Insurance policy information auditing method, device and equipment, storage medium and program product

By finding the text information of the compensation image in the knowledge graph instead of directly comparing the images, the problem of high computing resource consumption in the prior art is solved, and more efficient compensation image review is achieved.

CN118967335APending Publication Date: 2024-11-15PEOPLE'S INSURANCE COMPANY OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411345957.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-25
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In the prior art, the calculation resource consumption is high in determining whether the compensation image is forged.

Method used

By receiving the compensation pictures sent by the terminal device, identifying text information, naming entity recognition, filtering out the text to be queried, and finding relevant nodes in the preset knowledge graph. If the risk node is included, obtain the policy information corresponding to the compensation picture. Send to the audit terminal to reduce the consumption of computing resources.

Benefits of technology

It reduces the consumption of computing resources, improves efficiency, and reduces the demand for computing resources.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967335A_ABST
    Figure CN118967335A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides an insurance policy information auditing method and device, equipment, a storage medium and a program product. Belongs to the technical field of data processing and big data. The method comprises the following steps: receiving a compensation picture sent by terminal equipment; determining a to-be-queried text according to the claim payment picture; searching related nodes of the to-be-queried text in a preset knowledge graph; if the related nodes comprise risk nodes, insurance policy information corresponding to the claim payment picture is acquired; and sending the insurance policy information to an auditing terminal. By identifying the information in the picture and comparing the information in the knowledge graph, whether manual auditing is needed is determined, and the problem of high computing resource consumption of a current mode for determining whether the compensation image is forged or not is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data processing and big data technology, and in particular to a policy information review method, device, equipment, storage medium and program product. Background Art

[0002] With the continuous development of the economy and society, insurance has become a financial product that is widely accepted by the public. In order to protect the value of property, property insurance is usually required, and when there is an abnormality in the property, the property insurance company will pay compensation.

[0003] At present, in the prior art, when insurance companies make claims, they usually use image comparison algorithms to compare the images used in the claims payment to identify whether the claims payment images are forged.

[0004] However, the inventors have discovered that the prior art has at least the following technical problems: the current method of determining whether a compensation image is forged consumes a high amount of computing resources. Summary of the invention

[0005] The embodiments of the present application provide a policy information review method, apparatus, device, storage medium and program product to solve the problem of high computing resource consumption in the current method of determining whether a claim image is forged.

[0006] In a first aspect, an embodiment of the present application provides a method for reviewing insurance policy information, comprising: receiving a compensation picture sent by a terminal device; determining a text to be queried based on the compensation picture; searching for relevant nodes of the text to be queried in a preset knowledge graph; if the relevant nodes include a risk node, obtaining the insurance policy information corresponding to the compensation picture; and sending the insurance policy information to the review terminal.

[0007] In one possible implementation, determining the text to be queried based on the compensation picture includes: identifying the text information in the compensation picture; performing named entity recognition on the text information to obtain the text type corresponding to each text information; deleting the text information with a preset text type, and determining the remaining text information as the text to be queried.

[0008] In a possible implementation, before searching for relevant nodes of the text to be queried in a preset knowledge graph, it also includes: obtaining at least one training image; using the target training image to generate at least one enhanced training image; determining the target training image and the enhanced training image as a target training image group; determining the target training text according to the target training image group; deduplicating the training text to obtain deduplicated text; removing similar characters in the deduplicated text corresponding to each training image group to obtain a further deduplicated text; removing invalid information in the further deduplicated text to obtain a graph construction text; using the graph construction text to create a knowledge graph.

[0009] In a possible implementation, after using a graph to construct text and create a knowledge graph, it also includes: determining the topological attributes of each node in the knowledge graph; determining the feature vector of the target node based on the topological attributes of the target node; clustering the feature vectors of each node to obtain clusters; and determining abnormal nodes based on the clusters.

[0010] In a possible implementation, determining abnormal nodes according to clusters includes: determining a cluster including nodes exceeding a preset number threshold as an abnormal cluster; and determining nodes in the abnormal cluster as abnormal nodes.

