Methods for analyzing long sequences, methods for analyzing physical injury judgments, and systems using the same

Abstract

This paper discloses a method for analyzing long text data, a method for determining physical injury, and a system thereof. The method disclosed herein is executed using a composite text analysis module, which includes a Universal Information Extraction (UIE) model and a Universal Text Classification (UTC) model. First, the UIE model extracts multiple short text data related to multiple feature information from the long text data. Then, the UTC model is used to classify the data to obtain structured text data.

Description

Technical Field

[0001] This invention belongs to the field of software technology, specifically a method for processing long text data using a composite text analysis model. Prior Technology

[0002] In the field of property insurance, auto liability insurance is one of the main products. When a policyholder is involved in a car accident, the insurance company will compensate the other party for their losses, including vehicle repair costs, property damage, medical expenses for bodily injuries, and related ancillary costs. Most of these expenses have corresponding receipts for reimbursement, and the insurance company compensates for the actual amount incurred based on the principle of loss compensation. However, it is difficult to assess and quantify the emotional distress and physical and psychological trauma suffered by the victim. Because the degree of suffering varies from person to person and cannot be uniformly measured, in practice, compensation for emotional distress mainly depends on the judge's discretion, and there has long been no unified quantitative standard.

[0003] Currently, claims adjusters primarily rely on their own experience or refer to historical judgments published on the Judicial Yuan's official website when processing settlements for emotional distress compensation. However, the amount of judgment data from the past five years alone reaches a staggering 5.55 million records, with each judgment containing at least 5,000 words. Manually sifting through key paragraphs and determining payout amounts is extremely time-consuming and labor-intensive. Therefore, there is currently no efficient analytical tool capable of processing such a large volume of judgment data, placing a significant burden on claims adjusters. In light of this, there is an urgent need in this field for a long-text data analysis method and system to overcome the shortcomings of previous technologies. Summary of the Invention

[0004] This summary is intended to provide a simplified overview of the present disclosure to enable the reader to gain a basic understanding of it. It is not a complete summary of the present disclosure and is not intended to identify key / critical elements of the embodiments of the invention or to define the scope of the invention.

[0005] This invention relates to a method for analyzing long text data and its application in analyzing physical injury judgment data, in order to improve the efficiency of claims personnel in handling claims cases.

[0006] The first aspect of this invention is a long text data analysis method, which is executed using a composite text analysis model, wherein the composite text analysis model includes a general information extraction model (UIE) and a general text classification model (UTC), and the method includes the following steps: (a) Using a Universal Information Extraction (UIE) model to extract multiple short text data points related to multiple feature information from long text data; and (b) Classify these short text data using the Universal Text Classification (UTC) model to obtain a structured text data.

[0007] According to a non-limiting embodiment of the present invention, the feature information may include, but is not limited to, the location of the injury, the degree of injury, liability for the accident, the age of the parties involved, the type of vehicle driven by the plaintiff, the type of vehicle driven by the defendant, the plaintiff's illegal act, the defendant's illegal act, the plaintiff's financial status, and the defendant's financial status.

[0008] According to one embodiment of the present invention, the long text data has at least about 5,000 characters, for example, about 5,000, 6,000, 7,000, 8,000, 9,000, or more than 10,000 characters. In a preferred embodiment, the long text data has at least about 10,000 characters. According to another embodiment of the present invention, the short text data has about 150-250 characters.

[0009] In addition, the composite text analysis model of this invention can be trained in the following ways: The first training data set was processed using a natural language processing model to obtain a complex number of important variables; and After marking the second training data set with these important variables, the data is input into the composite text analysis model for training.

[0010] The second aspect of this invention relates to a long text data analysis system. The system operates within a computer host and includes a processor and a memory electrically connected to each other. The memory stores computer-readable instructions that, when executed, can be used to perform the methods described in any of the above embodiments.

