution process, and the corresponding business process is executed according to the determined execution process to obtain a car insurance claim risk assessment result, a car insurance claim fraud detection report, a car insurance claim optimization adjustment scheme, etc. for subsequent processing.

[0054] In some optional implementations of the embodiment, the obtaining of the insurance related information and the pre-processing of the insurance related information to obtain the standard insurance related information includes the following steps:

[0055] An information extraction identifier is obtained, and the insurance related information is extracted from a database according to the information extraction identifier;

[0056] In this embodiment, the information extraction identifier can be a claim number, a policy number, or other identifiers that can uniquely identify an insurance business. The system extracts relevant information from the insurance company's business database, customer information library, claim record library, and other data sources according to these identifiers. The extracted insurance-related information includes but is not limited to the following: the insured person's information, the insured person's information, the insurance contract information, the claim application information, the medical record, the accident report, and other insurance-related information.

[0057] The insurance-related information is subjected to data cleaning and denoising processing to obtain the standard insurance-related information.

[0058] In this embodiment, the data cleaning and denoising processing mainly deals with possible problems in the extracted raw data, including handling missing values, outliers, duplicate data, and inconsistent formats. For example, for missing customer information, the system will supplement it according to other associated data; for date data with inconsistent formats, the system will convert them to a standard format; for typos and special symbols in text information, the system will correct and standardize them. Through these processes, structured and standardized standard insurance-related information is obtained, laying the foundation for subsequent knowledge extraction.

[0059] This embodiment obtains the information extraction identifier, extracts the insurance-related information from the database according to the information extraction identifier, and performs data cleaning and denoising processing on the insurance-related information to obtain the standard insurance-related information. This effectively realizes the acquisition of standard insurance-related information to facilitate subsequent feature extraction processing.

[0060] In some optional implementations of this embodiment, the knowledge extraction of the standard insurance-related information to obtain claim entity information and claim relationship information includes the following steps:

[0061] Performing claim entity recognition on the standard insurance-related information based on a preset entity recognition model to obtain the claim entity information;

[0062] In this embodiment, the preset entity recognition model can use a BERT-based named entity recognition model that has been pre-trained and fine-tuned on a large amount of insurance domain corpus and can accurately identify key entities in insurance claim text. The claim entity information includes but is not limited to the following: person entities (such as the insured, the insured, and the beneficiary), insurance product entities (such as product names and protection types), event entities (such as accident types and disease names), time entities (such as the time of the accident and the time of the visit), and amount entities (such as the insurance amount and the claim amount).

[0063] Performing entity relationship extraction on the standard insurance-related information based on a preset relationship extraction model to obtain the claim relationship information.

[0064] In this embodiment, the preset relationship extraction model can adopt a relationship extraction network based on an attention mechanism. The model identifies the semantic relationship between entities by analyzing the context information between entities. The claim relationship information includes but is not limited to "insured person-purchase-insurance product", "insured person-suffer from-illness", "accident-cause-injury", "disease-need-treatment plan", and other semantic relationships. These relationships reflect the logical connection and business rules between entities in the insurance claim process.

[0065] In this embodiment, the claim entity information is obtained by performing claim entity recognition on the standard insurance-related information based on a preset entity recognition model, and the claim relationship information is obtained by performing entity relationship extraction on the standard insurance-related information based on a preset relationship extraction model. Thus, effective claim entity recognition and entity relationship extraction are effectively realized from the standard insurance-related information to facilitate subsequent knowledge fusion processing.

[0066] In some optional implementations of this embodiment, the knowledge fusion of the claim entity information and the claim relationship information to construct the claim knowledge graph includes the following steps:

[0067] performing entity alignment processing on the claim entity information to obtain effective entity information;

[0068] In this embodiment, entity alignment processing mainly solves the problem of different expressions of the same entity in different data sources. For example, the same customer may have different representations (such as "Zhang San" and "Mr. Zhang San") in different systems. The system merges different entity representations that refer to the same object by calculating the semantic similarity and attribute matching degree between entities, forms a unique entity identifier, and obtains effective entity information. In addition, the system also performs standardization processing on the entity to ensure that the entity name, attribute, and other information conform to the predefined standard format.