[0011] In a possible implementation, after using a graph to construct text and create a knowledge graph, it also includes: obtaining risk case images; determining risk text based on the risk case images; using the risk text to determine risk nodes in the knowledge graph; determining the topological attributes of each node in the knowledge graph; determining the feature vector of the target node based on the topological attributes of the target node; clustering the feature vectors of each node to obtain clusters; and determining abnormal nodes based on the clusters and risk nodes.

[0012] In the second aspect, an embodiment of the present application provides a policy information review device, including: a picture receiving module, used to receive a compensation picture sent by a terminal device; a text determination module, used to determine a text to be queried based on the compensation picture; a node search module, used to search for relevant nodes of the text to be queried in a preset knowledge graph; an information acquisition module, used to obtain the policy information corresponding to the compensation picture if the relevant nodes include a risk node; an information sending module, used to send the policy information to the review terminal.

[0013] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a memory, a processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory, so that the processor executes the first aspect above and / or various possible implementations of the first aspect.

[0014] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the first aspect above and / or various possible implementations of the first aspect.

[0015] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above first aspect and / or various possible implementation methods of the first aspect.

[0016] The insurance policy information review method, device, equipment, storage medium and program product provided in the embodiments of the present application, by receiving the claim picture sent by the terminal device, determine the text to be queried based on the claim picture, query the relevant nodes of the text to be queried in the knowledge graph, and when the relevant nodes include risk nodes, obtain the insurance policy information corresponding to the claim picture, and send the insurance policy information to the review terminal for manual review. Since the information is compared with the text to be queried instead of directly comparing with the picture, fewer computing resources are required. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0018] Figure 1 A schematic diagram of the scenario of the policy information review method provided for this application;

[0019] Figure 2 A flowchart of a method for reviewing insurance policy information provided in an embodiment of the present application;

[0020] Figure 3 A schematic diagram of the structure of the insurance policy information review device provided in an embodiment of the present application;

[0021] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0022] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0023] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0024] This application is used in the scenario of reviewing insurance policy information. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards, and provide corresponding operation portals for users to choose to authorize or refuse.

[0025] Figure 1 A schematic diagram of the scenario of the policy information review method provided for this application. Figure 1 In this scenario, it includes: terminal device 101, server 102 and audit terminal 103.

[0026] In the specific implementation process, the terminal device 101 and the audit terminal 103 may include a computer, a server, a tablet, a mobile phone, a PDA (Personal Digital Assistant) and a notebook, etc., which can input data.

[0027] The server 102 can be implemented using a server or a cluster of multiple servers with more powerful processing capabilities and higher security. If possible, it can also be replaced by a computer, laptop, etc. with stronger computing power.

[0028] The terminal device 101 can submit corresponding pictures to the server 102 during the underwriting and claims settlement stages.

[0029] Server 102 receives the image and obtains the information in the image. After filtering the information, it searches the knowledge graph to see if there is any associated risk node. If there is any associated risk node, it obtains the insurance policy information corresponding to the image and sends the insurance policy information to the review terminal 103, so that the review terminal 103 outputs the insurance policy information.

[0030] It is understandable that the structure illustrated in the embodiment of the present application does not constitute a specific limitation on the method for reviewing insurance policy information. In other feasible implementations of the present application, the above architecture may include more or fewer components than shown in the figure, or combine certain components, or split certain components, or arrange the components differently, which can be determined according to the actual application scenario and is not limited here. Figure 1 The components shown may be implemented by hardware, software, or a combination of software and hardware.

[0031] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0032] Figure 2 A flowchart of the insurance policy information review method provided in the embodiment of the present application. The execution subject of the embodiment of the present application may be Figure 1 The server 102 in the embodiment may also be a computer and / or a mobile phone, etc., which is not particularly limited in this embodiment. Figure 2 As shown, the method includes:

[0033] S201: Receive the compensation picture sent by the terminal device.