[0011] Another aspect of this invention relates to a method for analyzing bodily injury determination. The method includes the following steps: (a) Obtain the judgment data and label it with a plurality of important variables to obtain the labeled judgment data; (b) Using a composite text analysis model to process labeled decision data to obtain physical injury classification information, wherein the composite text analysis model consists of a Universal Information Extraction (UIE) model and a Universal Text Classification (UTC) model; and (c) The labeled judgment data is analyzed using the amount extraction model to obtain the judgment amount and correlate it with the physical injury classification data to obtain structured judgment data.

[0012] Specifically, in the above method, step (b) involves using a Universal Information Extraction (UIE) model to extract multiple short text data related to multiple feature information in the judgment data; and classifying these short text data using the Universal Text Classification (UTC) model to obtain physical injury classification information.

[0013] The composite text analysis model in the physical injury judgment analysis method of this invention can also be trained in the following way: The first training data set was processed using a natural language processing model to obtain a complex number of important variables: and After marking the second training data set with these important variables, the data is input into the composite text analysis model for training.

[0014] Furthermore, in one embodiment of the present invention, step (a) involves labeling using artificial intelligence technology.

[0015] In a non-limiting implementation, the significant variables include a plurality of primary variables and a plurality of secondary variables, wherein the primary variables are selected from the group consisting of: compensation for emotional distress, description of bodily injury, and liability for the accident, etc. Furthermore, the secondary variables are selected from the group consisting of: increased living expenses, court location, and illegal acts.

[0016] According to one embodiment of the present invention, the judgment data consists of multiple entries, and after steps (a) to (c), the adjudication amount and bodily injury classification data are obtained respectively, and a range of adjudication amounts is generated based on the adjudication amounts of these judgment data.

[0017] Furthermore, according to a third aspect of the present invention, there is a system for analyzing physical injury determination. The system includes a processor and a memory electrically connected to each other, wherein the memory stores computer-readable instructions that, when executed, can be used to perform the methods described in any of the above embodiments.

[0018] After reading the following embodiments, those skilled in the art will easily understand the basic spirit and other inventive objectives of the present invention, as well as the technical means and implementation methods adopted by the present invention. Simple Explanation of the Diagram

[0019] To make the above and other objects, features, advantages and embodiments of the present invention more apparent and understandable, the accompanying drawings are described below:

[0020] Figure 1 is a schematic diagram of a computer host 100 running the system or method of the present invention according to an embodiment of the present invention; Figure 2 is a flowchart of a long text data analysis method according to an embodiment of the present invention; Figure 3 illustrates a method for analyzing bodily injury determination according to another embodiment of the present invention; Figure 4 is a schematic diagram illustrating the application of the physical injury assessment analysis method shown in Figure 3; Figure 5 is a schematic diagram of a decision retrieval system 200 according to an embodiment of the present invention; and Figure 6 is a flowchart of the integrated physical injury judgment analysis method of the present invention, which is executed using the judgment retrieval system 200.

[0021] As is customary practice, the various features and elements in the drawings are not drawn to scale. The drawing method is intended to best represent the specific features and elements related to the present invention. Furthermore, similar elements / components are referred to by the same or similar element symbols across different drawings. Implementation

[0022] To make the description of this disclosure more detailed and complete, illustrative descriptions of embodiments and specific examples of the present invention are provided below; however, these are not the only forms of implementing or utilizing the specific examples of the present invention. The embodiments cover features of multiple specific examples and methods and steps for constructing and operating these specific examples, as well as their order. However, other specific examples may also be used to achieve the same or equivalent functions and order of steps.

[0023] Unless otherwise defined in this specification, the scientific and technical terms used herein have the same meaning as understood and commonly used by one of ordinary skill in the art to which this invention pertains. Furthermore, unless conflicting with the context, singular nouns used herein encompass their plural forms, and vice versa.

