Quality control method and device based on insurance claim settlement, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a quality control method and device based on insurance claim settlement, electronic equipment and a storage medium, belongs to the technical field of computers, and is suitable for the field of financial science and technology. The method comprises the steps that insurance claim settlement materials and the claim settlement number of the insurance claim settlement materials are acquired; determining at least two sequence number intervals according to a pre-acquired quality control concurrency number and a pre-acquired claim settlement number; the serial number intervals are obtained according to the interval arrangement sequence between the serial number intervals, a target interval is obtained, and the target interval comprises an interval right end point; the insurance claim settlement materials are screened through an asynchronous thread, a reference claim settlement material with a material sequence number and a target material category is obtained, and the material sequence number is located in a target interval; according to the material sequence number, the quality control model is called through the target material category and the asynchronous thread to sequentially carry out quality detection on the reference claim settlement material until the material sequence number is equal to the right end point of the interval, and a quality control result of quality detection is obtained, so that the quality control efficiency of the insurance claim settlement material can be improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, specifically to the field of financial technology, and particularly to a quality control method, apparatus, electronic device, and storage medium based on insurance claims. Background Technology

[0002] When users file an insurance claim with an insurance company, they upload supporting documents such as their ID card photo page and invoices through the service interface provided by the insurance company. To ensure the accuracy of these documents, quality control is required, such as checking the clarity of the ID card photo page. As the scale of insurance business expands, insurance companies receive a large number of claim documents. However, the traditional method of manual review for quality control of these documents results in a long quality control cycle and low efficiency. Summary of the Invention

[0003] The main objective of this application is to propose a quality control method, device, electronic device, and storage medium based on insurance claims, aiming to improve the quality control efficiency of insurance claims materials.

[0004] To achieve the above objectives, a first aspect of this application proposes a quality control method based on insurance claims, the method comprising:

[0005] Obtain insurance claim materials and the number of claims for which the materials are required;

[0006] Based on the pre-obtained concurrent quality control data and the number of claims processed, at least two sequence number ranges are determined;

[0007] The sequence number intervals are obtained according to the interval arrangement order between the sequence number intervals to obtain the target interval; wherein, the target interval includes the right endpoint of the interval;

[0008] The insurance claim materials are filtered through a preset asynchronous thread to obtain reference claim materials with material serial numbers and target material categories; wherein the material serial numbers are located within the target range.

[0009] According to the material serial number, the quality control model is invoked sequentially through the target material category and the asynchronous thread to perform quality inspection on the reference claim materials until the material serial number is equal to the right endpoint of the interval, and the quality control result of the quality inspection is obtained; wherein, the quality control result includes quality inspection passed or quality inspection failed.

[0010] In some embodiments, obtaining insurance claim materials and the number of claims for which the insurance claim materials are obtained includes:

[0011] Obtain the original claim materials;

[0012] The original claim materials are classified to obtain the predicted material categories;

[0013] Based on the predicted material category, the original claim materials are preliminarily inspected to obtain the material quality category;

[0014] The original claim materials are screened according to the predicted material category and the material quality category to obtain the insurance claim materials;

[0015] Obtain the number of claims for the insurance claim materials.

[0016] In some embodiments, classifying the original claim materials to obtain predicted material categories includes:

[0017] Feature extraction is performed on the original claim materials to obtain first modality claim features and second modality claim features;

[0018] Cross-attention calculation is performed on the first modality claim feature and the second modality claim feature to obtain attention features;

[0019] The first modal claim feature, the second modal claim feature, and the attention feature are fused to obtain the reference claim feature;

[0020] The original claim materials are classified according to the reference claim characteristics to obtain the predicted material category.

[0021] In some embodiments, the step of performing cross-attention calculation on the first modal claim feature and the second modal claim feature to obtain attention features includes:

[0022] The first modality of claims features are used for feature mapping to obtain query features;

[0023] Feature mapping is performed on the second modality claims features to obtain key features and value features;

[0024] Calculate the attention score based on the query features and the key features;

[0025] The attention features are obtained by weighted summation of the value features based on the attention scores.

[0026] In some embodiments, the feature fusion of the first modal claim feature, the second modal claim feature, and the attention feature to obtain the reference claim feature includes:

[0027] Calculate the gating weight value based on the attention characteristics;

[0028] Based on the gating weight value, the first modal claims feature and the attention feature are fused to obtain the first fused feature;

[0029] Based on the gating weight value, the second modality claims feature and the attention feature are fused to obtain the second fused feature;

[0030] The first fusion feature and the second fusion feature are fused to obtain the reference claim feature.

