Medical insurance data classification method and system based on text element analysis and extraction model
The judicial documents are processed through the text element analysis and extraction model, and structured recovery clues are generated, which solves the problem of medical insurance fund recovery, improves the recovery efficiency and accuracy, and reduces human dependence.
Patent Information
- Application Number
- CN202410638057.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-22
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-05-22
AI Technical Summary
It is difficult for medical insurance funds to recover in judicial litigation, and the medical insurance agency lacks specific methods to determine and recover the performance of the medical insurance fund, resulting in inefficient recovery.
The text element analysis and extraction model is used to process the judicial documents, extract medical insurance fund data and generate structured recovery clues, and data grading is performed through machine learning to avoid omissions and improve recovery efficiency.
It has achieved rapid verification of the performance of the medical insurance fund, generated post-judgment recovery clues, improved the recovery efficiency of the medical insurance fund, ensured data accuracy, and reduced human dependence.
Smart Images

Figure CN118212078B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of government-court linkage data technology, and specifically relates to a medical insurance data classification method and system based on a text element analysis and extraction model. Background Art
[0002] How to promote the business collaboration and data sharing of judicial and government information systems, build a multi-dimensional three-dimensional network for barrier-free communication between the government and the courts by building a government-court communication platform, and break the communication barrier between the government and the courts is also one of the current hot issues. Strengthening the regular supervision of the use of medical insurance funds is of great significance to ensuring the safe operation of medical insurance funds, improving the efficiency of fund use, standardizing medical service behaviors, and reducing the burden of medical treatment for the masses. Medical expenses should be borne by a third party according to law. If the third party does not pay or the third party cannot be identified, the basic medical insurance fund shall pay first. After the basic medical insurance fund pays first, it has the right to claim compensation from the third party. This can safeguard the legitimate rights and interests of citizens' social insurance and ensure that the insured persons who have been harmed can receive timely medical treatment. However, in judicial practice, in many tort lawsuits involving personal injury compensation, such as motor vehicle traffic accident liability disputes, medical damage liability disputes, health rights disputes, life rights disputes, etc., part of the medical expenses of the victim (plaintiff) have been paid by the medical insurance fund, but the tortfeasor (defendant) is fully recovered for medical expenses. Although the Social Insurance Law stipulates that the medical insurance fund has the right to seek compensation from the infringer after advance payment of expenses, it does not specify how to seek compensation specifically.
[0003] There are practical obstacles for medical insurance agencies to directly seek compensation from the infringer. Since they do not know the specific circumstances of the parties to the lawsuit, it is difficult for them to determine the actual repayment of the medical insurance fund and obtain clues for recovery of the medical insurance fund. Therefore, how to reasonably use the data of government-court linkage to quickly verify the actual performance of the medical insurance funds paid in advance and quickly provide clues for recovery is an issue that needs to be solved urgently. Summary of the invention
[0004] The purpose of the present invention is to provide a medical insurance data classification method and system based on a text element analysis and extraction model, which processes and checks the judicial documents according to specific element information, classifies the relevant medical insurance fund data and automatically generates a medical insurance fund return verification letter, solving the problem of the difficulty in accurately determining the quality of medical insurance fund recovery.
[0005] To achieve the above-mentioned object of the invention, the technical solution of the present invention is as follows:
[0006] A medical insurance data classification method based on a text element analysis and extraction model comprises the following steps:
[0007] Extract the judgment documents of closed cases with a settlement amount greater than 0;
[0008] Text analysis: Use the trained text element analysis and extraction model to analyze the sentence-level document elements of the judicial documents and extract element information;
[0009] Medical insurance fund data classification: judge the element information according to the conditions for medical insurance fund advance payment, and then classify the medical insurance fund data, and generate post-judgment recovery clue structured data for the medical insurance fund data that needs to be recovered.
[0010] The present invention creatively screens medical insurance fund advance payment cases from the perspective of closed judicial documents, thereby avoiding the possibility of omissions to the greatest extent; at the same time, based on machine learning, it classifies medical insurance fund data, thereby directly linking the medical insurance fund data in the judicial documents with subsequent processing, which can not only quickly verify the actual performance of the medical insurance advance payment fund, but also rely on the more comprehensive information in the judicial documents to directly generate post-judgment recovery clue structured data. The medical insurance bureau can directly recover based on the post-judgment recovery clue structured data, thereby improving the recovery efficiency.