[0069] performing relationship fusion processing on the claim relationship information to obtain effective relationship information;

[0070] In this embodiment, relationship fusion processing mainly solves the problems of relationship redundancy and relationship conflict. For different relationship expressions expressing the same semantics (such as "purchase" and "insure"), the system unifies them into a standard relationship type; for conflicting relationships (such as the same insured person being marked as different disease types), the system judges and processes them according to the reliability, time sequence, and other factors of the data source to obtain effective relationship information.

[0071] constructing a knowledge graph according to the effective entity information and the effective relationship information to obtain the claim knowledge graph.

[0072] In this embodiment, in the knowledge graph construction process, the system takes the effective entity as a node in the graph and takes the effective relationship as an edge in the graph to form a complete claim knowledge network. Each node contains attribute information of the entity (such as the age, gender, occupation, etc. of the customer), and each edge contains the type and attribute of the relationship (such as the effective date and guarantee range of the insurance contract). The completed claim knowledge graph can intuitively display the complex associations between various elements in the claim case and provide support for subsequent analysis and decision-making.

[0073] In this embodiment, the effective entity information is obtained by performing entity alignment processing on the claim entity information; the effective relationship information is obtained by performing relationship fusion processing on the claim relationship information; and the claim knowledge graph is obtained by performing knowledge graph construction according to the effective entity information and the effective relationship information. Thus, the effective claim knowledge graph is effectively constructed according to the effective entity information and the effective relationship information, facilitating subsequent text encoding and representation learning processing.

[0074] In some optional implementation manners of this embodiment, the text encoding of the claim knowledge graph to obtain the semantic representation vector and the contrast learning of the semantic representation vector based on the SimCSE algorithm to obtain the semantic embedding vector include the following steps:

[0075] The text information of the claim knowledge graph is encoded based on the pre-trained language model to obtain the semantic representation vector;

[0076] In this embodiment, the pre-trained language model adopts a BERT model fine-tuned for the financial insurance field. By inputting the entity name, attribute description, and other text content in the insurance knowledge graph into the BERT model, the corresponding vector representation is obtained. Then, the vectors are averaged or spliced to obtain the semantic representation vector of the entire knowledge graph.

[0077] The semantic positive and negative sample pairs are constructed based on the SimCSE algorithm for the semantic representation vector;

[0078] In this embodiment, the SimCSE (Simple Contrastive Learning of Sentence Embeddings) algorithm is an effective contrastive learning method that can improve the quality of semantic representation. In this embodiment, the system constructs semantic positive and negative sample pairs based on the SimCSE algorithm: for each original text, generate its positive sample (semantically the same but expressed slightly differently) by adding different dropout masks, and use other texts in the batch as negative samples (semantically different). Specifically, in the insurance knowledge graph, different descriptions of the same entity or descriptions of related entities can be used as positive sample pairs, and descriptions of unrelated entities can be used as negative sample pairs. By continuously optimizing the parameters of the SimCSE model, the semantic similarity of positive sample pairs is higher, and the semantic similarity of negative sample pairs is lower, so that more accurate semantic embedding vectors are obtained.

[0079] According to the similarity and difference of the semantic positive and negative sample pairs, the semantic embedding vector is obtained.

[0080] In this embodiment, during the contrastive learning process, the system maximizes the similarity between positive sample pairs and minimizes the similarity with negative samples, so that texts with similar semantics are closer in vector space and texts with different semantics are farther apart in vector space. In this way, the semantic embedding vector obtained by the system has better semantic discrimination ability and can more accurately express the semantic information of the claim knowledge.

[0081] This embodiment encodes and processes the text information of the claim knowledge graph based on a pre-trained language model to obtain the semantic representation vector; constructs semantic positive and negative sample pairs based on the SimCSE algorithm; and performs contrastive learning of similarity and difference according to the semantic positive and negative sample pairs to obtain the semantic embedding vector. Thus, the semantic embedding vector for text understanding of the claim knowledge graph is effectively obtained to facilitate subsequent execution flow judgment operations.

[0082] In some optional implementations of this embodiment, the representation learning of the claim knowledge graph based on the GNN model includes the following steps:

[0083] Loading and standardizing the original node features of the claim knowledge graph based on the GNN model to obtain an initial node feature matrix;

[0084] In this embodiment, the GNN (Graph Neural Network) model adopts the GraphSAGE structure, which can effectively process large-scale heterogeneous graph data. First, the system loads the original features of each node in the claim knowledge graph (such as entity attribute information, semantic vectors of text descriptions, etc.) as initial features, and performs standardization processing (such as normalization, mean removal, etc.) to form an initial node feature matrix.

