Car insurance claim settlement evaluation system and method based on large language model
Through the auto insurance claims evaluation system based on the big language model, the problems of inefficient claims and uncertainty in traditional auto insurance claims are solved, and fast and accurate claims evaluation is achieved, which improves customer satisfaction.
Patent Information
- Application Number
- CN202411356593.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-26
- Publication Date
- 2024-12-27
AI Technical Summary
The traditional auto insurance claims process relies on manual review, which leads to inefficiency and subjective interference, resulting in uncertainty in claims results.
Adopt a car insurance claim evaluation system based on a large language model, including a data preprocessing module, a large language model training module and a multi-task processing module. Through recording and translator, information extraction, model training and task processing, auto insurance claim evaluation report is generated to reduce manual labor The subjectivity and inconsistency of judgment.
It greatly shortens the claims processing time, improves the accuracy and fairness of the evaluation results, reduces disputes and disputes in manual judgments, and improves customer experience and trust.
Smart Images
Figure CN119205367A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of automobile insurance claims assessment, and in particular to an automobile insurance claims assessment system and method based on a large language model. Background Art
[0002] Motor vehicle insurance (auto insurance for short) is an important form of commercial insurance used to protect motor vehicles from personal injury and property loss caused by natural disasters or accidents. With the continuous increase in the total number of vehicles, auto insurance claims business has also ushered in huge challenges and opportunities.
[0003] The traditional auto insurance claims process relies on manual review and judgment. The manual judgment process takes a lot of time, resulting in low efficiency. In addition, the manual judgment process is easily interfered by subjective factors, resulting in uncertainty in the claims results.
[0004] In order to improve claims efficiency and enhance customer experience, the industry urgently needs an intelligent and automated auto insurance claims assessment system. Summary of the invention
[0005] In response to the above problems, this application provides a car insurance claim assessment system based on a large language model, including the following contents:
[0006] In a first aspect, the present application provides a car insurance claim assessment system based on a large language model, the system comprising:
[0007] Data preprocessing module, large language model training module, multi-task processing module and result output module, among which:
[0008] The data preprocessing module is used to transcribe the communication recordings between the front-line accident investigators and the customers into text, and extract information based on preset keywords;
[0009] The large language model training module is used to train the accident loss situation judgment and vehicle damage amount estimation tasks, the auto insurance accident responsibility judgment tasks, the insurance exemption clause judgment tasks and the insurance product compensation reasoning tasks based on the large language basic model through the preset prompt word prompt and the training data set;
[0010] The multi-task processing module is used to use the trained large language model to simultaneously perform the accident loss situation judgment and vehicle damage amount estimation tasks, the auto insurance accident responsibility judgment task, the insurance exemption clause judgment task and the insurance product compensation reasoning task, so as to judge the accident loss situation, estimate the vehicle damage amount, judge the accident responsibility, review the insurance exemption clause and reason the insurance product compensation;
[0011] The result output module is used to integrate the processing results of each task and generate a car insurance claim assessment report.
[0012] Optionally, the data preprocessing module is also used for data cleaning and feature extraction.
[0013] Optionally, when the large language model training module is based on the large language basic model, and trains the accident loss situation judgment and vehicle damage amount estimation tasks through preset prompt words and training data sets, the preset prompt words and training data sets include transcribed text of communication recordings between front-line accident investigators and customers in historical accident cases, which are used to enable the trained large model to judge the vehicle damage of both parties to the accident and the loss of surrounding environmental facilities, and then make amount predictions based on the vehicle damage of both parties to the accident and the loss of surrounding environmental facilities.
[0014] Optionally, when the large language model training module trains the motor vehicle accident responsibility judgment task based on the large language basic model through preset prompt words and training data sets, the preset prompt words prompt and training data sets include historical accident responsibility determination documents, so that the trained large language model can judge the responsibilities of both parties in the accident and give relevant road traffic safety law basis. The types of responsibilities of both parties in the accident include full responsibility, primary responsibility, secondary responsibility, no responsibility and equal responsibility.
[0015] Optionally, when the large language model training module trains the insurance exemption clause judgment task based on the large language basic model through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set contain exemption clause instances, which are used to enable the trained large model to recognize and determine whether the accident triggers the exemption clause.
[0016] Optionally, when the large language model training module trains the insurance product claim reasoning task based on the large language basic model through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set include a variety of claim cases, which are used to enable the trained large language model to automatically determine the type of claim, claim ratio and amount.