[0031] In some embodiments, the feature fusion of the first fusion feature and the second fusion feature to obtain the reference claim feature includes:

[0032] Calculate the similarity between the first fused feature and the second fused feature;

[0033] Based on the similarity, principal component analysis is performed on the first fusion feature to obtain the first key feature;

[0034] Principal component analysis was performed on the second fusion feature to obtain the second key feature;

[0035] The first key feature and the second key feature are fused to obtain the reference claim feature.

[0036] In some embodiments, the preliminary inspection of the original claim materials based on the predicted material category to obtain the material quality category includes:

[0037] Optical character recognition is performed on the original claim materials to obtain a character sequence;

[0038] The character sequence is subjected to keyword recognition to obtain the target keywords;

[0039] Calculate the quality score of the original claim materials based on the predicted material category and the target keywords;

[0040] The original claim materials are classified according to the quality score to obtain the material quality category.

[0041] To achieve the above objectives, a second aspect of this application provides a quality control device based on insurance claims, the device comprising:

[0042] The material acquisition module is used to acquire insurance claim materials and the quantity of the insurance claim materials required for the claim.

[0043] The interval determination module is used to determine at least two sequence intervals based on the pre-acquired concurrent quality control number and the number of claims.

[0044] An interval acquisition module is used to acquire the numbered intervals according to the interval arrangement order between the numbered intervals to obtain a target interval; wherein, the target interval includes the right endpoint of the interval;

[0045] The filtering module is used to filter the insurance claim materials through a preset asynchronous thread to obtain reference claim materials with material serial numbers and target material categories; wherein the material serial numbers are located in the target range;

[0046] The quality control module is used to sequentially perform quality inspections on the reference claims materials according to the material serial number, through the target material category and the asynchronous thread, by calling the quality control model until the material serial number is equal to the right endpoint of the interval, and obtain the quality control result of the quality inspection; wherein, the quality control result includes quality inspection passed or quality inspection failed.

[0047] To achieve the above objectives, a third aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0048] To achieve the above objectives, a fourth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect.

[0049] The present application proposes a quality control method, device, electronic device, and computer-readable storage medium based on insurance claims. This method acquires insurance claim materials and the number of claims for each material, enabling quality control of the materials based on the claim volume. When the number of claims is high, to improve quality control efficiency, at least two sequence number intervals are determined based on pre-acquired concurrent quality control and claim volume. This allows for the allocation of insurance claim materials to different sequence number intervals, resulting in more rational workload distribution and improved resource utilization. Sequence number intervals are obtained according to their arrangement order to obtain a target interval for quality control of insurance claim materials located within that interval. To further improve efficiency, an asynchronous thread is initiated, and the insurance claim materials are filtered using a preset asynchronous thread to obtain reference claim materials with material serial numbers and target material categories. The material serial numbers are located within the target interval to determine the insurance claim materials that the asynchronous thread needs to control. To automate the quality control process, the quality control model is called sequentially according to the material serial number, through the target material category and asynchronous thread, to perform quality inspection on the reference claim materials until the material serial number is equal to the right endpoint of the interval. This yields the quality control results of the insurance claim materials whose material serial numbers are located in the target interval, which greatly improves the quality control efficiency compared to manual quality control methods. Attached Figure Description

[0050] Figure 1 This is a flowchart of a quality control method based on insurance claims provided in an embodiment of this application;

[0051] Figure 2 yes Figure 1 The flowchart of step S110 in the middle;

[0052] Figure 3 yes Figure 2 The flowchart of step S220 in the text;

[0053] Figure 4 yes Figure 3 The flowchart of step S320 in the middle;

[0054] Figure 5 yes Figure 3 The flowchart of step S330 in the text;

[0055] Figure 6 yes Figure 5 The flowchart of step S540 in the text;

[0056] Figure 7 yes Figure 2 The flowchart of step S230 in the text;

[0057] Figure 8 This is a schematic diagram of the structure of the quality control device based on insurance claims provided in the embodiments of this application;

[0058] Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0060] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0061] 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 herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0062] When users file an insurance claim with an insurance company, they upload supporting documents such as their ID card photo page and invoices through the service interface provided by the insurance company. To ensure the accuracy of these documents, quality control is required, such as checking the clarity of the ID card photo page. As the scale of insurance business expands, insurance companies receive a large number of claim documents. However, the traditional method of manual review for quality control of these documents results in a long quality control cycle and low efficiency.

[0063] Based on this, embodiments of this application provide a quality control method, a quality control device, an electronic device, and a computer-readable storage medium based on insurance claims, aiming to improve the quality control efficiency of insurance claims materials.