[0024] To make the description of this disclosure more detailed and complete, illustrative descriptions of embodiments and specific examples of the present invention are provided below; however, these are not the only forms of implementing or utilizing the specific examples of the present invention. The embodiments cover features of multiple specific examples and methods and steps for constructing and operating these specific examples, as well as their order. However, other specific examples may also be used to achieve the same or equivalent functions and order of steps.

[0025] As is understood, the "system" typically includes at least some form of storage media, communication unit, and processing unit. The storage media includes electrically and non-electrically oriented, removable and non-removable media, which can be used with appropriate methods or techniques to store desired information (e.g., computer-readable instructions, data structures, application modules, and other data). Storage units include, but are not limited to, RAM, ROM, EEPROM, flash memory, or other memory technologies, CD-ROM, digital versatile discs (DVDs), or other optical storage, magnetic cartridges, magnetic tapes, magnetic disk storage, and other magnetic storage devices, or any other media capable of storing required information and accessible to the processor. Generally, the communication unit can materialize computer-readable instructions, data, structures, application modules, and other data into various data signals, and can transmit them through any communication media. By way of example and not limitation, the communication unit includes wired media (such as wired networks or direct wired connections) and wireless media (such as sound waves, infrared, radio waves, microwaves, spread spectrum technology, and other wireless media technologies). Furthermore, the communication unit is implemented using a communication chip, such as a component supporting signal transmission for Ethernet, fiber optic communication networks, telecommunication cable networks, Global System for Mobile communication (GSM), Personal Handyphone System (PHS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), Worldwide Interoperability for Microwave Access (WiMAX), Wireless Fidelity (Wi-Fi), Bluetooth, or other chips supporting standard communication specifications defined by the Institute of Electrical and Electronics Engineers (IEEE). The processing unit communicates with the communication unit and storage unit to perform the necessary calculations and functions for commercial insurance underwriting risk assessment methods.In one embodiment of this disclosure, the processing unit may be, for example, a Central Processing Unit (CPU), or other programmable general-purpose or special-purpose microprocessors, digital signal processors (DSPs), programmable controllers, application-specific integrated circuits (ASICs), or other similar components or combinations thereof; this disclosure is not limited thereto.

[0026] By way of example and not limitation, the "system" operates via a computer device, which may be a desktop computer, server computer, handheld or laptop device, personal digital assistant, multiprocessor system, microprocessor-based system, set-top box, programmable consumer electronics, mobile phone (especially smartphone), network computer, minicomputer, mainframe computer, distributed computing environment including any of the above systems or devices, and similar devices. In one particular embodiment, the "system" operates within a computer mainframe.

[0027] The “composite text analysis model” and “monetary analysis model” mentioned herein are based on the LLM (Large Language Model) to achieve the tasks required by the Universal Information Extraction Model (UIE) and / or the Universal Text Classification Model (UTC).

[0028] The term "user" as used herein refers to an individual who uses the long text data analysis method, physical injury judgment analysis method, or system of the present invention, such as an insurance agent. In a preferred embodiment, the user refers to a claims adjuster.

[0029] To improve the efficiency of insurance operations, this invention provides a method for analyzing long text data, a method for analyzing bodily injury judgments, and a system for executing the methods. This invention utilizes a novel computational process to efficiently process large-scale long text data with limited computer performance and generate structured text data. Furthermore, using the technical features of this invention in bodily injury judgment analysis, it is possible to summarize and present the relevant claim amounts for bodily injury cases, particularly compensation for emotional distress, for claims personnel to use as a reference.

[0030] Figure 1 shows a computer host 100 that executes the long text data analysis system or the physical injury judgment analysis system of the present invention. The system of the present invention operates in the computer host 100, which includes a processor 110 and a memory 120 electrically connected to each other. The memory 120 stores a computer-readable instruction, which, when executed, can be used to perform the long text data analysis or physical injury judgment analysis method disclosed in the present invention.