[0011] In the above-mentioned medical insurance data classification method based on the text element analysis and extraction model, the medical insurance fund data classification includes judging whether there is a medical insurance fund advance payment according to the conditions for the occurrence of the medical insurance fund advance payment, and judging whether the medical insurance fund advance payment has been fulfilled in combination with the actual case payment information;
[0012] The medical insurance fund data are divided into three levels: no medical insurance fund advance payment, medical insurance fund advance payment and it has been fully fulfilled, and medical insurance fund advance payment that has not been fully fulfilled.
[0013] In the present invention, the element information extracted by the text element analysis and extraction model can be judged according to the conditions for the advance payment of medical insurance funds, including "whether personal injury is involved", "whether medical expenses are included", "whether medical expenses need to be compensated" and the actual case payment information, including "the amount of medical expense compensation" and "whether it has been fulfilled". With the help of the text element analysis and extraction model, no manual processing is required, and the element information of the medical insurance fund data is extracted efficiently and quickly, and directly graded so as to be directly linked to post-processing.
[0014] In the above-mentioned medical insurance data classification method based on text element analysis and extraction model, the two levels of medical insurance fund advance payment that has been fulfilled and medical insurance fund advance payment that has not been fulfilled both generate corresponding medical insurance fund return verification letters and post-judgment recovery clue structured data.
[0015] In order to improve the perfection of the medical insurance fund data processing by the technical solution of the present invention, as long as there is an advance payment of the medical insurance fund, a medical insurance fund return verification letter and post-judgment recovery clue structured data will be generated.
[0016] Since the medical expenses paid by the third party are not included in the payment scope of the medical insurance fund, the large model can effectively extract part of the medical expenses that should be borne by the third party and pay them through the medical insurance fund. Later, they sued the infringer in court to demand compensation for the medical expenses. After the judgment was made and took effect, the court has ruled that the infringer should compensate for part of the expenses reimbursed through the medical insurance fund. According to the law, the expenses should be returned to the medical insurance fund. The structured data of the post-judgment recovery clues mainly includes the case number, case filing time, case cause, parties, ID number, accident time, accident location, medical expenses compensation judgment amount, whether it has been automatically fulfilled, medical insurance fund verification letter, and evidence materials. The structured data queries the case information based on the case number, extracts the basic case information and party information from the case information, and obtains the case filing time, parties, ID number and other information based on the field name to extract and generate.
[0017] In the above-mentioned medical insurance data classification method based on text element analysis and extraction model, the judgment of whether the fulfillment has been completed includes extracting the amount of medical expense compensation, extracting structured information of case funds received, and judging the fulfillment of the case funds in combination with the amount of medical expense compensation.
[0018] The amount of medical expenses compensation can be directly extracted based on the prediction results of the text factor analysis and extraction model, and the case funds receipt status can be queried from the court’s internal database, thereby ensuring the accuracy of the data and avoiding errors or omissions.
[0019] In the above-mentioned medical insurance data classification method based on text element analysis and extraction model, the three-level medical insurance fund data are respectively connected to different post-processing, as follows:
[0020] If there is no advance payment from the medical insurance fund, the medical insurance fund data processing process for the judgment document ends.
[0021] If there is an advance payment from the medical insurance fund and it has been fulfilled, a medical insurance fund refund verification letter and post-judgment recovery clue structured data will be generated, and the court will verify and push it to the medical insurance bureau.
[0022] If there is any unfulfilled advance payment of medical insurance funds, a medical insurance fund return verification letter and post-judgment recovery clue structured data will be generated. After verification by the court, the data will be pushed to the Medical Insurance Bureau, which will feedback to the court after the recovery is completed.
[0023] For judicial documents that do not involve any advance payment from medical insurance funds, no further processing is required and the case can be concluded directly. However, for medical insurance fund data that involve advance payment from medical insurance funds, since the changes in money are involved, the generated medical insurance fund return verification letter and post-judgment recovery clue structured data need to be verified by the court before being pushed, in order to ensure the stability of the verification letter and recovery clues, and to provide the most comprehensive protection for the medical insurance fund.
[0024] In the above-mentioned medical insurance data classification method based on text element analysis and extraction model, the text analysis includes the following steps:
[0025] Perform machine reading comprehension on judicial documents and output the document sentences to be processed.
[0026] Output the element instance sentence containing the document elements through the trained text element analysis and extraction model.
[0027] Post-process the model prediction results to extract feature information.
[0028] The specific steps of text analysis in the present invention are as follows:
[0029] (1) Perform machine reading and comprehension of judicial documents and output the results;
[0030] (2) Based on the questions of “whether personal injury is involved”, “whether medical expenses are included”, “whether medical expenses need to be compensated”, and “the amount of medical expenses compensation”, the trained text element analysis and extraction model is used to perform sentence-level document element analysis on the document sentences to be processed, and the document sentences containing element semantics (element instance sentences) are output;
[0031] (3) After extracting the document elements, complete the similarity detection between the document elements and the questions to enhance the results of machine reading comprehension;
[0032] (4) Integrate the answer and document elements and output the model prediction results;
[0033] (5) Classify the questions according to the question keywords, post-process the prediction results, and extract the final question results, that is, the element information that needs to be extracted.