[0064] The quality control method, device, electronic device, and computer-readable storage medium based on insurance claims provided in this application are specifically described through the following embodiments. First, the quality control method based on insurance claims in this application embodiment is described.

[0065] The quality control method based on insurance claims provided in this application relates to the field of computer technology. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the quality control method based on insurance claims, but is not limited to the above forms.

[0066] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0067] Figure 1 This is an optional flowchart of the quality control method based on insurance claims provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S110 to S150.

[0068] Step S110: Obtain insurance claim materials and the number of claims required for the insurance claim.

[0069] Step S120: Determine at least two sequence number ranges based on the pre-acquired concurrent quality control and claim numbers;

[0070] Step S130: Obtain the sequence number intervals according to the interval arrangement order between the sequence number intervals to obtain the target interval; wherein, the target interval includes the right endpoint of the interval;

[0071] Step S140: The insurance claim materials are filtered through a preset asynchronous thread to obtain reference claim materials with material serial numbers and target material categories; wherein, the material serial numbers are located in the target range;

[0072] Step S150: According to the material serial number, the quality control model is called sequentially through the target material category and asynchronous thread to perform quality inspection on the reference claim materials until the material serial number is equal to the right endpoint of the interval, and the quality control result of the quality inspection is obtained; wherein, the quality control result includes whether the quality inspection passed or failed.

[0073] Steps S110 to S150, as shown in the embodiments of this application, automate the quality control of insurance claim materials. The entire quality control process does not require manual intervention, thus improving the efficiency of quality control in insurance claims.

[0074] Please see Figure 2In some embodiments, step S110 may include, but is not limited to, steps S210 to S250:

[0075] Step S210: Obtain the original claim materials;

[0076] Step S220: Classify the original claim materials to obtain the predicted material categories;

[0077] Step S230: Conduct a preliminary inspection of the original claim materials based on the predicted material category to obtain the material quality category;

[0078] Step S240: Screen the original claim materials according to the predicted material category and material quality category to obtain the insurance claim materials;

[0079] Step S250: Obtain the number of insurance claim materials required for the claim.

[0080] In step S210 of some embodiments, when a user needs to apply for insurance claims, they upload various supporting documents through the business service interface provided by the insurance institution to obtain the original claim materials. The original claim materials are related to the type of insurance to be claimed. For example, if the insurance type is health insurance, the original claim materials include identity documents, medical diagnosis results, medical records, medical expense invoices, etc. Identity documents can be ID cards or household registration books. As another example, if the insurance type is car insurance, the original claim materials include identity documents, accident scene photos, vehicle repair invoices, accident liability determination letters, etc.

[0081] In step S220 of some embodiments, the original claim materials include various supporting documents. Different supporting documents have different quality control standards. For example, the front of an ID card needs to contain key information such as name and ID number, while an invoice needs to contain information such as receipt and amount. To improve the accuracy of insurance claim quality control, it is necessary to determine the material type of the supporting documents so that subsequent quality control can be performed based on the material type. Specifically, the original claim materials are classified using a classification model, assigning them to predefined labels to obtain predicted material categories. The classification model can be a convolutional neural network, a recurrent neural network, a transformer model, etc. Predefined labels include front and back of ID card, household registration book, invoice, document, and others. Medical records, medical diagnosis results, and accident liability determination letters are all labeled as "documents." The original claim materials include multiple supporting documents, and the predicted material category includes the material type of each supporting document. The predicted material category includes at least one of the following: front and back of ID card, household registration book, invoice, document, and others.

[0082] In step S230 of some embodiments, the original claim materials include multiple supporting documents, and the predicted material category includes the material type of each supporting document. For each supporting document, a preliminary inspection is performed based on the material type to obtain the inspection result. If the inspection result for each supporting document is "passed," the material quality category of the original claim materials is determined to be "qualified." If the inspection result for any supporting document is "failed," the material quality category of the original claim materials is determined to be "unqualified."

[0083] In step S240 of some embodiments, the preset material category refers to the material types of various supporting documents that should be uploaded for the type of insurance claim to be processed. The predicted material category is compared with the preset material category. If the predicted material category includes all material types of the preset material category and the material quality category is "quality meets standards," then the original claim materials are used as insurance claim materials. If the predicted material category lacks any material types of the preset material category, it indicates that the user's uploaded original claim materials are missing a certain supporting document, or the predicted material category is "quality does not meet standards." In this case, a prompt message is sent to the user through the business service interface, prompting the user to re-upload the missing supporting document or the supporting document that failed the inspection, without requiring subsequent quality control processes, thus improving the quality control efficiency of insurance claims.