[0085] An adjacency matrix is constructed for the entities and relationships of the claim knowledge graph to obtain a graph structure representation.

[0086] In this embodiment, during the adjacency matrix construction process, the system generates an adjacency matrix describing the graph structure according to the entity relationships in the claim knowledge graph. For a heterogeneous graph (containing multiple types of nodes and edges), the system will construct multiple relationship-specific adjacency matrices to describe the connection of different types of relationships.

[0087] The neighbor node features in the graph structure representation are processed by an aggregation function to obtain an aggregated neighbor feature representation.

[0088] In this embodiment, in the aggregation function processing stage, the system aggregates the neighbor node features of each node. Common aggregation methods include average aggregation, maximum value aggregation, attention weighted aggregation, etc. This embodiment uses an aggregation method based on attention mechanism, which weights and aggregates the neighbor nodes and the center node according to the importance of the relationship, to obtain a more representative neighbor feature representation.

[0089] The initial node feature matrix and the aggregated neighbor feature representation are processed by an update function to obtain an updated node feature matrix.

[0090] In this embodiment, in the update function processing stage, the system combines the initial features of the nodes with the aggregated neighbor features (such as concatenation, weighted summation, etc.), and generates updated node features through nonlinear transformation. This process can be iterated multiple times, so that the representation of each node can integrate more extensive graph structure information, thereby obtaining an effective updated node feature matrix.

[0091] The updated node feature matrix is extracted to obtain the claim knowledge graph structure.

[0092] In this embodiment, in the update function processing stage, the system combines the initial features of the nodes with the aggregated neighbor features (such as concatenation, weighted summation, etc.), and generates updated node features through nonlinear transformation. This process can be iterated multiple times, so that the representation of each node can integrate more extensive graph structure information, thereby obtaining an effective claim knowledge graph structure.

[0093] The embodiment loads and standardizes the original node features of the claim knowledge graph based on the GNN model to obtain an initial node feature matrix; constructs an adjacency matrix for entities and relationships of the claim knowledge graph to obtain a graph structure representation; performs an aggregation function processing on neighbor node features in the graph structure representation to obtain an aggregated neighbor feature representation; performs an update function processing on the initial node feature matrix and the aggregated neighbor feature representation to obtain an updated node feature matrix; and performs an extraction processing on the updated node feature matrix to obtain the claim knowledge graph structure. Thus, the claim knowledge graph structure is effectively obtained to facilitate subsequent execution flow judgment operations.

[0094] In some optional implementations of the embodiment, the obtaining of the scene type information, the execution flow judgment based on the semantic embedding vector and the claim knowledge graph structure according to the scene type information, the obtaining of an execution flow result, and the execution operation of the business flow according to the execution flow result include the following steps:

[0095] performing an encoding processing on the scene type information to obtain a scene feature vector;

[0096] In the embodiment, the scene type information refers to specific scene classifications involved in claim business processing, such as “claim risk judgment”, “claim fraud detection”, “claim process optimization”, etc. Different claim scenes may require different processing flows and decision rules. The system first performs an encoding processing on the scene type information to convert the text description into a numerical vector representation, i.e., a scene feature vector.

[0097] performing fusion on the scene feature vector, the semantic embedding vector, and the claim knowledge graph structure to obtain a fused feature vector;

[0098] In the embodiment, in the feature fusion stage, the system fuses the scene feature vector, the semantic embedding vector (reflecting the semantic information of the claim case), and the claim knowledge graph structure (reflecting the association relationship between claim elements). The fusion methods include simple vector splicing, weighted summation, or more complex attention mechanism fusion, etc. The embodiment adopts a multi-head attention mechanism for feature fusion, which can adaptively focus on important parts of different features to obtain a more expressive fused feature vector.

[0099] inputting the fused feature vector into a pre-trained process classification model to obtain an execution flow result;

[0100] In the embodiment, the pre-trained process classification model is a multi-layer neural network that is trained on a large number of historical claim cases and can predict the most suitable execution flow according to the input fused feature vector. The execution flow result includes claim risk assessment, claim fraud detection, claim process optimization, etc.