[0034] In this step, it can be achieved by receiving messages, data packets, application messages, etc.

[0035] S202: Determine the text to be queried based on the compensation image.

[0036] In this step, it may include using a text recognition model to collect the text to be queried in the compensation image; it may also include obtaining fields in the compensation image and screening the fields to obtain the text to be queried.

[0037] S203: Searching for relevant nodes of the text to be queried in a preset knowledge graph.

[0038] In this step, the process may include writing the text to be queried into a preset query statement template, obtaining a query statement, executing the query statement, and obtaining the relevant nodes of the text to be queried. The process may also include determining the node corresponding to the text to be queried in the preset knowledge graph as the basic node, and determining the node within N hops of the basic node as the relevant node.

[0039] Here, N is greater than or equal to 1, such as 2, 3, 4, etc.

[0040] S204: If the relevant nodes include risk nodes, the insurance policy information corresponding to the compensation picture is obtained.

[0041] In this step, the risk node may be pre-calibrated. Obtaining the insurance policy information corresponding to the compensation picture may include obtaining the insured information, insurance policy logo, claim application, relevant certificates, etc. corresponding to the compensation picture.

[0042] S205: Send the policy information to the review terminal.

[0043] This step may include writing the policy information into a message or a data packet, and sending the message or the data packet to the review terminal.

[0044] From the description of the above embodiments, it can be seen that the embodiments of the present application receive the compensation picture sent by the terminal device, determine the text to be queried based on the compensation picture, query the relevant nodes of the text to be queried in the knowledge graph, and when the relevant nodes include risk nodes, obtain the insurance policy information corresponding to the compensation picture, and send the insurance policy information to the review terminal for manual review. Since the text to be queried is used for information comparison instead of directly comparing the picture, fewer computing resources are required.

[0045] In a possible implementation, in the above step S202, the text to be queried is determined based on the compensation image and includes:

[0046] S2021: Identify text information in compensation images.

[0047] In this step, it may include using a preset text recognition model to recognize the text information in the compensation image; it may also include first using a text type recognition model to recognize the language type in the compensation image, and then using a text recognition model corresponding to the language type to recognize the text information in the compensation image.

[0048] S2022: Perform named entity recognition on the text information to obtain the text type corresponding to each piece of text information.

[0049] In this step, the text information may be input into a pre-trained named entity recognition model to obtain a text type corresponding to each text information output by the named entity recognition model.

[0050] Among them, the named entity recognition model can be implemented using technologies such as recurrent neural networks, long short-term memory networks, gated recurrent units, and convolutional neural networks.

[0051] S2023: Delete text information of a preset text type, and determine the remaining text information as the text to be queried.

[0052] In this step, for example, text information of the vehicle model, automobile manufacturer, and time are deleted, and the remaining text information is determined as the text to be queried.

[0053] From the description of the above embodiments, it can be seen that the embodiments of the present application identify the text information in the compensation image and perform named entity recognition to obtain the text type corresponding to the text information, delete the text information with a preset type, and determine the remaining text information as the text to be queried, thereby filtering the text information and selecting appropriate text information for subsequent information search and other operations, thereby reducing the difficulty of executing subsequent operations and resource usage.

[0054] In a possible implementation, before searching for relevant nodes of the text to be queried in the preset knowledge graph in step S203, the following is further included:

[0055] S220: Acquire at least one training image.

[0056] In this step, the training images may be pre-stored in the database by the staff, and may include images of underwriting, claims settlement, and the like.

[0057] S221: Generate at least one enhanced training image using the target training image.

[0058] In this step, the target training image may be rotated, flipped, translated, cropped, random salt and pepper noise may be added, etc. to obtain an enhanced training image. The enhanced training image may also be generated by deep enhancement, which generates more complex and high-quality images through a self-attention model.

[0059] The target training image may be any training image, and the enhanced training image may refer to an image obtained by data enhancement.