[0084] In step S250 of some embodiments, the number of insurance claim materials is obtained to determine the claim quantity. It should be noted that multiple supporting documents for the same insurance incident constitute one insurance claim material, and the claim quantity is 1.

[0085] Through steps S210 to S250, insurance claim materials that pass preliminary inspection can be screened from a large number of claim materials, reducing the number of claims that need to be processed for subsequent quality control and improving the efficiency of insurance claim quality control.

[0086] Please see Figure 3 In some embodiments, step S220 may include, but is not limited to, steps S310 to S340:

[0087] Step S310: Extract features from the original claim materials to obtain the first modality claim features and the second modality claim features;

[0088] Step S320: Perform cross-attention calculation on the first modality claims features and the second modality claims features to obtain attention features;

[0089] Step S330: Perform feature fusion on the first modality claims features, the second modality claims features, and the attention features to obtain reference claims features;

[0090] Step S340: Classify the original claim materials according to the reference claim characteristics to obtain the predicted material categories.

[0091] In step S310 of some embodiments, the classification model includes multiple feature extraction layers and classification layers connected in series. Features are extracted from the original claim materials using the first feature extraction layer to obtain the claim features output by the first feature extraction layer. The claim features output by the previous feature extraction layer are input into the current feature extraction layer for feature extraction to obtain the claim features output by the current feature extraction layer. This process is repeated until the claim features output by the last feature extraction layer are obtained. The claim features output by any two feature extraction layers are used as the first modality claim features and the second modality claim features. The first modality claim features and the second modality claim features are claim features of different modalities extracted by different feature extraction layers.

[0092] In step S320 of some embodiments, in order to capture the complex dependency relationship between the first modality claims feature and the second modality claims feature, and to better understand the mutual influence of features between different feature extraction layers, cross-attention calculation is performed on the first modality claims feature and the second modality claims feature of each pair of feature extraction layers to obtain multiple attention features.

[0093] In step S330 of some embodiments, in order to reduce information loss, enhance feature representation ability and the generalization ability of the classification model, the first modality claims feature, the second modality claims feature and multiple attention features are concatenated or added to obtain reference claims features.

[0094] In step S340 of some embodiments, reference claim features are input to the classification layer to classify the original claim materials and obtain the predicted material category. The classification layer may be a fully connected layer, a multilayer perceptron, etc.

[0095] Through the above steps S310 to S340, the predicted material category can be obtained, and the original claim materials can be judged based on the predicted material category to determine whether there are missing supporting materials. This achieves the preliminary detection and screening of the original claim materials, thereby improving the quality control efficiency of insurance claims.

[0096] Please see Figure 4 In some embodiments, step S320 may include, but is not limited to, steps S410 to S440:

[0097] Step S410: Perform feature mapping on the first modality claims features to obtain query features;

[0098] Step S420: Perform feature mapping on the second modality claims features to obtain key features and value features;

[0099] Step S430: Calculate the attention score based on the query features and key features;

[0100] Step S440: The attention features are obtained by weighted summation of the value features based on the attention scores.

[0101] In step S410 of some embodiments, the query weight and the first modality claims feature are multiplied to obtain the query feature.

[0102] In step S420 of some embodiments, the key weights are multiplied by the second modal claims feature to obtain the key feature. The value weights are multiplied by the second modal claims feature to obtain the value feature.

[0103] In step S430 of some embodiments, a dot product is calculated between the query feature and the key feature to obtain the similarity between them. A scaling factor is determined based on the feature dimension of the key feature, and the similarity is scaled according to the scaling factor. The scaled similarity is then normalized using the softmax function to obtain the attention score. The formula for calculating the attention score is as follows:

[0104]

[0105] Where Q represents the query feature; K represents the key feature; T represents the transpose operation; and d represents the feature dimension of the key feature. This represents the scaling factor.

[0106] In step S440 of some embodiments, the attention score and the value feature are multiplied to obtain the attention feature. If there are multiple attention heads, the attention score output by each attention head can be multiplied by the value feature, and the results of multiple multiplications can be added together to obtain the attention feature.

[0107] Each pair of feature extraction layers influences the others, not unidirectionally. To ensure the accuracy of attention feature extraction, the query weight is multiplied by the second modality claims feature to obtain the intermediate query feature. The key weight is multiplied by the first modality claims feature to obtain the intermediate key feature, and the value weight is multiplied by the first modality claims feature to obtain the intermediate value feature. An intermediate attention score is calculated based on the intermediate query and key features. The intermediate value feature is then weighted and summed based on the intermediate attention score to obtain the intermediate attention feature. Finally, the average of the attention feature and the intermediate attention feature is calculated to obtain the final attention feature.