[0034] In the above-mentioned medical insurance data classification method based on text element analysis and extraction model, the text element analysis and extraction model calculates the similarity between the trained element instance sentences containing specific elements and the document sentences of the judicial documents, and then matches the corresponding semantics.
[0035] The text element analysis and extraction model in the present invention includes a Bert-Chinese model, an attention layer, a maximum pooling layer, and an output layer. In order to ensure the smooth application of the corresponding data under the classification of medical insurance fund data, the present invention also optimizes the relevant parameters of the medical insurance data, including the size of the vocabulary, the dimension of the word vector, the maximum size of the text input, the maximum sequence length, the learning rate, and the stride, etc.; thereby improving the accuracy of the model prediction results.
[0036] A medical insurance data classification system based on a text element analysis and extraction model, the system is used to execute the above-mentioned medical insurance data classification method based on a text element analysis and extraction model, including: a scheduled task module, a document analysis module, a medical insurance fund data classification module and a post-processing module.
[0037] In the above-mentioned medical insurance data classification system based on text element analysis and extraction model, the scheduled task module includes:
[0038] Regularly check all unmarked closed cases with a settlement amount greater than 0.
[0039] Extract the judgment documents of the cases in the case sequence in sequence and hand them over to the document analysis module for processing;
[0040] The document analysis module uses the trained text element analysis and extraction model to analyze the document elements at the sentence level and extract element information.
[0041] The medical insurance fund data classification module judges the element information according to the conditions for the advance payment of the medical insurance fund, and then classifies the medical insurance fund data, and generates post-judgment recovery clue structured data for the medical insurance fund data that needs to be recovered.
[0042] The scheduled inspection in the present invention can be set to be conducted at 23:00 every day, or at 24:00 every day, etc. The present invention can conduct inspection of medical insurance fund data on a daily basis, can locate the medical insurance funds paid in a timely and efficient manner, and has less reliance on manpower, thereby effectively improving the verification and recovery of medical insurance fund advances by the medical insurance bureau.
[0043] In the above-mentioned medical insurance data classification system based on text element analysis and extraction model, in the post-processing module,
[0044] If structured data of medical insurance recovery clues has been generated and a medical insurance fund return verification letter has been generated, the case will be marked as handled, reminding the judge to handle the medical insurance recovery event and verify the information.
[0045] If it is determined that there is no advance payment from the medical insurance fund, the process ends and the next case is processed until it is completed.
[0046] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0047] (1) The present invention creatively screens medical insurance fund advance payment cases from the perspective of closed judicial documents, thereby avoiding the possibility of omissions to the greatest extent; at the same time, based on machine learning, it classifies medical insurance fund data, thereby directly linking the medical insurance fund data in the judicial documents with subsequent processing, which can not only quickly verify the actual implementation of the medical insurance advance payment fund, but also directly generate post-judgment recovery clue structured data based on the more comprehensive information in the judicial documents. The medical insurance bureau can directly recover the money based on the post-judgment recovery clue structured data, thereby improving the recovery efficiency.
[0048] (2) In the present invention, the element information extracted by the text element analysis and extraction model can be judged based on the conditions for the occurrence of medical insurance fund advance payment, including "whether personal injury is involved", "whether medical expenses are included", "whether medical expenses need to be compensated" and the actual case payment information, including "the amount of medical expenses compensation" and "whether it has been fulfilled". With the help of the text element analysis and extraction model, no manual processing is required, and the element information of the medical insurance fund data is extracted efficiently and quickly and directly graded so as to be directly linked to post-processing.
[0049] (3) The amount of medical expense compensation can be directly obtained based on the prediction results of the text factor analysis and extraction model, and the receipt of case funds can be queried from the court’s internal database, thereby ensuring the accuracy of the data and avoiding errors or omissions.
[0050] (4) For judgment documents that do not involve advance payment of medical insurance funds, no further processing is required and the case can be closed directly. However, for medical insurance fund data that involve advance payment of medical insurance funds, since the changes in money are involved, the generated medical insurance fund return verification letter and post-judgment recovery clue structured data need to be verified by the court before being pushed, in order to ensure the stability of the verification letter and recovery clues and provide the most comprehensive protection for the medical insurance funds.