[0101] Specifically, the claim risk assessment utilizes SimCSE to generate semantic embedding vectors for the text content in the claim application, such as accident descriptions and customer information, and combines GNN's graph structure analysis to consider the correlation between claim cases and the entity relationships in the knowledge graph. For example, through GNN model analysis of the relationships between entities such as insurance products, customers, and risk events involved in different claim cases, it is determined whether there are potential risk transmission paths or common risk factors. Then, according to the analysis results, the claim risk is quantitatively evaluated. Historical claim data and statistical analysis methods can be combined to calculate claim risk indicators, such as the mean, standard deviation, and payout rate of claim amounts. At the same time, the predictive ability of the GNN model is used to predict and warn of future claim risks. Finally, the insurance company is provided with a claim risk assessment report to help the insurance company develop reasonable claim strategies and risk management measures. For example, according to the risk assessment results, the insurance company can adjust the claim review process, strengthen risk control measures, and optimize insurance product design, etc.

[0102] Claim fraud detection utilizes SimCSE's semantic similarity calculation function to compare the text descriptions between different claim cases to detect whether there are abnormally similar claim applications. For example, for similar text content such as accident descriptions and customer information in multiple claim applications, the semantic similarity between them is calculated. If the similarity exceeds a certain threshold, there may be suspicion of claim fraud. Then, combined with GNN's graph structure analysis, other entities and relationships related to the suspicious claim case are found. For example, through GNN model analysis of the relationships between entities such as insurance products, customers, and agents involved in the suspicious claim case, and the correlation with other claim cases, if it is found to be related to known claim fraud patterns or high-risk entities, the likelihood of fraud is further increased. Finally, a variety of fraud detection methods such as rule engines and machine learning algorithms are used to confirm and classify claim fraud. For example, pre-defined fraud rules such as false accident reports and repeated claims, as well as machine learning algorithms such as decision trees and random forests, can be combined to further analyze and judge suspicious claim cases. Thus, the insurance company is provided with a claim fraud detection report to help the insurance company timely detect and handle claim fraud behavior, reducing claim costs and risks.

[0103] The claim settlement process optimization is to analyze the relationships in the knowledge graph and the characteristics of the claim settlement cases, identify the bottlenecks and optimization points in the claim settlement process. For example, using the GNN model to analyze the processing time, approval process, resource allocation, etc. of each link in the claim settlement case, to find out the links that may cause delay or low efficiency in claim settlement. Combined with the semantic understanding function of SimCSE, the text description in the claim settlement application is automatically classified and prioritized. For example, for different types of claim settlement applications, such as car insurance claim settlement, health insurance claim settlement, etc., the semantic embedding vector is generated by the SimCSE algorithm, and then the automatic classification and priority sorting are performed by using the machine learning algorithm, so as to better arrange the claim settlement resources and processing order. Finally, the claim settlement process optimization suggestions are put forward, such as simplifying the approval process, automating part of the links, strengthening communication with customers, etc. For example, according to the analysis results, the insurance company can use the automatic claim settlement system to quickly process some simple claim settlement applications; or strengthen communication with customers and timely feedback on the claim settlement progress to improve customer satisfaction. The optimized claim settlement process can also be monitored and evaluated to continuously improve and perfect the claim settlement service. For example, by collecting and analyzing claim settlement data, the effect of the optimized claim settlement process is evaluated, and the optimization measures are adjusted and improved in a timely manner to ensure the efficiency and quality of the claim settlement service.

[0104] According to the execution process result, the business process mapping is performed, and the mapped business process is executed.

[0105] In this embodiment, the business process mapping is to convert the execution process result into specific business operation steps. For example, for the claim settlement case of "claim settlement risk assessment", the system will automatically generate the claim settlement risk assessment result and push it to the relevant personnel; for the case of "claim settlement fraud detection", the system will automatically generate the claim settlement fraud detection report and push it for display; for the case of "claim settlement process optimization", the system will generate the claim settlement optimization adjustment scheme for subsequent viewing and processing.

[0106] In this embodiment, the scene type information is encoded to obtain a scene feature vector; the scene feature vector, the semantic embedding vector, and the claim settlement knowledge graph structure are fused to obtain a fused feature vector; the fused feature vector is input into a pre-trained process classification model to obtain an execution process result; and according to the execution process result, the business process mapping is performed, and the mapped business process is executed. Thus, multi-dimensional comprehensive judgment is effectively realized according to the scene feature vector, the semantic embedding vector, and the claim settlement knowledge graph structure, so as to realize effective information analysis and process execution.