[0108] If the attention feature is represented as h1 and the intermediate attention feature is represented as h2, then the final attention feature is represented as (h1+h2) / 2.

[0109] Through the above steps S410 to S440, the complex dependencies between different modal features can be captured, providing a more comprehensive feature representation, thereby improving the accuracy of claims material classification.

[0110] Please see Figure 5 In some embodiments, step S330 may include, but is not limited to, steps S510 to S540:

[0111] Step S510: Calculate the gating weight value based on the attention characteristics;

[0112] Step S520: Based on the gating weight value, perform feature fusion on the first modality claims feature and attention feature to obtain the first fused feature;

[0113] Step S530: Based on the gating weight value, the second modality claims feature and attention feature are fused to obtain the second fused feature;

[0114] Step S540: Perform feature fusion on the first fusion feature and the second fusion feature to obtain the reference claim feature.

[0115] In step S510 of some embodiments, for each attention feature, feature mapping is performed on the attention feature to obtain a gating weight value, the value of which ranges from [0,1]. The gating weight value allows dynamic control over the retention or suppression of feature dimensions, enhancing the fine-grained feature selection capability of the classification model. The calculation process of feature mapping is expressed as follows:

[0116] g = σ(WH + b),

[0117] Where g is the gating weight; σ represents the activation function, which can be the sigmoid function; W and b are the projection weight and bias, respectively; and H is the attention feature.

[0118] In step S520 of some embodiments, in order to achieve effective feature fusion without increasing computational complexity, the first modality claims feature and the attention feature are added together to obtain the first modality feature. This incorporates the contextual information introduced by the attention feature into the first modality claims feature, thereby enhancing the feature representation capability and enabling the classification model to better capture global and local information, improving its understanding of complex insurance claims materials. The gating weight value is then multiplied by the first modality feature to dynamically adjust the feature importance of the first modality feature based on the gating weight value, resulting in the first fused feature.

[0119] In step S530 of some embodiments, the second modality claims feature and the attention feature are added together to incorporate the contextual information introduced by the attention feature into the second modality claims feature, thus obtaining the second modality feature. The gating weight value and the second modality feature are multiplied together to dynamically adjust the feature importance of the second modality feature according to the gating weight value, thus obtaining the second fused feature.

[0120] In step S540 of some embodiments, the first fusion feature and the second fusion feature calculated based on each attention feature are concatenated or added together to obtain the reference claim feature.

[0121] Steps S510 to S540 above utilize a gating mechanism to achieve fine-grained feature selection, reducing unnecessary feature computation. Introducing contextual information through an attention mechanism enhances feature representation capabilities and improves the model's adaptability and generalization ability.

[0122] Please see Figure 6 In some embodiments, step 540 may include, but is not limited to, steps S610 to S640:

[0123] Step S610: Calculate the similarity between the first fusion feature and the second fusion feature;

[0124] Step S620: Perform principal component analysis on the first fusion feature based on similarity to obtain the first key feature;

[0125] Step S630: Perform principal component analysis on the second fusion feature to obtain the second key feature;

[0126] Step S640: Perform feature fusion on the first key feature and the second key feature to obtain reference claim features.

[0127] In step S610 of some embodiments, the Euclidean distance between the first fused feature and the second fused feature is calculated, and the reciprocal of the Euclidean distance is used as the similarity. The larger the Euclidean distance, the smaller the similarity between the first fused feature and the second fused feature.

[0128] In step S620 of some embodiments, if the similarity is greater than or equal to a preset similarity threshold, it indicates that the first fused feature and the second fused feature are quite similar and have high feature redundancy. Then, Principal Component Analysis (PCA) is used to perform PCA on the first fused feature to identify the main features and obtain the first key feature. The preset similarity threshold is a pre-set threshold used to determine whether features are similar; it can be set according to actual conditions, such as 0.8. If the similarity is less than the preset similarity threshold, it indicates that the similarity between the first fused feature and the second fused feature is small. Then, feature fusion is performed on the first fused feature and the second fused feature to obtain a reference claim feature.

[0129] Specifically, the first fused feature is divided into multiple row vectors or column vectors, resulting in multiple vectors. Each vector is standardized to have a mean of 0 and a variance of 1, resulting in multiple base vectors. Based on the positions of the base vectors in the first fused feature, a matrix is ​​constructed from these base vectors to obtain a base matrix. The covariance matrix of the base matrix is ​​calculated, and the eigenvalues ​​and eigenvectors of the covariance matrix are solved. The eigenvectors with the largest eigenvalues ​​are selected to obtain the projection matrix. The projection matrix is ​​multiplied by the first fused feature to obtain the first key feature.