[0017] Optionally, the system also includes a user interaction interface for receiving accident information input by the user, displaying the claims assessment process and results, providing clear claims progress tracking and result query functions, and improving user experience and satisfaction.
[0018] In a second aspect, the present application provides a method for evaluating automobile insurance claims based on a large language model, the method comprising:
[0019] Obtain recordings of communications between frontline accident investigators and clients and transcribe them into text;
[0020] Use the pre-trained large model to judge the accident loss and estimate the amount of vehicle damage, judge the auto insurance accident, judge the insurance exemption clauses and analyze the insurance product compensation based on the transcribed text;
[0021] A motor vehicle insurance claim assessment report is generated based on the analysis results of the pre-trained large model, wherein the assessment report includes an accident overview, loss situation, liability determination, analysis of the applicability of exemption clauses, and compensation recommendations.
[0022] Optionally, the method also includes data cleaning and feature extraction.
[0023] Optionally, when the large language model training module is based on the large language basic model, and trains the accident loss situation judgment and vehicle damage amount estimation tasks through preset prompt words and training data sets, the preset prompt words and training data sets include transcribed text of communication recordings between front-line accident investigators and customers in historical accident cases, so that the trained large model can judge the vehicle damage of both parties to the accident and the loss of surrounding environmental facilities, and then make amount predictions based on the vehicle damage of both parties to the accident and the loss of surrounding environmental facilities.
[0024] Optionally, when the large language model training module trains the motor vehicle accident responsibility judgment task based on the large language basic model through preset prompt words and training data sets, the preset prompt words prompt and training data sets include historical accident responsibility determination documents, so that the trained large language model can judge the responsibilities of both parties in the accident and give relevant road traffic safety law basis. The types of responsibilities of both parties in the accident include full responsibility, primary responsibility, secondary responsibility, no responsibility and equal responsibility.
[0025] Optionally, when the large language model training module trains the insurance exemption clause judgment task based on the large language basic model through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set contain exemption clause instances, so that the trained large model can recognize and determine whether the accident triggers the exemption clause.
[0026] Optionally, when the large language model training module trains the insurance product claim reasoning task based on the large language basic model through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set include a variety of claim cases, so that the trained large language model automatically determines the claim type, claim ratio and amount.
[0027] Optionally, the method further includes receiving accident information input by a user through a user interaction interface, displaying the claims assessment process and results through the user interaction interface, and obtaining claims progress tracking and query results.
[0028] In a third aspect, the present application provides a device comprising a memory and a processor, wherein the memory is used to store instructions or codes, and the processor is used to execute the instructions or codes, so that the device executes the auto insurance claim assessment method based on a large language model introduced in any implementation of the aforementioned second aspect.
[0029] In a fourth aspect, the present application provides a computer-readable storage medium, in which a code is stored. When the code is executed, the device executing the code implements the auto insurance claim assessment method based on a large language model introduced in any implementation method of the aforementioned second aspect.
[0030] The present application provides a motor vehicle insurance claim assessment system based on a large language model, the system comprising a data preprocessing module, a large language model training module, a multi-task processing module and a result output module, wherein: the data preprocessing module is used to transcribe the communication recordings between the front-line accident investigators and the customers into text, and extract information based on preset keywords; the large language model training module is used to train the accident loss situation judgment and vehicle damage amount estimation tasks, the motor vehicle insurance accident responsibility judgment tasks, the insurance exemption clause judgment tasks and the insurance product compensation reasoning tasks based on the large language basic model through preset prompt words prompt and training data sets; the multi-task processing module is used to use the trained large language model to simultaneously execute the accident loss situation judgment and vehicle damage amount estimation tasks, the motor vehicle insurance accident responsibility judgment tasks, the insurance exemption clause judgment tasks and the insurance product compensation reasoning tasks, so as to judge the accident loss situation, estimate the vehicle damage amount, judge the accident responsibility, review the insurance exemption clauses and reason the insurance product compensation; the result output module is used to integrate the processing results of each task to generate a motor vehicle insurance claim assessment report.