[0051] (5) The present invention can check the medical insurance fund data on a daily basis, and can locate the medical insurance funds that have been advanced in a timely and efficient manner, with less reliance on manpower, thereby effectively improving the medical insurance bureau's verification and recovery of the medical insurance fund advance payment situation. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] Figure 1 It is a schematic diagram of the process of the medical insurance data text element analysis, extraction and classification system of the present invention;
[0053] Figure 2 Schematic diagram of the network architecture of the text element analysis and extraction model of the present invention. DETAILED DESCRIPTION
[0054] Specific implementation modes are listed below to further explain the technical solution of the present invention in detail.
[0055] Example 1
[0056] like Figure 2 As shown, this embodiment provides a medical insurance data text element analysis, extraction and classification system, the specific process is as follows:
[0057] S1 Scheduled Task Module: Check all unmarked closed cases with a settlement amount greater than 0 at 23:00 every day;
[0058] For one of the cases in the case sequence, retrieve the court’s electronic file and extract the judgment document.
[0059] S2 Document Analysis Module: Combined Figure 1 The network architecture diagram of the text feature analysis and extraction model shown is as follows:
[0060] (1) Perform machine reading and comprehension of judicial documents and output the results;
[0061] (2) Based on the questions of “whether personal injury is involved”, “whether medical expenses are included”, “whether medical expenses need to be compensated”, and “the amount of medical expenses compensation”, the trained text element analysis and extraction model is used to perform sentence-level document element analysis on the judgment document and output document sentences containing element semantics (element instance sentences);
[0062] (3) After extracting the document elements, complete the similarity detection between the document elements and the questions to enhance the results of machine reading comprehension;
[0063] (4) Integrate the answer and document elements and output the model prediction results;
[0064] (5) Classify the questions according to the question keywords, post-process the prediction results, and extract the final question results, that is, the element information that needs to be extracted.
[0065] S3 medical insurance fund data classification module and post-processing module
[0066] (1) Judgment of element information - whether personal injury is involved - No: Level 1 data, not pushed, the process ends; Yes: continue to judge.
[0067] (2) Element information judgment - whether medical expenses are included - No: Level 1 data, not pushed, the process ends; Yes: continue to judge.
[0068] (3) Element information judgment - whether medical expenses need to be compensated - No: Level 1 data, do not push, the process ends; Yes: continue to extract.
[0069] (4) For medical insurance fund data that has been advanced by the medical insurance fund, extract element information - extract the amount of medical expense compensation.
[0070] (5) Extract case structured data and check whether the case payment has been completed.
[0071] Completion completed: Level 2 data, generate a medical insurance fund return verification letter and generate post-judgment recovery clue structured data. The judge will check and click to push; the level 2 data can be received by medical insurance.
[0072] Not fully fulfilled: Level 3 data, generates a medical insurance fund return verification letter, generates post-judgment recovery clue structured data, and the judge checks and clicks to push; after the level 3 data medical insurance recovery is completed, it needs to be fed back to the court and confirmed by the judge.
[0073] (6) Generate structured data of medical insurance recovery clues and generate a medical insurance fund return verification letter, marking the case as processed. Remind the judge to handle the medical insurance recovery event and verify the information.
[0074] S4 continues to process the next case until the processing is completed.
[0075] The system provided in this embodiment can check the medical insurance fund data on a daily basis, can locate the advanced medical insurance funds in a timely and efficient manner, and is less dependent on manpower, thereby effectively improving the medical insurance bureau's verification and recovery of the advanced medical insurance fund situation.
[0076] It should be appreciated that embodiments of the present invention may be implemented or enforced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The method may be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if desired, the program may be implemented in an assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed ASIC for this purpose.
[0077] Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes described herein (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that is executed collectively on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions that may be executed by one or more processors.
[0078] Further, the method can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, a RAM, a ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or part thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques described in the present invention, the present invention also includes the computer itself. The computer program can be applied to input data to perform the functions described herein, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents a physical and tangible object, including a specific visual depiction of the physical and tangible object produced on a display.
[0079] As used in this application, the terms "component", "module", "system", etc. are intended to refer to a computer-related entity, which can be hardware, firmware, a combination of hardware and software, software, or software in operation. For example, a component can be, but is not limited to: a process running on a processor, a processor, an object, an executable file, a thread in execution, a program, and / or a computer. As an example, an application running on a computing device and the computing device can both be components. One or more components can exist in a process and / or thread in execution, and a component can be located in a computer and / or distributed between two or more computers. In addition, these components can be executed from various computer-readable media having various data structures thereon. These components can communicate in a local and / or remote process manner, such as according to a signal having one or more data packets (e.g., data from a component that interacts with another component in a local system, a distributed system, and / or interacts with other systems in a signal manner through a network such as the Internet).