[0107] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related hardware through computer readable instructions, and the computer readable instructions can be stored in a computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiment methods. Among them, the storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0108] It should be understood that although each step in the flowchart of the accompanying drawings is shown in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and they can be executed in other orders. Moreover, at least part of the steps in the flowchart of the accompanying drawings can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence is not necessarily sequential, but can be alternately executed with other steps or sub-steps or stages of other steps.

[0109] Further referring to Figure 3 , as an implementation of the method shown in Figure 1 , the present application provides an embodiment of a claim information analysis processing device, which corresponds to the method embodiment shown in Figure 1 , and the device can be applied to various electronic devices.

[0110] As shown in Figure 3 , the claim information analysis processing device 700 described in the embodiment includes an information acquisition module 701, a knowledge extraction module 702, a knowledge fusion module 703, a contrast learning module 704, a representation learning module 705, and a process execution module 706. Among them:

[0111] The information acquisition module 701 is configured to acquire insurance-related information, pre-process the insurance-related information, and obtain standard insurance-related information.

[0112] The knowledge extraction module 702 is configured to extract knowledge from the standard insurance-related information to obtain claim entity information and claim relationship information.

[0113] The knowledge fusion module 703 is configured to fuse the claim entity information and the claim relationship information to construct a claim knowledge graph.

[0114] The contrast learning module 704 is configured to perform text encoding on the claim settlement knowledge graph to obtain a semantic representation vector, and perform contrast learning on the semantic representation vector based on a SimCSE algorithm to obtain a semantic embedding vector.

[0115] The representation learning module 705 is configured to perform representation learning on the claim settlement knowledge graph based on a GNN model to obtain a claim settlement knowledge graph structure.

[0116] The process execution module 706 is configured to obtain scene type information, perform process execution determination based on the semantic embedding vector and the claim settlement knowledge graph structure according to the scene type information to obtain a process execution result, and perform a business process execution operation according to the process execution result.

[0117] The claim settlement information analysis processing device can obtain insurance related information, pre-process the insurance related information to obtain standard insurance related information, extract knowledge from the standard insurance related information to obtain claim settlement entity information and claim settlement relationship information, fuse the claim settlement entity information and the claim settlement relationship information to construct a claim settlement knowledge graph, perform text encoding on the claim settlement knowledge graph to obtain a semantic representation vector, perform contrast learning on the semantic representation vector based on a SimCSE algorithm to obtain a semantic embedding vector, perform representation learning on the claim settlement knowledge graph based on a GNN model to obtain a claim settlement knowledge graph structure, obtain scene type information, perform process execution determination based on the semantic embedding vector and the claim settlement knowledge graph structure according to the scene type information to obtain a process execution result, and perform a business process execution operation according to the process execution result. Thus, the claim settlement information can be quickly and accurately analyzed, and the corresponding business process execution can be effectively realized.

[0118] In some optional implementations of the present embodiment, the information acquisition module 701 includes an identifier extraction unit and an information preprocessing unit. Wherein:

[0119] The identifier extraction unit is configured to obtain an information extraction identifier, and extract the insurance related information from a database according to the information extraction identifier.

[0120] The information preprocessing unit is configured to perform data cleaning and denoising processing on the insurance related information to obtain the standard insurance related information.

[0121] The information acquisition module 701 including the identifier extraction unit and the information preprocessing unit is configured to effectively obtain standard and effective user basic information, user behavior information, and user related information, so as to facilitate subsequent feature extraction processing.

[0122] In some optional implementations of the embodiment, the knowledge extraction module 702 includes an entity recognition unit and a relationship extraction unit. Wherein:

[0123] The entity recognition unit is configured to perform claim settlement entity recognition on the standard insurance-related information based on a preset entity recognition model to obtain the claim settlement entity information.

[0124] The relationship extraction unit is configured to perform entity relationship extraction on the standard insurance-related information based on a preset relationship extraction model to obtain the claim settlement relationship information.

[0125] The embodiment sets the knowledge extraction module 702 including the entity recognition unit and the relationship extraction unit, thereby effectively realizing hierarchical feature extraction on the user basic information, the user behavior information, and the user related information to facilitate subsequent feature clustering processing.