[0031] This invention relates to a method for efficiently processing and analyzing long text data. It primarily utilizes a composite text analysis model that integrates a Universal Information Extraction (UIE) model and a Universal Text Classification (UTC) model. Referring to Figure 2, the method proposed in this invention processes long text data in two stages. First, the UIE model is used to extract multiple short text data points related to multiple feature information points from the long text data (step 202). For example, the feature information may include, but is not limited to, the location of the injury, the degree of injury, liability for the accident, the age of the parties involved, the type of vehicle driven by the plaintiff, the type of vehicle driven by the defendant, the plaintiff's illegal act, the defendant's illegal act, the plaintiff's financial status, and the defendant's financial status. For example, in one embodiment of this invention, the UIE model extracts multiple short text data points based on the feature information of the location of the injury, the degree of injury, the age of the parties involved, and the type of vehicle driven by the defendant. Examples include open pelvic fracture, plaintiff under 18 years old, and defendant driving a private passenger car. Those skilled in the art will understand that specific feature information can be selected depending on the purpose of long text data analysis. After long text data is extracted and segmented based on feature information using a general information extraction module, multiple short text data containing the aforementioned feature information are obtained. Next, the multiple short text data are classified using a Universal Text Classification (UTC) model to obtain structured text data (step 204).

[0032] In an optional implementation, the long text data has at least about 5,000 characters, for example, about 5,000, 6,000, 7,000, 8,000, 9,000, or 10,000 characters, and the short text data has about 150-250 characters.

[0033] It should be noted that the long text data analysis method of the present invention, through the integration of a universal information extraction model (UIE) and a universal text classification model (UTC) in a two-stage process, can run on limited computer performance and improve the efficiency of long text data analysis.

[0034] Furthermore, the composite text analysis model of the present invention can be trained by processing a first training data set using a natural language processing model to obtain a plurality of important variables: and After labeling the second training data set with these key variables, the data is input into the composite text analysis model for training. In an optional implementation, the first and / or second training data sets each contain text data of multiple lengths.

[0035] [Experimental Example]

[0036] The method described in Figure 2 of this invention is analyzed using 1000 manually labeled data points for model training. Of these, 800 are training data, 100 are test data during model training, and the last 100 are the final test data after model training is complete. The table below shows the accuracy of the final model training test data.

[0037] Therefore, the results of this experiment can be found in Table 1. It can be seen that the composite text analysis of this invention is better than other computational models in terms of precision, recall and F1.

[0038] [surface] [1] Information Extraction Model Precision Recall rate (Recall) F1 Baseline (Zero Shot UIE) 0.22 0.38 0.28 UIE 0.80 0.83 0.82 GPT 3.5 0.26 0.06 0.1 Text classification Model Baseline (Ernie) 0.4 0.46 0.43 Chunking (Ernie) 0.58 0.56 0.57 Recurrent (XLNet) 0.63 0.46 0.53 UTC (Truncated) 0.82 0.56 0.66 Composite Text Analysis Model (UIE+UTC) 0.87 0.89 0.88 GPT3.5 0.76 0.35 0.70

[0039] Figure 3 illustrates a bodily injury judgment analysis method according to another embodiment of the present invention. The bodily injury judgment analysis method is based on the long text data analysis method in Figure 2, which analyzes judgment data related to bodily injury to generate structured judgment data for subsequent analysis and application.

[0040] In the field of property insurance, the claims process primarily follows the principle of reimbursement, meaning that expenses arising from property loss or injury are mainly reimbursed based on invoices. However, for bodily injury cases, there is no clear standard for determining the amount of compensation for emotional distress. It typically requires consideration of numerous factors, including the facts of the case and the background of the parties involved, with the final amount determined by the judge's opinion. Therefore, without established standards, claims adjusters often spend considerable time searching historical judgments for reference when processing cases, a very time-consuming process. This invention proposes a novel method for analyzing bodily injury judgments, which, with limited computer resources, can obtain structured judgment data through a composite text analysis module.