[0080] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
Claims
1. A medical insurance data classification method based on a text element analysis and extraction model, characterized in that: The steps include: Extract the judgment documents of closed cases with a settlement amount greater than 0; Text analysis: Perform machine reading comprehension on judicial documents and output document sentences to be processed; Based on the questions of "whether personal injury is involved", "whether medical expenses are included", "whether medical expenses need to be compensated", and "the amount of medical expenses compensation", the trained text element analysis and extraction model is used to analyze the document elements at the sentence level of the judgment document, and output the element instance sentences containing the document elements; Integrate the answer to the question and the document elements and output the model prediction results; Classify questions according to question keywords, post-process the prediction results, and extract the final question results as factor information; Based on the conditions for advance payment of medical insurance funds, including "whether personal injury is involved", "whether medical expenses are included", "whether medical expenses need to be compensated" and the actual case payment information, including "the amount of medical expense compensation" and "whether it has been fulfilled", the element information extracted by the text element analysis and extraction model is judged; Medical insurance fund data classification: The element information is judged according to the conditions for the occurrence of medical insurance fund advance payment, and then the medical insurance fund data is classified, and the structured data of post-judgment recovery clues is generated for the medical insurance fund data that needs to be recovered; the structured data of post-judgment recovery clues includes case number, case filing time, case cause, parties, ID number, accident time, accident location, medical expense compensation judgment amount, whether it has been automatically fulfilled, medical insurance fund verification letter, and evidence materials; the structured data is generated by querying case information based on the case number and extracted from the case information; The medical insurance fund data classification includes determining whether there is a medical insurance fund advance payment based on the conditions for the occurrence of the medical insurance fund advance payment and determining whether the medical insurance fund advance payment has been fulfilled based on the actual case payment information; The medical insurance fund data are divided into three levels: no medical insurance fund advance payment, medical insurance fund advance payment and fulfillment, and medical insurance fund advance payment but not fulfillment; The three-level medical insurance fund data are connected to different post-processing, as follows: If there is no advance payment from the medical insurance fund, the medical insurance fund data processing process for the judgment document ends. If there is a medical insurance fund advance payment and it has been fulfilled, a medical insurance fund return verification letter and post-judgment recovery clue structured data will be generated, and the structural information will be marked as fulfilled. After verification by the court, it will be pushed to the Medical Insurance Bureau. If there is any unfulfilled advance payment of medical insurance funds, a medical insurance fund return verification letter and post-judgment recovery clue structured data will be generated, and the structural information will be marked as unfulfilled. After verification by the court, it will be pushed to the Medical Insurance Bureau, and the Medical Insurance Bureau will feedback to the court after the recovery is completed.
2. The medical insurance data classification method based on the text element analysis and extraction model according to claim 1 is characterized in that: Both levels, namely, medical insurance fund advance payment that has been fulfilled and medical insurance fund advance payment that has not been fulfilled, generate corresponding medical insurance fund return verification letters and post-judgment recovery clue structured data.
3. The medical insurance data classification method based on the text element analysis and extraction model according to claim 2 is characterized in that: The judgment of whether the fulfillment has been completed includes extracting the amount of medical expenses compensation, extracting the structured information of the case funds received, and judging the fulfillment of the case funds based on the amount of medical expenses compensation.
4. A medical insurance data classification system based on a text element analysis and extraction model, characterized by: The system is used to execute a medical insurance data classification method based on a text element analysis and extraction model as described in any one of claims 1-3, including: a scheduled task module, a document analysis module, a medical insurance fund data classification module and a post-processing module.
5. The medical insurance data classification system based on text element analysis and extraction model as claimed in claim 4, characterized in that: The timing task module includes: Regularly check all unmarked closed cases with a settlement amount greater than 0. Extract the judgment documents of the cases in the case sequence in sequence and hand them over to the document analysis module for processing; The document analysis module uses the trained text element analysis and extraction model to analyze the document elements at the sentence level and extract element information. The medical insurance fund data classification module judges the element information according to the conditions for the advance payment of the medical insurance fund, and then classifies the medical insurance fund data, and generates post-judgment recovery clue structured data for the medical insurance fund data that needs to be recovered.
6. The medical insurance data classification system based on text element analysis and extraction model according to claim 5, characterized in that: In the post-processing module, if the medical insurance recovery clue structured data has been generated and a medical insurance fund return verification letter has been generated, the case is marked as processed, and the judge is reminded to handle the medical insurance recovery event and verify the information.