[0060] S222: Determine the target training image and the enhanced training image as a target training image group.

[0061] In this step, for example, if the target training image is A and there are 5 corresponding enhanced training images, then image A and the 5 enhanced training images, a total of 6 images, are jointly determined as the target training image group. For another example, if the target training image is B and there are 10 corresponding enhanced training images, then image B and the 10 enhanced training images, a total of 11 images, are jointly determined as the target training image group.

[0062] S223: Determine the target training text according to the target training image group.

[0063] In this step, the text of each image in the target training image group may be extracted to obtain the target training text.

[0064] S224: De-duplicate the training text to obtain de-duplicate text.

[0065] In this step, the text group can be deduplicated and integrated based on the voting principle by comparing similarity based on character strings and corpora. It can also include calculating the cosine similarity of each training text, retaining a group of training texts with a cosine similarity greater than a preset similarity threshold, and obtaining a deduplicated text.

[0066] Among them, the duplicate removal and integration of text groups based on the voting principle may include obtaining multiple training texts with the same field in the corresponding corpus, voting according to the recognition results of the training texts, and determining the recognized result with a larger number of votes as the text after duplicate removal. For example, currently there are 8 training texts, all corresponding to "automobile factory" in the corpus. Among them, the text recognition results of 5 training texts are "automobile factory", and the text recognition results of the remaining 3 training texts are "automobile wide". Then, "automobile factory" gets 5 votes, which is more, and "automobile factory" is taken as the text after duplicate removal. Another example, currently there are 10 training texts, and the cosine similarity between them is greater than the preset similarity threshold. Among them, the content of 8 training texts is "wood", and the content of 2 training texts is "grain wood". Then, "wood" is taken as the text after duplicate removal.

[0067] S225: Remove the similar-shaped characters in the text after duplicate removal corresponding to each training image group to obtain the text after further duplicate removal.

[0068] In this step, the duplicate removal step in the above step S224 is executed again using the text after duplicate removal corresponding to each training image group, so as to remove the similar-shaped characters of different training image groups and obtain the text after further duplicate removal.

[0069] S226: Remove the invalid information in the text after further duplicate removal to obtain the text for graph construction.

[0070] This step may be similar to the above step S2023. By deleting the text information of the preset type in the text after further duplicate removal, the remaining text after further duplicate removal is determined as the text for graph construction.

[0071] S227: Use the text for graph construction to create a knowledge graph.

[0072] In this step, it may include creating triples using the text for graph construction and creating a knowledge graph using the triples.

[0073] As can be seen from the description of the above embodiments, in the embodiments of the present application, by obtaining training images and performing data augmentation to obtain enhanced training images, determining the target training texts from the target training images and the enhanced training images, removing duplicates from the target training texts to obtain the text after duplicate removal, removing the similar-shaped characters in the text after duplicate removal corresponding to different training images to obtain the text after further duplicate removal, removing the invalid information in the text after further duplicate removal to obtain the text for graph construction, and using the text for graph construction to create a knowledge graph, it is convenient to search for relevant nodes using the knowledge graph subsequently.

[0074] In a possible implementation manner, after using the text for graph construction to create a knowledge graph in the above step S227, it further includes:

[0075] S230: Determine the topological attributes of each node in the knowledge graph.

[0076] In this step, the topological attributes may include HITS (Hyperlink-Induced Topic Search), pagerank, degree centrality, betweenness centrality and other attributes.

[0077] S231: Determine a feature vector of the target node according to the topological attributes of the target node.

[0078] In this step, the topological attributes of the target node may be combined in a preset order to obtain a feature vector of the target node.

[0079] Among them, the target node can be any node in the knowledge graph.

[0080] S232: Cluster the feature vectors of each node to obtain clusters.

[0081] In this step, the K-means algorithm can be used to cluster the feature vectors to obtain clusters.

[0082] S233: Determine abnormal nodes based on clustering clusters.