[0126] In some optional implementations of the embodiment, the knowledge fusion module 703 includes an entity alignment unit, a relationship fusion unit, and a knowledge graph construction unit. Wherein:

[0127] The entity alignment unit is configured to perform entity alignment processing on the claim settlement entity information to obtain effective entity information.

[0128] The relationship fusion unit is configured to perform relationship fusion processing on the claim settlement relationship information to obtain effective relationship information.

[0129] The knowledge graph construction unit is configured to construct a knowledge graph based on the effective entity information and the effective relationship information to obtain the claim settlement knowledge graph.

[0130] The embodiment sets the knowledge fusion module 703 including the entity alignment unit, the relationship fusion unit, and the knowledge graph construction unit, thereby effectively realizing accurate and effective user feature clustering on the first user feature, the second user feature, and the third user feature to obtain a clustered user feature set, facilitating subsequent scene demand matching processing.

[0131] In some optional implementations of the embodiment, the contrast learning module 704 includes an encoding processing unit, a sample pair construction unit, and a contrast learning unit. Wherein:

[0132] The encoding processing unit is configured to perform encoding processing on text information of the claim settlement knowledge graph based on a pre-trained language model to obtain the semantic representation vector.

[0133] The sample pair construction unit is configured to construct a semantic positive-negative sample pair based on the semantic representation vector and the SimCSE algorithm.

[0134] The contrast learning unit is configured to perform similarity and difference contrast learning according to the semantic positive and negative sample pairs, and obtain the semantic embedding vector.

[0135] The contrast learning module 704 includes an encoding processing unit, a sample pair construction unit, and a contrast learning unit. The contrast learning module 704 is configured to perform category classification on the user feature set according to the specific requirement setting of the scene, so as to facilitate subsequent user similarity matching processing according to the classified scene requirement classification result.

[0136] In some optional implementation manners of the embodiment, the representation learning module 705 includes an initial matrix obtaining unit, a graph structure representation unit, an aggregation processing unit, an updated matrix unit, and a graph structure obtaining unit.

[0137] The initial matrix obtaining unit is configured to load and perform standardization processing on original node features of the claim knowledge graph based on the GNN model, and obtain an initial node feature matrix.

[0138] The graph structure representation unit is configured to perform adjacency matrix construction on entities and relationships of the claim knowledge graph, and obtain a graph structure representation.

[0139] The aggregation processing unit is configured to perform an aggregation function processing on neighbor node features in the graph structure representation, and obtain an aggregated neighbor feature representation.

[0140] The updated matrix unit is configured to perform an updated function processing on the initial node feature matrix and the aggregated neighbor feature representation, and obtain an updated node feature matrix.

[0141] The graph structure obtaining unit is configured to perform extraction processing on the updated node feature matrix, and obtain the claim knowledge graph structure.

[0142] The representation learning module 705 includes an initial matrix obtaining unit, a graph structure representation unit, an aggregation processing unit, an updated matrix unit, and a graph structure obtaining unit. The representation learning module 705 is configured to effectively arrange user similarity situations, so as to obtain a corresponding user similarity ranking table, and facilitate subsequent real-time user information matching.

[0143] In some optional implementation manners of the embodiment, the flow execution module 706 includes an information encoding unit, a feature fusion unit, a model processing unit, and an operation execution unit.

[0144] In some optional implementation manners of the embodiment, the flow execution module 706 includes an information encoding unit, a feature fusion unit, a model processing unit, and an operation execution unit.

[0145] The information encoding unit is configured to perform encoding processing on the scene type information, and obtain a scene feature vector.

[0146] The feature fusion unit is used to fuse the scene feature vector, the semantic embedding vector, and the claims knowledge graph structure to obtain a fused feature vector;

[0147] The model processing unit is used to input the fused feature vector into a pre-trained process classification model to obtain the execution process result;

[0148] The operation execution unit is used to map business processes based on the execution process results and to perform operations on the mapped business processes.

[0149] This embodiment effectively achieves similar user data search of real-time user information by setting up a process execution module 706 including an information encoding unit, a feature fusion unit, a model processing unit, and an operation execution unit, thereby effectively improving the accuracy and efficiency of similar user data search.