[0130] In step S630 of some embodiments, referring to step S620, principal component analysis is performed on the second fusion feature using a principal component analysis algorithm to extract the main features of the second fusion feature and obtain the second key feature.

[0131] In step S640 of some embodiments, the first key feature and the second key feature calculated based on each attention feature are concatenated or added together to obtain the reference claim feature.

[0132] Through steps S610 to S640, key claim features can be identified, reducing the complexity of feature calculation and thus improving the efficiency of insurance claim material classification.

[0133] Please see Figure 7 In some embodiments, step S230 may include, but is not limited to, steps S710 to S740:

[0134] Step S710: Perform optical character recognition on the original claim materials to obtain a character sequence;

[0135] Step S720: Perform keyword recognition on the character sequence to obtain the target keywords;

[0136] Step S730: Calculate the quality score of the original claim materials based on the predicted material category and target keywords;

[0137] Step S740: Classify the original claim materials according to the quality score to obtain the material quality category.

[0138] In step S710 of some embodiments, the original claim materials are subjected to optical character recognition by an optical character recognition (OCR) engine to extract the text content of the original claim materials and convert the text content into a text format that can be processed by a computer to obtain a character sequence.

[0139] In step S720 of some embodiments, keyword recognition is performed on the character sequence using a pre-trained language model such as BERT to extract keywords from the character sequence and obtain target keywords.

[0140] In step S730 of some embodiments, the predicted material category includes the material type of each supporting document, and the category weight of the material type is obtained. The category weight is used to measure the importance of the material type to the quality score; the larger the category weight, the greater the importance. The value of the category weight ranges from [0,1]. Reference keywords and their keyword weights corresponding to the material type are obtained. The keyword weight is used to measure the importance of the keyword to the quality score; the keyword weight is proportional to the importance, and the value of the keyword weight ranges from [0,1]. The target keyword and the reference keyword are compared. If the target keyword and the reference keyword are the same, the category weight and the keyword weight are multiplied to obtain the keyword score of the target keyword. If the target keyword and the reference keyword are different, the keyword score of the target keyword is 0. The keyword scores of the target keywords are added together to obtain the quality score of the original claim material.

[0141] In step S740 of some embodiments, if the quality score is greater than or equal to a preset scoring threshold, the material quality category of the original claim materials is determined to be compliant. If the quality score is less than the preset scoring threshold, the material quality category of the original claim materials is determined to be non-compliant. The preset scoring threshold is a threshold set to determine whether the quality of the original claim materials is compliant, and can be set according to actual conditions.

[0142] Through the above steps S710 to S740, the quality of claim materials can be preliminarily inspected. For claim materials that do not meet the quality standards, there is no need to carry out subsequent quality control procedures, thereby improving the quality control efficiency of insurance claims.

[0143] In step S120 of some embodiments, the insurance claim materials have material serial numbers. These materials can be sorted according to their upload time or other criteria to obtain the material serial numbers. The earlier the upload time, the smaller the material serial number. The quality control concurrency quantity is the number of insurance claim materials that need to be processed in a batch, and this quantity can be configured and adjusted. The claim quantity is divided by the quality control concurrency quantity, and the quotient is rounded up to obtain the interval quantity. The serial number interval is determined based on the claim quantity, the quality control concurrency quantity, and the interval quantity. The interval length is the quality control concurrency quantity or the remainder obtained from the division, and the interval length is the number of material serial numbers within the interval. For example, if the claim quantity is 23 and the quality control concurrency quantity is 7, the interval quantity is 4. Based on the claim quantity 23, the interval quantity 4, and the quality control concurrency quantity 7, the serial number intervals are determined as [1,7], [8,14], [15,21], and [22,23].

[0144] In step S130 of some embodiments, the sequence number interval includes a left endpoint and a right endpoint. The sequence number intervals can be sorted from smallest to largest according to either the left endpoint or the right endpoint to obtain the interval arrangement order. The sequence number intervals are obtained according to the interval arrangement order to obtain the target interval, which includes the right endpoint of the interval.

[0145] In step S140 of some embodiments, an asynchronous thread is started to filter insurance claim materials, selecting those with material serial numbers within the target range to obtain reference claim materials with the target material category. The target material category is the predicted material category obtained in step S220.