[0083] In this step, it may include determining, based on the size of the cluster and the density of the cluster, a cluster whose size is greater than a preset size threshold or whose density is greater than a preset density threshold as an abnormal cluster, and determining the nodes in the abnormal cluster as abnormal nodes. Since most of the information is irrelevant, it is possible that each node in the cluster obtained by clustering is abnormal.

[0084] From the description of the above embodiments, it can be seen that the embodiments of the present application determine the topological attributes of the nodes in the knowledge graph, construct feature vectors from the topological attributes, cluster the feature vectors into clusters, and determine the abnormal nodes from the clusters, so that it is possible to subsequently determine whether there are related abnormal nodes in the knowledge graph.

[0085] In a possible implementation, in the above step S233, determining abnormal nodes according to the clustering clusters includes:

[0086] S233A1: Determine a cluster whose number of nodes exceeds a preset number threshold as an abnormal cluster.

[0087] In this step, for example, if the number of nodes included in cluster A is 10 and the preset number threshold is 5, cluster A is determined as an abnormal cluster. For another example, if the number of nodes included in cluster B is 7 and the preset number threshold is 3, cluster B is determined as an abnormal cluster.

[0088] S233A2: Determine the nodes in the abnormal cluster as abnormal nodes.

[0089] In this step, for example, if the abnormal cluster contains nodes a, b, c, and d, nodes a, b, c, and d are all determined as abnormal nodes; for another example, if the abnormal cluster contains nodes e, f, g, h, and i, nodes e, f, g, h, and i are all determined as abnormal nodes.

[0090] From the description of the above embodiments, it can be seen that the embodiments of the present application determine the abnormal clusters in each clustering cluster according to the number of nodes contained, and determine the nodes in the abnormal clusters as abnormal nodes, so as to facilitate the subsequent use of the knowledge graph to determine whether the relevant nodes contain abnormal nodes.

[0091] In a possible implementation, in the above step S233, determining abnormal nodes according to the clustering clusters includes:

[0092] S233B1: If the target cluster includes marked nodes, all nodes in the target cluster are determined as abnormal nodes, wherein the marked nodes are nodes marked by the staff according to the training images.

[0093] In this step, the marked nodes may also be obtained by expanding and recalling the corresponding nodes of the abnormal cases.

[0094] Among them, the method of expanding the recall may include determining all nodes within the 2-degree propagation of the node corresponding to the abnormal case as marked nodes.

[0095] For example, if there are currently marked nodes A, B, and C, and the target cluster includes nodes E, F, and A, then nodes E, F, and A are all determined as abnormal nodes. For another example, if there are currently marked nodes G and H, and the target cluster includes nodes G, H, I, P, Q, and O, then nodes G, H, I, P, Q, and O are all determined as abnormal nodes.

[0096] From the description of the above embodiments, it can be seen that the embodiments of the present application determine all nodes in the same cluster as the marked nodes as abnormal nodes by combining the marked nodes pre-marked by the staff, thereby realizing accurate marking of abnormal nodes in combination with manual judgment, and facilitating the subsequent use of the knowledge graph to determine whether abnormal nodes are included in related nodes.

[0097] In a possible implementation, after the above step S227 uses the graph to construct text and creates the knowledge graph, it also includes:

[0098] S240: Obtain risk case images.

[0099] In this step, the risk case images can be images obtained through image duplication checking and anti-tampering model identification, or they can be images pre-marked by staff.

[0100] S241: Determine the risk text based on the risk case picture.

[0101] This step is similar to the above-mentioned steps S220 to S226, and the risk text is obtained by text recognition, deduplication, and removal of invalid information.

[0102] S242: Use risk text to determine risk nodes in the knowledge graph.

[0103] In this step, it may include searching for nodes corresponding to the risk text in the knowledge graph to obtain risk nodes; it may also include converting the risk text into a triple or a vector, and determining the node corresponding to the triple or the vector in the knowledge graph as a risk node; it also includes determining the node corresponding to the risk text in the knowledge graph as a basic node, and determining the node within 2 degrees of propagation (or 2 hops) of the basic node as a risk node.