[0150] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0151] The computer device 8 includes a memory 81, a processor 82, and a network interface 83 that are interconnected via a system bus. It should be noted that only the computer device 8 with components 81-83 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0152] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0153] The memory 81 includes at least one type of readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 81 can be an internal storage unit of the computer device 8, such as a hard disk or a memory of the computer device 8. In other embodiments, the memory 81 can also be an external storage device of the computer device 8, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device 8. Of course, the memory 81 can also include both an internal storage unit and an external storage device of the computer device 8. In this embodiment, the memory 81 is generally used to store an operating system and various application software installed on the computer device 8, such as computer readable instructions of the claim information analysis processing method, etc. In addition, the memory 81 can also be used to temporarily store various data that have been output or will be output.

[0154] The processor 82 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 82 is generally used to control the overall operation of the computer device 8. In this embodiment, the processor 82 is used to run computer readable instructions or process data stored in the memory 81, such as computer readable instructions of the claim information analysis processing method.

[0155] The network interface 83 can include a wireless network interface or a wired network interface, and is generally used to establish a communication connection between the computer device 8 and other electronic devices.

[0156] The embodiment can obtain insurance related information, preprocess the insurance related information to obtain standard insurance related information, extract knowledge from the standard insurance related information to obtain claim entity information and claim relationship information, fuse the claim entity information and the claim relationship information to construct a claim knowledge graph, perform text coding on the claim knowledge graph to obtain a semantic representation vector, perform comparative learning on the semantic representation vector based on a SimCSE algorithm to obtain a semantic embedding vector, perform representation learning on the claim knowledge graph based on a GNN model to obtain a claim knowledge graph structure, obtain scene type information, perform execution process judgment based on the semantic embedding vector and the claim knowledge graph structure according to the scene type information to obtain an execution process result, and perform business process execution operation according to the execution process result. Therefore, the fast and accurate analysis of the claim information and the corresponding business process execution are effectively realized.

[0157] The application also provides another implementation, that is, a computer readable storage medium storing computer readable instructions executable by at least one processor to cause the at least one processor to perform the steps of the claim information analysis processing method as described above.

[0158] The embodiment can obtain insurance related information, preprocess the insurance related information to obtain standard insurance related information, extract knowledge from the standard insurance related information to obtain claim entity information and claim relationship information, fuse the claim entity information and the claim relationship information to construct a claim knowledge graph, perform text coding on the claim knowledge graph to obtain a semantic representation vector, perform comparative learning on the semantic representation vector based on a SimCSE algorithm to obtain a semantic embedding vector, perform representation learning on the claim knowledge graph based on a GNN model to obtain a claim knowledge graph structure, obtain scene type information, perform execution process judgment based on the semantic embedding vector and the claim knowledge graph structure according to the scene type information to obtain an execution process result, and perform business process execution operation according to the execution process result. Therefore, the fast and accurate analysis of the claim information and the corresponding business process execution are effectively realized.

[0159] Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and necessary general hardware platform, of course, also can be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of software product, and the computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a plurality of instructions to make a terminal device (may be a mobile phone, computer, server, air conditioner, or network device, etc.) execute the method described in each embodiment of the present application.

[0160] Obviously, the above-described embodiments are only a part of the embodiments of the present application, rather than all the embodiments, and the preferred embodiments of the present application are given in the drawings, but do not limit the patent scope of the present application. The present application can be realized in many different forms, and conversely, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing specific embodiments, or make equivalent replacements to some technical features. Any equivalent structure made by using the contents of the specification and drawings, directly or indirectly applied to other related technical fields, is also within the scope of the patent protection of the present application.

[0161] The non-company software tools or components appearing in the embodiments of the present application are only examples for introduction, not representing actual use.

Claims

1. A method for analyzing and processing claims information, characterized in that, Includes the following steps: Obtain insurance-related information, preprocess the insurance-related information to obtain standard insurance-related information; Knowledge extraction is performed on the standard insurance-related information to obtain claims entity information and claims relationship information; Knowledge fusion is performed on the claims entity information and the claims relationship information to construct a claims knowledge graph; The claims knowledge graph is text-encoded to obtain a semantic representation vector, and the semantic representation vector is compared and learned based on the SimCSE algorithm to obtain a semantic embedding vector; The claims knowledge graph structure is obtained by performing representation learning on the GNN model. Obtain scenario type information, and based on the scenario type information, determine the execution process according to the semantic embedding vector and the claims knowledge graph structure to obtain the execution process result, and perform business process execution operations according to the execution process result.