[0031] The above system can significantly shorten the claims processing time by integrating automated processes such as data preprocessing, large language model training, multi-task parallel processing and result integration. The data preprocessing module transcribes the recording into text and extracts key information, and combines it with the trained large language model to perform the claims assessment task, replacing traditional manual judgment. The result output module can generate a detailed and clear auto insurance claims assessment report, which is convenient for customers to understand the claims process and results, and has the effect of improving customer satisfaction and trust. In addition, the large language model is trained based on massive data and complex algorithms, which can reduce the subjectivity and inconsistency of human judgment. In key links such as accident loss assessment, liability determination, exemption clause review and compensation reasoning, the system can conduct objective analysis based on preset rules and logic to ensure the fairness and accuracy of the claims results, and can effectively avoid disputes and disputes during manual judgment. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] In order to more clearly illustrate the technical solutions in this embodiment or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0033] Figure 1 A schematic diagram of the structure of a car insurance claim assessment system based on a large language model provided in an embodiment of the present application;
[0034] Figure 2 A schematic diagram of the structure of a user interaction interface provided in an embodiment of the present application;
[0035] Figure 3 A flowchart of a method for evaluating a motor vehicle insurance claim based on a large language model is provided in an embodiment of the present application. DETAILED DESCRIPTION
[0036] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with the relevant laws, regulations and standards of relevant countries and regions.
[0037] Figure 1 A structural diagram of a car insurance claim assessment system based on a large language model provided in an embodiment of the present application, according to Figure 1 As shown, the automobile insurance claim assessment system based on the large language model in the present application includes: a data preprocessing module 101, a large language model training module 102, a multi-task processing module 103 and a result output module 104.
[0038] The data preprocessing module 101 is used to transcribe the communication recordings between the front-line accident investigators and the customers into text, and extract information based on preset keywords.
[0039] Recording to text: Use speech recognition technology (ASR, Automatic Speech Recognition) to convert the recorded conversation between the front-line personnel at the accident site and the customer into text. This step needs to ensure a high transcription accuracy to preserve the integrity and accuracy of the conversation.
[0040] Keyword extraction: Based on a preset keyword list (such as "damage", "liability", "compensation", etc.), extract information related to the keyword list from the transcribed text. These keywords can help the large language model understand the accident situation and customer needs more accurately in subsequent processing.
[0041] In one implementation of the embodiment of the present application, the data preprocessing module is also used for data cleaning and feature extraction. The data cleaning and feature extraction process includes directly processing the communication recordings between the front-line accident investigators and the customers and processing the data after the communication recordings between the front-line accident investigators and the customers are transcribed into text.
[0042] Data processing process of the data preprocessing module:
[0043] Before the recording is transcribed into text, the recording file is processed initially to remove those parts of the audio that are useless for subsequent analysis, such as long periods of silence, background noise, irrelevant conversation fragments, etc. Audio processing technology can be used to identify and trim invalid parts of the recording. Thresholds can be set to determine whether the audio is valid, for example, audio below a certain volume or a certain frequency range may be considered invalid; or the pre-processing function of the speech recognition software can be used to assist in removing invalid audio.
[0044] After the recording is transcribed into text, the text data is further processed to remove text information that is not useful for claim assessment, such as speech recognition errors, irrelevant parentheses, repeated sentences, etc. The text can be cleaned by using natural language processing technology (NLP), including removing punctuation errors, correcting spelling errors, deleting meaningless filler words, etc. Identify and delete text content that is not relevant to claim assessment, such as polite language, small talk topics, etc.
[0045] The data quality of subsequent claims assessment tasks can be significantly improved by processing the data in the above data preprocessing module. Removing invalid audio directly based on the recording can reduce interference in the transcription process and improve the transcription accuracy; further processing of the transcribed text can ensure the accuracy and usefulness of the text data, laying a solid foundation for subsequent feature extraction and model analysis.
[0046] It should be noted that the above two steps may be adjusted and optimized according to specific circumstances in actual implementation. For example, in some cases, the recording may be preliminarily processed before transcription to reduce the amount of calculation and error rate in the transcription process; in other cases, the transcription may be performed directly and then the text after transcription is processed to process the text data more flexibly.
[0047] The large language model training module 102 is used to train the accident loss situation judgment and vehicle damage amount estimation tasks, the auto insurance accident responsibility judgment tasks, the insurance exemption clause judgment tasks and the insurance product compensation reasoning tasks based on the large language basic model through preset prompt words and training data sets.
[0048] The large language model or basic model described in the embodiments of the present application may refer to the Qwen-72B basic model based on Alibaba. Qwen-72B is an important member of the Tongyi Qianwen large model series developed by Alibaba Cloud, with a parameter scale of 72 billion. This model has demonstrated strong capabilities in the field of natural language processing, especially in pre-training in Chinese and English. Qwen-72B-Chat is an AI assistant based on a large language model built on the basis of Qwen-72B using an alignment mechanism, which further enhances the model's dialogue and understanding capabilities.