[0104] S243: Determine the topological attributes of each node in the knowledge graph.

[0105] This step is similar to the above step S230 and will not be repeated here.

[0106] S244: Determine a feature vector of the target node according to the topological attributes of the target node.

[0107] In this step, this step is similar to the above step S231 and will not be repeated here.

[0108] S245: Cluster the feature vectors of each node to obtain clusters.

[0109] This step is similar to the above step S232 and will not be repeated here.

[0110] S246: Determine abnormal nodes based on clusters and risk nodes.

[0111] In this step, the cluster containing the risk node may be determined as an abnormal cluster, and the nodes in the abnormal cluster may be determined as abnormal nodes.

[0112] From the description of the above embodiments, it can be seen that the embodiments of the present application obtain risk case pictures, use the risk case pictures to determine the risk text, find the risk nodes in the knowledge graph from the risk text, cluster the nodes in the knowledge graph, and then determine the nodes in the cluster containing the risk nodes as abnormal nodes, thereby marking the abnormal nodes in the knowledge graph, which facilitates the subsequent use of the knowledge graph to determine whether the policy information corresponding to the payment picture needs manual review again.

[0113] In a possible implementation, at the stage of creating a knowledge graph, a group of images (including M images) from the underwriting and claims process is used, and data enhancement is performed on each image to generate N images. The texts in these images are extracted to obtain X groups of texts. These texts are compared for similarity of strings and corpora, and the voting principle is used to remove duplicates, and each group of images obtains Y groups of texts. Since it is possible that area A in image 1 is clear and area B is not clear, and area B in image 2 is clear and area A is not clear, it is very likely that the OCR recognition process will be incorrectly recognized. Therefore, the M×Y groups of texts are compared for similarity, and the text groups that are incorrectly recognized are eliminated. At the same time, the repetition and ambiguity of the same text content are reduced to obtain L text groups. Named entity recognition is performed on these L text groups, and the preset type of text groups in the text groups are removed by the results of named entity recognition to obtain graph creation text, and the graph creation text is used to create a knowledge graph, so that the vehicle and case information in the claims stage are integrated into the knowledge graph as attributes. The knowledge graph is stored in the database, and expanded on the basis of the nodes corresponding to the risk cases in the knowledge graph (including determining all nodes within two hops of the original risk case corresponding nodes as abnormal nodes, clustering the nodes, determining clusters with more nodes and clusters with abnormal nodes as abnormal clusters, and determining nodes in abnormal clusters as abnormal nodes) to obtain more abnormal nodes. After receiving a new picture, the knowledge graph is searched according to the text in the picture. If there is an abnormal node within 3 degrees of the node corresponding to the picture, the insurance policy information corresponding to the picture is sent to the review terminal, or directly intercepted. After the new picture is reviewed, the information corresponding to the new picture can also be added to the knowledge graph as a new node.

[0114] Figure 3 This is a schematic diagram of the structure of the insurance policy information review device provided in the embodiment of the present application. Figure 3 As shown, the insurance policy information review device 300 includes: a picture receiving module 301, a text determination module 302, a node search module 303, an information acquisition module 304 and an information sending module 305.

[0115] The picture receiving module 301 is used to receive the compensation picture sent by the terminal device;

[0116] The text determination module 302 is used to determine the text to be queried based on the compensation picture;

[0117] The node search module 303 is used to search for relevant nodes of the text to be queried in a preset knowledge graph;

[0118] The information acquisition module 304 is used to acquire the insurance policy information corresponding to the compensation picture if the relevant node includes a risk node;

[0119] The information sending module 305 is used to send the insurance policy information to the review terminal.

[0120] The device provided in this embodiment can be used to execute the technical solution of the above method embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be repeated here.

[0121] In one possible implementation, the text determination module 302 is used to identify text information in the compensation image; perform named entity recognition on the text information to obtain the text type corresponding to each text information; delete the text information with a preset text type, and determine the remaining text information as the text to be queried.