[0041] Referring to Figure 3, firstly, multiple judgment data are obtained and labeled with multiple important variables to obtain labeled judgment data (step 302). Next, the labeled judgment data is processed using a composite text analysis model; specifically, an information extraction model (UIE) is used to extract multiple short text data related to multiple feature information from the judgment data. These short text data are then classified using a Universal Text Classification (UTC) model to obtain physical injury classification information (step 304). In fact, the calculation process of step 304 is the same as that of the long text data analysis shown in Figure 2, so it will not be described again here. Furthermore, the labeled judgment data is analyzed using a monetary extraction model to obtain a judgment amount and correlate it with the physical injury classification data to obtain structured judgment data, wherein the monetary extraction model is the Universal Information Extraction (UIE) model (step 306). In a specific embodiment, the judgment amount includes information on mental distress compensation and other judgment amounts.

[0042] In a non-limiting embodiment, the marking process in step 302 can be achieved through artificial intelligence technology, such as using Chat GPT. Furthermore, the structured judgment data generated by the method of this invention can be further analyzed and applied. For example, by obtaining multiple judgment amount results, a judgment amount range can be statistically determined, or further, an artificial intelligence model can be combined to predict the judgment amount range. In a specific embodiment, the judgment amount can be selected according to the user's actual use, such as compensation for mental distress or damages. In one embodiment, the judgment amount range is a range for compensation for mental distress, and a statistical distribution of compensation for mental distress is provided, such as median, average, minimum, or maximum, preferably displayed in a chart for easy reading by the user. Furthermore, those skilled in the art should understand that a corresponding judgment amount range can be generated by filtering the location or severity of bodily injury classification data.

[0043] In addition, in the method shown in this embodiment, the composite text analysis model and / or the amount extraction model can also be trained in the following way: the first training data set is processed by a natural language processing model to obtain a plurality of important variables; and the second training data set is labeled based on these important variables and then input into the composite text analysis model for analysis and training.

[0044] Figure 4 is a schematic diagram illustrating the application of the bodily injury judgment analysis method of the present invention according to a specific embodiment of the invention. In this embodiment, claims adjusters need to search for relevant expenses related to specific injuries in historical judgments, especially compensation for emotional distress, as a reference for case claims or settlements. Please refer to Figures 3 and 4 simultaneously. The method of the present invention can be divided into two main frameworks: using a composite text analysis model to classify text and obtain bodily injury classification information, and using a monetary analysis model to analyze monetary information and obtain various monetary information. Finally, the bodily injury classification information and monetary information are correlated to obtain structured judgment data 402.

[0045] Specifically, the judgment data is first labeled with a plurality of important variables to obtain labeled judgment data (step 302). In an optional implementation, the important variables include primary variables and secondary variables, wherein the primary variables include the following groups: compensation for mental distress, description of bodily injury, and liability for the accident, etc. The secondary variables include the following groups: increased living expenses, court location, and illegal acts, etc. Next, the labeled judgment data is processed using a composite text analysis model, which is carried out in two stages. First, a UIE model is used to extract a plurality of short text data related to a plurality of feature information in the judgment data. Then, these short text data are classified using a Universal Text Classification (UTC) model to obtain bodily injury classification information (step 304). The feature information may include the location of bodily injury, the degree of injury, liability for the accident, the age of the parties, the type of vehicle driven by the plaintiff, the type of vehicle driven by the defendant, the illegal acts of the plaintiff, the illegal acts of the defendant, the financial status of the plaintiff and / or the financial status of the defendant. Meanwhile, the long text data is analyzed using a monetary extraction model, which is an Information Extraction (UIE) model. Depending on the purpose, it extracts monetary-related feature information, such as compensation for emotional distress, descriptions of bodily injury, and liability for the accident. Furthermore, in a preferred embodiment, the compensation for emotional distress is the amount ordered by the court. This is achieved by using the monetary extraction module to understand the semantics of the text and extracting the amount of compensation for emotional distress from the relevant paragraphs of the final court judgment. Finally, the results of text classification and monetary analysis are integrated; that is, the bodily injury classification information obtained in steps 304 and 306 is correlated with the aforementioned monetary information to obtain structured judgment data 402, facilitating subsequent analysis and application.