2. The claims information analysis and processing method according to claim 1, characterized in that, The steps of obtaining insurance-related information and preprocessing the insurance-related information to obtain standard insurance-related information specifically include: Obtain an information extraction identifier, and extract the insurance-related information from the database based on the information extraction identifier; The insurance-related information is cleaned and denoised to obtain the standard insurance-related information.

3. The claims information analysis and processing method according to claim 1, characterized in that, The step of extracting knowledge from the standard insurance-related information to obtain claims entity information and claims relationship information specifically includes: Based on a preset entity recognition model, the standard insurance-related information is used to identify claim entities to obtain the claim entity information; Based on a preset relationship extraction model, entity relationships are extracted from the standard insurance-related information to obtain the claims relationship information.

4. The claims information analysis and processing method according to claim 1, characterized in that, The step of performing knowledge fusion on the claims entity information and the claims relationship information to construct a claims knowledge graph specifically includes: The claim entity information is aligned to obtain valid entity information; The claims relationship information is processed by relationship fusion to obtain valid relationship information; The claims knowledge graph is constructed based on the valid entity information and the valid relationship information.

5. The claims information analysis and processing method according to claim 1, characterized in that, The steps of text encoding the claims knowledge graph to obtain semantic representation vectors, and then performing comparative learning on the semantic representation vectors based on the SimCSE algorithm to obtain semantic embedding vectors, specifically include: The text information of the claims knowledge graph is encoded based on a pre-trained language model to obtain the semantic representation vector; Based on the SimCSE algorithm, construct semantic positive and negative sample pairs from the semantic representation vector; The semantic embedding vector is obtained by comparing and learning the similarity and differences between the semantic positive and negative sample pairs.

6. The claims information analysis and processing method according to claim 1, characterized in that, The step of performing representation learning on the claims knowledge graph based on the GNN model to obtain the claims knowledge graph structure specifically includes: Based on the GNN model, the original node features of the claims knowledge graph are loaded and standardized to obtain an initial node feature matrix; An adjacency matrix is ​​constructed from the entities and relationships of the claims knowledge graph to obtain a graph structure representation; The neighbor node features in the graph structure representation are processed by an aggregation function to obtain an aggregated neighbor feature representation; The initial node feature matrix and the aggregated neighbor feature representation are processed by an update function to obtain the updated node feature matrix; The feature matrix of the updated node is extracted and processed to obtain the claims knowledge graph structure.

7. The claims information analysis and processing method according to claim 1, characterized in that, The steps of obtaining scenario type information, determining the execution process based on the semantic embedding vector and the claims knowledge graph structure according to the scenario type information, obtaining the execution process result, and performing business process execution operations based on the execution process result specifically include: The scene type information is encoded to obtain a scene feature vector; The scene feature vector, the semantic embedding vector, and the claims knowledge graph structure are fused to obtain a fused feature vector. The fused feature vector is input into a pre-trained process classification model to obtain the execution process result; Based on the execution process results, business processes are mapped, and the mapped business processes are executed.

8. A claims information analysis and processing device, characterized in that, include: The information acquisition module is used to acquire insurance-related information, preprocess the insurance-related information, and obtain standard insurance-related information. The knowledge extraction module is used to extract knowledge from the standard insurance-related information to obtain claims entity information and claims relationship information; The knowledge fusion module is used to perform knowledge fusion on the claims entity information and the claims relationship information to construct a claims knowledge graph. The contrastive learning module is used to encode the claims knowledge graph into text to obtain a semantic representation vector, and to perform contrastive learning on the semantic representation vector based on the SimCSE algorithm to obtain a semantic embedding vector. The representation learning module is used to learn the representation of the claims knowledge graph based on the GNN model to obtain the claims knowledge graph structure; The process execution module is used to obtain scenario type information, perform execution process judgment based on the scenario type information according to the semantic embedding vector and the claims knowledge graph structure, obtain the execution process result, and perform business process execution operations based on the execution process result.

9. A computer device, characterized in that, The method includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the claims information analysis and processing method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the claims information analysis and processing method as described in any one of claims 1 to 7.

Images