[0122] The device provided in this embodiment can be used to execute the technical solution of the above method embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be repeated here.

[0123] In a possible implementation, the insurance policy information review device 300 further includes: a graph creation module 306.

[0124] The graph creation module 306 is used to obtain at least one training image; use the target training image to generate at least one enhanced training image; determine the target training image and the enhanced training image as a target training image group; determine the target training text based on the target training image group; deduplicate the training text to obtain deduplicated text; remove similar characters in the deduplicated text corresponding to each training image group to obtain a further deduplicated text; remove invalid information in the further deduplicated text to obtain a graph construction text; use the graph construction text to create a knowledge graph.

[0125] The device provided in this embodiment can be used to execute the technical solution of the above method embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be repeated here.

[0126] In a possible implementation, the insurance policy information review device 300 further includes: an abnormality determination module 307.

[0127] The abnormality determination module 307 is used to determine the topological attributes of each node in the knowledge graph; determine the feature vector of the target node according to the topological attributes of the target node; cluster the feature vectors of each node to obtain clusters; and determine the abnormal nodes according to the clusters.

[0128] The device provided in this embodiment can be used to execute the technical solution of the above method embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be repeated here.

[0129] In a possible implementation, the abnormality determination module 307 is specifically configured to determine a cluster whose number of nodes exceeds a preset number threshold as an abnormal cluster; and determine a node in the abnormal cluster as an abnormal node.

[0130] The device provided in this embodiment can be used to execute the technical solution of the above method embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be repeated here.

[0131] In a possible implementation, the anomaly determination module 307 is also used to obtain risk case images; determine risk texts based on the risk case images; use the risk texts to determine risk nodes in the knowledge graph; determine the topological properties of each node in the knowledge graph; determine the feature vector of the target node based on the topological properties of the target node; cluster the feature vectors of each node to obtain clusters; and determine abnormal nodes based on the clusters and risk nodes.

[0132] The device provided in this embodiment can be used to execute the technical solution of the above method embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be repeated here.

[0133] In order to implement the above embodiment, the embodiment of the present application also provides an electronic device.

[0134] refer to Figure 4 , which shows a schematic diagram of the structure of an electronic device 400 suitable for implementing the embodiment of the present application, and the electronic device 400 may be a terminal device or a server. The terminal device may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, personal digital assistants (PDAs), tablet computers (Portable Android Devices, PADs), portable multimedia players (Portable Media Players, PMPs), vehicle terminals (such as vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 4 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0135] like Figure 4As shown, the electronic device 400 may include a processor (such as a central processing unit, a graphics processing unit, etc.) 401, and a memory 402 that is communicatively connected to the processor, which can perform various appropriate actions and processes according to the program stored in the memory 402, the computer execution instruction, or the program loaded from the storage device 408 to the random access memory (Random Access Memory, referred to as RAM) 403, to implement the insurance policy information review method in any of the above embodiments, wherein the memory may be a read-only memory (Read Only Memory, referred to as ROM). In the RAM 403, various programs and data required for the operation of the electronic device 400 are also stored. The processing device 401, the memory 402, and the RAM 403 are connected to each other via a bus 404. The input / output (I / O) interface 405 is also connected to the bus 404.

[0136] Typically, the following devices may be connected to the I / O interface 405: an input device 406 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 407 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 408 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 409. The communication device 409 may allow the electronic device 400 to communicate with other devices wirelessly or by wire to exchange data. Figure 4 The electronic device 400 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0137] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 409, or installed from the storage device 408, or installed from the memory 402. When the computer program is executed by the processing device 401, the above-mentioned functions defined in the method of the embodiment of the present application are executed.

[0138] It should be noted that the computer-readable storage medium mentioned above in the present application may be a computer-readable signal medium or a computer storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable storage medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0139] The computer-readable storage medium may be included in the electronic device, or may exist independently without being installed in the electronic device.