[0049] By training a large language model, it can handle specific tasks in auto insurance claims. The specific process includes:
[0050] Choose a base model: Choose a large language model suitable for natural language processing tasks as a basis.
[0051] Prepare training data sets: Collect and annotate a large number of data sets on tasks such as accident loss, vehicle damage estimation, accident liability judgment, insurance exemption clauses, and insurance product compensation.
[0052] Design prompts: Design appropriate prompts or instructions for each task to guide the model to understand and perform specific tasks.
[0053] Model training: Use labeled data sets and prompt words to train the model, and improve the performance of the model on various tasks by adjusting model parameters and optimizing algorithms.
[0054] Model evaluation: Evaluate model performance through the test set to ensure that the model can achieve the expected accuracy and efficiency on different tasks.
[0055] The multi-task processing module 103 is used to use the trained large language model to simultaneously execute the accident loss situation judgment and vehicle damage amount estimation tasks, the motor vehicle insurance accident responsibility judgment task, the insurance exemption clause judgment task and the insurance product compensation reasoning task, so as to judge the accident loss situation, estimate the vehicle damage amount, judge the accident responsibility, review the insurance exemption clause and reason the insurance product compensation.
[0056] The pre-processed data is fed into different parts of the model to handle tasks such as accident loss judgment, vehicle damage amount estimation, accident liability judgment, insurance exemption clause review, and insurance product compensation reasoning. The parallel computing capability of the model is used to process multiple tasks at the same time to improve the processing speed. After each task is processed, the processing results of each task are integrated to form a comprehensive analysis and evaluation of the accident.
[0057] The result output module 104 is used to integrate the processing results of each task and generate a motor vehicle insurance claim assessment report.
[0058] The result output module can organize the output results of the multi-task processing module into an easy-to-understand report for reference by claims adjusters.
[0059] The specific implementation process may include:
[0060] Result analysis: Further analyze the output results of the multi-task processing module to ensure the accuracy and rationality of the results.
[0061] Report generation: Based on the analysis results, the auto insurance claim assessment report is automatically generated. The report should include accident loss, vehicle damage amount estimation, accident responsibility judgment, insurance exemption clause review results and insurance product compensation reasoning.
[0062] In one implementation of the embodiment of the present application, the large language model training module is based on the large language basic model. When training the accident loss situation judgment and vehicle damage amount estimation tasks through preset prompt words and training data sets, the preset prompt words and training data sets include transcribed text of communication recordings between front-line accident investigators and customers in historical accident cases, which are used to enable the trained large model to judge the vehicle damage of both parties to the accident and the loss of surrounding environmental facilities, and then predict the amount based on the vehicle damage of both parties to the accident and the loss of surrounding environmental facilities.
[0063] Training data set: Contains the transcribed text of the communication recordings between the front-line accident investigators and customers in historical accident cases. These texts record in detail the situation at the accident scene, the damage to both vehicles, and the loss of surrounding facilities.
[0064] Prompt: used to guide the model to focus on specific information points, such as "Please analyze the car damage in the following conversation" or "Please estimate the amount of car damage based on the description."
[0065] Model training: Through training, the model can identify key information in the conversation, determine the damage to the vehicles of both parties involved in the accident and the loss of surrounding facilities, and further combine this information to predict the amount.
[0066] In one implementation of the embodiment of the present application, the large language model training module is based on the large language basic model. When training the motor vehicle accident responsibility judgment task through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set include historical accident responsibility determination documents, which are used to enable the trained large language model to judge the responsibilities of both parties in the accident and provide relevant road traffic safety law basis. The types of responsibilities of both parties in the accident include full responsibility, primary responsibility, secondary responsibility, no responsibility and equal responsibility.
[0067] Training data set: includes historical accident responsibility determination documents, which record in detail the division of responsibilities between the two parties in the accident and the provisions of the Road Traffic Safety Law.
[0068] Model training: Through training, the model can analyze the accident description, determine the type of responsibility of both parties in the accident (full responsibility, primary responsibility, secondary responsibility, no responsibility, equal responsibility), and provide relevant legal basis.
[0069] In one implementation of the embodiment of the present application, the large language model training module is based on the large language basic model. When training the insurance exemption clause judgment task through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set contain exemption clause instances, which are used to enable the trained large model to recognize and determine whether an accident triggers the exemption clause.