[0046] Furthermore, the method of the present invention can integrate artificial intelligence models to predict the amount awarded by the court, preferably compensation for emotional distress, and can further provide a range of predictions for compensation for emotional distress.

[0047] Furthermore, the long text data analysis method or physical injury judgment analysis method of the present invention can also be applied to the judgment retrieval system disclosed in the applicant's previous Taiwan Utility Model Patent Application No. 112206536, the contents of which are also incorporated into the present invention.

[0048] Figure 5 is a schematic diagram of a judgment retrieval system 200 according to an embodiment of the present invention. As shown in Figure 5, the judgment retrieval system 200 of the present invention operates within the computer host 100 shown in Figure 1. The judgment retrieval system 200 includes a judgment database 210, a communication unit 220, a storage unit 230, and a processing unit 240, which are communicatively connected to each other. According to the embodiment of the present disclosure, the judgment database 210 contains a plurality of judgment records; the communication unit 220 is used to transmit and receive information; the storage unit 230 is used to store multiple modules; and the processing unit 240 is communicatively connected to the communication unit 220 and the storage unit 230, and executes the modules in the storage unit 230.

[0049] According to certain embodiments of this disclosure, the decision retrieval system 200 is connected to a user terminal, and the user transmits operation commands through a personal endpoint device (e.g., desktop computer, server computer, handheld or laptop device, smartphone, etc.) to use the decision retrieval system 200 to perform a retrieval.

[0050] According to certain embodiments of this disclosure, the judgment database 210 contains a plurality of judgment records derived from historical judgment files in a publicly accessible database. In a preferred embodiment, the publicly accessible database is a court judgment database. In some embodiments, the judgment database 210 may be updated periodically or immediately, manually or automatically, to obtain the latest judgment files; for example, the judgment database 210 may be updated daily or semi-annually, depending on the frequency of judgment file updates in the court judgment database, or it may be set to be updated weekly, bi-weekly, monthly, bi-monthly, quarterly, or annually, depending on usage needs. In alternative embodiments, the judgment database 210 is communicatively connected to and monitors the court judgment database in real time. In this case, the update status of the court judgment database is checked periodically, and the data in the judgment database 210 is updated in batches to obtain the latest judgment files.

[0051] Furthermore, to improve the efficiency of users (i.e., claims personnel) in filtering valuable judgment data, the judgment database 210 provides additional case information for each of the plurality of judgment data, displaying the content classification of these judgment data and the historical records of their citation by users. In a preferred embodiment of this disclosure, the additional case information includes one or more keyword tags assigned to each judgment data based on the content of the judgment, the number of times each judgment data has been cited in the system, expert annotations made by other users in the judgment database 210, or other case information that can be used as a filtering basis.

[0052] The communication unit 220 is configured to receive retrieval requests, key decision annotation requests, and other operation instructions from the user terminal, and transmit these received requests and / or instructions to the processing unit 240 to initiate the processing unit 240 to execute the corresponding module to complete the user terminal's requests and / or instructions. After completing the user terminal's requests and / or instructions, the communication unit 220 will transmit the results to the user terminal.

[0053] Storage unit 230 is used to store a plurality of modules, wherein these modules are the necessary program code required for the execution of the judgment retrieval system 200, and can each correspond to different requests from the user terminal. According to some embodiments of this disclosure, storage unit 230 includes a retrieval module 232, a key judgment annotation module 234, and a claim amount calculation module 236.