[0146] In step S150 of some embodiments, according to the material serial number, the asynchronous thread sequentially calls the quality control model for each material type in the target material category. The invoked quality control model performs quality checks on the reference claim materials until the material serial number equals the right endpoint of the interval. The asynchronous thread then triggers a callback function to send the quality control result corresponding to the target interval to the main thread. The main thread writes the quality control result to the log and, referring to steps S130 to S150, continues to obtain the next target interval. The asynchronous thread performs quality checks on the next target interval until the quality control result for the last target interval is obtained. The quality control result includes whether the quality check passed or failed. Each material type corresponds to one quality control model, and the corresponding material type's quality control model can be called in parallel according to each material type in the target material category to improve the quality control efficiency of the reference claim materials. During the quality check process, the main thread is not blocked. When the asynchronous task is time-consuming, the main thread can perform other business operations. After the asynchronous task is completed, the asynchronous thread will call back the result to the main thread. The main thread does not need to wait for the asynchronous thread to return the quality check result, greatly improving the quality control efficiency of insurance claims.

[0147] This application provides a quality control method based on insurance claims, including: a main thread reads insurance claim materials from a test path and determines the claim quantity x. Based on the claim quantity x, the sequence number range is determined to be [1, x]. Based on the concurrent quality control quantity n and the sequence number range [1, x], the sequence number interval is determined to be [start_index, end_index], where start_index and end_index are the left and right endpoints of the sequence number interval, respectively. The initial value of start_index is 1, and end_index = min{start_index + n - 1, x}. An asynchronous thread sequentially performs quality checks on insurance claim materials whose sequence numbers fall within the sequence number interval. After completing the quality check on the insurance claim materials whose sequence number equals end_index, the asynchronous thread triggers a callback function to return the quality control result corresponding to the sequence number interval to the main thread. The main thread sets start_index to start_index = start_index + n, initiates the next batch of asynchronous n concurrent tests, and sets end_index to the minimum value between the current start_index + n - 1 and x to determine the current sequence number interval. The quality control process ends when the end_index is greater than x, by using an asynchronous thread to perform quality checks on insurance claim materials whose serial numbers are within the current serial number range.

[0148] By calling asynchronous threads from the main thread, the main thread is not blocked and can execute other business operations simultaneously. Upon completion of the asynchronous task, the result is returned to the main thread. During the callback, in addition to writing the quality control result to the log file, the next asynchronous call in the loop is initiated based on the sequence number, without the main thread needing to wait. After completing the current batch of asynchronous concurrent tasks, the next batch of asynchronous concurrent tasks is initiated at the callback node. Compared to serial testing, this reduces the number of times the loop control code is executed, improving the quality control efficiency of insurance claims.

[0149] Please see Figure 8 This application also provides a quality control device based on insurance claims, which can implement the above-mentioned quality control method based on insurance claims. The quality control device based on insurance claims includes:

[0150] The material acquisition module 810 is used to acquire insurance claim materials and the number of insurance claim materials required for the claim.

[0151] The interval determination module 820 is used to determine at least two sequence intervals based on the pre-acquired concurrent quality control and claim volume.

[0152] The interval acquisition module 830 is used to acquire the sequence intervals according to the interval arrangement order between the sequence intervals to obtain the target interval; wherein, the target interval includes the right endpoint of the interval;

[0153] The filtering module 840 is used to filter insurance claim materials through a preset asynchronous thread to obtain reference claim materials with material serial numbers and target material categories; wherein, the material serial number is located in the target range;

[0154] The quality control module 850 is used to sequentially perform quality inspections on the reference claims materials according to the material serial number, through the target material category and asynchronous thread, until the material serial number is equal to the right endpoint of the interval, and obtain the quality control result of the quality inspection; wherein, the quality control result includes whether the quality inspection passed or failed.

[0155] The specific implementation of the quality control device based on insurance claims is basically the same as the specific implementation of the quality control method based on insurance claims described above, and will not be repeated here.

[0156] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned quality control method based on insurance claims. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0157] Please see Figure 9 , Figure 9 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:

[0158] The processor 910 can be implemented using a general-purpose central processing unit (CPU), microprocessor, application specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0159] The memory 920 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 920 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 920 and is called and executed by the processor 910 to implement the quality control method based on insurance claims in the embodiments of this application.

[0160] The input / output interface 930 is used to implement information input and output;

[0161] The communication interface 940 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0162] Bus 950 transmits information between various components of the device (e.g., processor 910, memory 920, input / output interface 930, and communication interface 940);

[0163] The processor 910, memory 920, input / output interface 930 and communication interface 940 are connected to each other within the device via bus 950.

[0164] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described quality control method based on insurance claims.