[0070] Training dataset: Contains various insurance exemption clause examples and their application scenarios. These examples show the triggering conditions of the exemption clauses in different situations.
[0071] Model training: By learning from these examples, the model can recognize and determine whether a given accident situation triggers a specific exemption clause.
[0072] In one implementation of the embodiment of the present application, the large language model training module is based on the large language basic model. When training the insurance product claim reasoning task through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set include a variety of claim cases, which are used to enable the trained large language model to automatically determine the type of claim, claim ratio and amount.
[0073] Training data set: includes a variety of claims cases, which record in detail the claims ratios, amounts and processes for different types of insurance and under different circumstances.
[0074] Model training: Through training, the model can automatically analyze the accident situation, determine the applicable compensation insurance type, compensation ratio and amount, and improve the accuracy and efficiency of compensation processing.
[0075] In one implementation of the embodiment of the present application, the system further includes a user interaction interface for receiving accident information input by the user, displaying the claim assessment process and results, and providing claim progress tracking and result query functions. Figure 2 As shown, Figure 2 A schematic diagram of the structure of a user interaction interface provided in an embodiment of the present application, including an information input module 201 and a result display module 202. The text obtained by processing the data preprocessing module is input as input information into the information input module 201, and after being processed by a pre-trained large language model, the claim assessment process and the final auto insurance claim assessment report can be displayed in the result display module. The existence of the user interaction interface makes the entire claim process more transparent and convenient, and users can understand the claim progress and results at any time, thereby improving user satisfaction and trust.
[0076] The present application also provides a method embodiment that is consistent with the above embodiment. The explanation based on the same name meaning is the same as the above embodiment, and has the same technical effect as the above embodiment, which will not be repeated here. Figure 3 A flowchart of a method for evaluating automobile insurance claims based on a large language model provided in an embodiment of the present application. Figure 3 As shown, the auto insurance claim assessment method based on a large language model provided in the embodiment of the present application may include:
[0077] S301. Obtain the recordings of the communication between the front-line accident investigators and the customers and transcribe them into text;
[0078] S302, using a pre-trained large model to perform accident loss judgment and vehicle damage amount estimation, vehicle insurance accident judgment, insurance exemption clause judgment and insurance product compensation reasoning analysis based on the transcribed text;
[0079] S303: Generate a motor vehicle insurance claim assessment report based on the analysis results of the pre-trained large model, wherein the assessment report includes an accident overview, loss situation, liability determination, analysis of the applicability of exemption clauses, and compensation recommendations.
[0080] In one implementation of the embodiment of the present application, the method further includes performing data cleaning and feature extraction.
[0081] In one implementation of the embodiment of the present application, the large language model training module is based on the large language basic model. When training the accident loss situation judgment and vehicle damage amount estimation tasks through preset prompt words and training data sets, the preset prompt words and training data sets include transcribed text of communication recordings between front-line accident investigators and customers in historical accident cases, so that the trained large model can judge the vehicle damage of both parties to the accident and the loss of surrounding environmental facilities, and then make amount predictions based on the vehicle damage of both parties to the accident and the loss of surrounding environmental facilities.
[0082] In one implementation of the embodiment of the present application, the large language model training module is based on the large language basic model. When training the motor vehicle accident responsibility judgment task through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set include historical accident responsibility determination documents, so that the trained large language model can judge the responsibilities of both parties in the accident and give relevant road traffic safety law basis. The types of responsibilities of both parties in the accident include full responsibility, primary responsibility, secondary responsibility, no responsibility and equal responsibility.
[0083] In one implementation of the embodiment of the present application, the large language model training module is based on the large language basic model. When training the insurance exemption clause judgment task through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set contain exemption clause instances, so that the trained large model can recognize and determine whether the accident triggers the exemption clause.
[0084] In one implementation of the embodiment of the present application, the large language model training module is based on the large language basic model. When training the insurance product claim reasoning task through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set include a variety of claim cases, so that the trained large language model automatically determines the type of claim, claim ratio and amount.
[0085] In one implementation of the embodiment of the present application, the method further includes receiving accident information input by a user through a user interaction interface, displaying the claim assessment process and results through the user interaction interface, and obtaining claim progress tracking and query results.
[0086] The embodiments of the present application also provide corresponding devices and computer storage media for implementing the solutions provided by the embodiments of the present application.
[0087] The device includes a memory and a processor, the memory is used to store instructions or codes, and the processor is used to execute the instructions or codes so that the device executes the method described in any embodiment of the present application.