[0140] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

[0141] The computer program code for performing the operation of the present application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0142] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0143] The modules involved in the embodiments described in this application can be implemented by software or hardware. The name of the unit does not limit the module itself in some cases. For example, the picture receiving module can also be described as a "compensation picture receiving module".

[0144] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0145] The present application also provides a computer-readable storage medium, which stores computer execution instructions. When the processor executes the computer execution instructions, the technical solution of the insurance policy information review method in any of the above-mentioned embodiments is implemented. The implementation principle and beneficial effects are similar to the implementation principle and beneficial effects of the insurance policy information review method. Please refer to the implementation principle and beneficial effects of the insurance policy information review method, which will not be repeated here.

[0146] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0147] The present application also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the technical solution of the insurance policy information review method in any of the above-mentioned embodiments. Its implementation principle and beneficial effects are similar to the implementation principle and beneficial effects of the insurance policy information review method. Please refer to the implementation principle and beneficial effects of the insurance policy information review method, and no further details will be given here.

[0148] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

[0149] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0150] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for reviewing insurance policy information, characterized in that: include: Receive compensation pictures sent by terminal devices; Determine the text to be queried based on the compensation picture; In a preset knowledge graph, search for nodes related to the text to be queried; If the relevant nodes include risk nodes, obtaining the insurance policy information corresponding to the compensation picture; The insurance policy information is sent to the review terminal.

2. The method according to claim 1, characterized in that Determining the text to be queried according to the compensation picture includes: Recognizing text information in the compensation image; Performing named entity recognition on the text information to obtain the text type corresponding to each text information; The text information of the preset text type is deleted, and the remaining text information is determined as the text to be queried.

3. The method according to claim 1 or 2, characterized in that: Before searching for the relevant nodes of the text to be queried in the preset knowledge graph, the method further includes: acquiring at least one training image; generating at least one enhanced training image using the target training image; Determine the target training image and the enhanced training image as a target training image group; Determining target training text according to the target training image group; Deduplicating the training text to obtain deduplicated text; Removing similar characters from the deduplicated text corresponding to each training image group to obtain a re-deduplicated text; Removing invalid information from the deduplicated text to obtain a graph construction text; The graph is used to construct text and create the knowledge graph.

4. The method according to claim 3, characterized in that After the graph is used to construct the text and the knowledge graph is created, the method further includes: Determine the topological attributes of each node in the knowledge graph; Determining a characteristic vector of the target node according to a topological attribute of the target node; Cluster the feature vectors of each node to obtain clusters; According to the clusters, abnormal nodes are determined.

5. The method according to claim 4, characterized in that The step of determining abnormal nodes according to the clustering clusters includes: Determine a cluster whose number of nodes exceeds a preset number threshold as an abnormal cluster; The nodes in the abnormal cluster are determined as abnormal nodes.

6. The method according to claim 3, characterized in that: After the graph is used to construct the text and the knowledge graph is created, the method further includes: Get risk case images; Determine the risk text according to the risk case picture; Using the risk text, determining risk nodes in the knowledge graph; Determine the topological attributes of each node in the knowledge graph; Determining a characteristic vector of the target node according to a topological attribute of the target node; Cluster the feature vectors of each node to obtain clusters; An abnormal node is determined according to the clusters and the risk nodes.

7. A device for verifying insurance policy information, characterized in that: include: The picture receiving module is used to receive the compensation picture sent by the terminal device; A text determination module, used to determine the text to be queried based on the compensation picture; A node search module is used to search for relevant nodes of the text to be queried in a preset knowledge graph; An information acquisition module, configured to acquire the insurance policy information corresponding to the compensation picture if the relevant node includes a risk node; The information sending module is used to send the insurance policy information to the review terminal.

8. An electronic device, characterized in that: include: Memory, processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor executes the insurance policy information review method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the insurance policy information review method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that It comprises a computer program which, when executed by a processor, implements the insurance policy information review method as described in any one of claims 1 to 6.