[0165] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0166] The quality control method, device, electronic equipment, and computer storage medium based on insurance claims provided in this application embodiment automatically control the quality of insurance claim materials, eliminating the need for manual intervention throughout the entire quality control process and improving the efficiency of insurance claim quality control.

[0167] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0168] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0169] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0170] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0171] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0172] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0173] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0174] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0175] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0176] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0177] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A quality control method based on insurance claims, characterized in that, The method includes: Obtain insurance claim materials and the number of claims for which the materials are required; Based on the pre-obtained concurrent quality control data and the number of claims processed, at least two sequence number ranges are determined; The sequence number intervals are obtained according to the interval arrangement order between the sequence number intervals to obtain the target interval; wherein, the target interval includes the right endpoint of the interval; The insurance claim materials are filtered through a preset asynchronous thread to obtain reference claim materials with material serial numbers and target material categories; wherein, the material serial numbers are located within the target range; According to the material serial number, the quality control model is invoked sequentially through the target material category and the asynchronous thread to perform quality inspection on the reference claim materials until the material serial number is equal to the right endpoint of the interval, and the quality control result of the quality inspection is obtained; wherein, the quality control result includes quality inspection passed or quality inspection failed.

2. The method according to claim 1, characterized in that, The acquisition of insurance claim materials and the quantity of such insurance claim materials include: Obtain the original claim materials; The original claim materials are classified to obtain the predicted material categories; Based on the predicted material category, the original claim materials are preliminarily inspected to obtain the material quality category; The original claim materials are screened according to the predicted material category and the material quality category to obtain the insurance claim materials; Obtain the number of claims for the insurance claim materials.

3. The method according to claim 2, characterized in that, The process of classifying the original claim materials to obtain predicted material categories includes: Feature extraction is performed on the original claim materials to obtain first modality claim features and second modality claim features; Cross-attention calculation is performed on the first modality claim feature and the second modality claim feature to obtain attention features; The first modal claim feature, the second modal claim feature, and the attention feature are fused to obtain the reference claim feature; The original claim materials are classified according to the reference claim characteristics to obtain the predicted material category.

4. The method according to claim 3, characterized in that, The step of performing cross-attention calculation on the first modal claim features and the second modal claim features to obtain attention features includes: The first modality of claims features are used for feature mapping to obtain query features; Feature mapping is performed on the second modality claims features to obtain key features and value features; Calculate the attention score based on the query features and the key features; The attention features are obtained by weighted summation of the value features based on the attention scores.

5. The method according to claim 3, characterized in that, The feature fusion of the first modality claim feature, the second modality claim feature, and the attention feature to obtain the reference claim feature includes: Calculate the gating weight value based on the attention characteristics; Based on the gating weight value, the first modal claims feature and the attention feature are fused to obtain the first fused feature; Based on the gating weight value, the second modality claims feature and the attention feature are fused to obtain the second fused feature; The first fusion feature and the second fusion feature are fused to obtain the reference claim feature.

6. The method according to claim 5, characterized in that, The process of fusing the first fused feature and the second fused feature to obtain the reference claim feature includes: Calculate the similarity between the first fused feature and the second fused feature; Based on the similarity, principal component analysis is performed on the first fusion feature to obtain the first key feature; Principal component analysis was performed on the second fusion feature to obtain the second key feature; The first key feature and the second key feature are fused to obtain the reference claim feature.

7. The method according to claim 2, characterized in that, The preliminary inspection of the original claim materials based on the predicted material category to obtain the material quality category includes: Optical character recognition is performed on the original claim materials to obtain a character sequence; The character sequence is subjected to keyword recognition to obtain the target keywords; Calculate the quality score of the original claim materials based on the predicted material category and the target keywords; The original claim materials are classified according to the quality score to obtain the material quality category.

8. A quality control device based on insurance claims, characterized in that, The device includes: The material acquisition module is used to acquire insurance claim materials and the quantity of the insurance claim materials required for the claim. The interval determination module is used to determine at least two sequence intervals based on the pre-acquired concurrent quality control number and the number of claims. An interval acquisition module is used to acquire the numbered intervals according to the interval arrangement order between the numbered intervals to obtain a target interval; wherein, the target interval includes the right endpoint of the interval; The filtering module is used to filter the insurance claim materials through a preset asynchronous thread to obtain reference claim materials with material serial numbers and target material categories; wherein the material serial numbers are located in the target range; The quality control module is used to sequentially perform quality inspections on the reference claims materials according to the material serial number, through the target material category and the asynchronous thread, by calling the quality control model until the material serial number is equal to the right endpoint of the interval, and obtain the quality control result of the quality inspection; wherein, the quality control result includes quality inspection passed or quality inspection failed.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 7.

Images