[0088] The computer storage medium stores codes, and when the codes are executed, a device executing the codes implements the method described in any embodiment of the present application.
[0089] Through the description of the above implementation methods, it can be known that those skilled in the art can clearly understand that all or part of the steps in the above-mentioned embodiment method can be implemented by means of software plus a general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a storage medium, such as a read-only memory (ROM) / RAM, a magnetic disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network communication device such as a router) to execute the methods described in each embodiment of the present application or some parts of the embodiments.
[0090] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0091] It should also be noted that the various embodiments in this specification are described in a progressive manner, and the same and similar parts between the various embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments. The device and system embodiments described above are merely schematic, in which the units described as separate components may or may not be physically separated, and the components indicated as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0092] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
Claims
1. A car insurance claim assessment system based on a large language model, characterized in that: The system includes a data preprocessing module, a large language model training module, a multi-task processing module and a result output module, wherein: The data preprocessing module is used to transcribe the communication recordings between the front-line accident investigators and the customers into text, and extract information based on preset keywords; The large language model training module is used to train the accident loss situation judgment and vehicle damage amount estimation tasks, the auto insurance accident responsibility judgment tasks, the insurance exemption clause judgment tasks and the insurance product compensation reasoning tasks based on the large language basic model through the preset prompt word prompt and the training data set; The multi-task processing module is used to use the trained large language model to simultaneously perform the accident loss situation judgment and vehicle damage amount estimation tasks, the auto insurance accident responsibility judgment task, the insurance exemption clause judgment task and the insurance product compensation reasoning task, so as to judge the accident loss situation, estimate the vehicle damage amount, judge the accident responsibility, review the insurance exemption clause and reason the insurance product compensation; The result output module is used to integrate the processing results of each task and generate a car insurance claim assessment report.
2. The system according to claim 1, characterized in that The data preprocessing module is also used for data cleaning and feature extraction.
3. The system according to claim 1, characterized in that The large language model training module is based on the large language basic model. When training the accident loss situation judgment and vehicle damage amount estimation tasks through preset prompt words and training data sets, the preset prompt words and training data sets include transcribed text of communication recordings between front-line accident investigators and customers in historical accident cases, which are used to enable the trained large model to judge the vehicle damage of both parties in the accident and the loss of surrounding environmental facilities, and then predict the amount based on the vehicle damage of both parties in the accident and the loss of surrounding environmental facilities.
4. The system according to claim 1, characterized in that The large language model training module is based on the large language basic model. When training the motor vehicle accident responsibility judgment task through preset prompt words and training data sets, the preset prompt words and training data sets include historical accident responsibility determination documents, which are used to enable the trained large language model to judge the responsibilities of both parties in the accident and provide relevant road traffic safety law basis. The types of responsibilities of both parties in the accident include full responsibility, primary responsibility, secondary responsibility, no responsibility and equal responsibility.
5. The system according to claim 1, characterized in that The large language model training module is based on the large language basic model. When training the insurance exemption clause judgment task through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set contain exemption clause instances, which are used to enable the trained large model to recognize and determine whether an accident triggers the exemption clause.
6. The system according to claim 1, characterized in that The large language model training module is based on the large language basic model. When training the insurance product claim reasoning task through a preset prompt word prompt and a training data set, the preset prompt word prompt and the training data set include a variety of claim cases, which are used to enable the trained large language model to automatically determine the type of claim, the claim ratio and the amount.
7. The system according to any one of claims 1 to 6, characterized in that The system also includes a user interaction interface for receiving accident information input by the user, displaying the claims assessment process and results, and providing claims progress tracking and result query functions.
8. A method for evaluating automobile insurance claims based on a large language model, characterized in that: The method comprises: Obtain recordings of communications between frontline accident investigators and clients and transcribe them into text; Use the pre-trained large model to judge the accident loss and estimate the amount of vehicle damage, judge the auto insurance accident, judge the insurance exemption clauses and analyze the insurance product compensation based on the transcribed text; A motor vehicle insurance claim assessment report is generated based on the analysis results of the pre-trained large model, wherein the assessment report includes an accident overview, loss situation, liability determination, analysis of the applicability of exemption clauses, and compensation recommendations.
9. A computing device, characterized in that The computing device includes: a memory and a processor; The memory is used to store computer programs; The processor is configured to implement the method according to claim 8 when executing the computer program.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to claim 8 is implemented.