[0054] Please also refer to Figure 6, which is a flowchart of the integration of the bodily injury judgment analysis method of the present invention using the judgment retrieval system 200. It should be noted that the data in the judgment database 210 of the present invention can first be structured using the bodily injury judgment analysis method shown in Figure 3 of the present invention. In a non-limiting embodiment, the user communicates with the judgment retrieval system 200 through a personal device (i.e., user terminal), sets one or more search conditions according to case requirements (e.g., type of compensation (e.g., damages, tort damages), case type (e.g., civil, criminal), expense category or amount (e.g., funeral expenses, alimony, mental anguish compensation), etc.), and transmits a search request to the judgment retrieval system 200 of the present invention. The search module 232 outputs at least one judgment document related to the search conditions (step 401). Here, the search module 232 can perform the search according to the search logic described in any of the above embodiments, for example, sorting the relevant judgment documents by relevance based on expert annotations, citation counts, and relevance scores, thereby prioritizing the presentation of the most relevant judgments to the user terminal for viewing. Next, users can annotate the relevant judgment data according to their actual needs if they believe it to be of reference value, through the judgment retrieval system 200. The user terminal issues a key judgment annotation request, and through the key judgment annotation module 234, the relevant judgment data is marked as key judgment data. As needed, the key judgment annotation module 234 further presents some or all of the key judgment data as a key judgment list and transmits it to the user terminal (step 403). Furthermore, the judgment retrieval system 200 of the present invention can further calculate the compensation amount range of the key judgments. Specifically, the compensation amount calculation module 236 uses the bodily injury judgment analysis method in Figure 3 of the present invention to statistically obtain the compensation amount range of the key judgments. Furthermore, the compensation amount corresponding to the at least one key judgment list, preferably multiple key judgments, is calculated, and a reasonable estimated compensation amount (including the average, median, maximum, and minimum compensation amount) is calculated according to different types of compensation and transmitted to the user terminal (step 405). Furthermore, the judgment retrieval system 200 can also use artificial intelligence technology to predict the range of the judgment amount. Therefore, the judgment retrieval system 200 of this invention can assist users in retrieving judgment information and estimating compensation amounts.

[0055] In summary, the long text data analysis method, physical injury judgment analysis method, and system proposed in this invention, through vertical integration of the Universal Information Extraction (UIE) model and the Universal Text Classification (UTC) model, and horizontal merging of a monetary analysis model based on the UIE model, can effectively process long text data and generate structured data for subsequent applications and analysis. Furthermore, the methods and systems proposed in this invention can assist users in analyzing the range of compensation for emotional distress in historical judgments and provide statistical indicators for reference, thereby predicting reasonable compensation amounts for emotional distress and improving the efficiency and accuracy of claims processing.

[0056] The main component symbols of this invention are listed below: 100: Computer host 110: Processor 120: Memory Steps 202, 204, 302-306, 401-405: 200: Judgment Retrieval System 210: Database 220: Communication Unit 230: Storage unit 232: Retrieval Module 234: Key Decision Annotation Module 236: Claim Amount Calculation Module 240: Processing Unit 402: Structured Judgment Data

Claims

1. A method for analyzing bodily injury judgments, comprising the following steps: (a) using a judgment retrieval system to search a judgment database based on one or more search criteria input by a user, and obtaining multiple judgment records related to the search criteria, wherein the search criteria include case type and compensation type; (b) sorting the multiple judgment records obtained in step (a) according to expert annotations, citation counts, and relevance scores, and using artificial intelligence technology to label the multiple judgment records with multiple important variables to obtain multiple labeled key judgment records; (c) processing the multiple labeled key judgment records using a composite text analysis model, wherein firstly, a Universal Information Extraction (UIE) model is used to extract and segment the multiple labeled judgment records to obtain multiple short text records containing feature information, and then a Universal Text Classification (UTC) model is used to classify the multiple short text records containing feature information to obtain bodily injury classification information; (d) The data is analyzed using a monetary extraction model, and the data is extracted using features related to monetary amounts to obtain monetary information related to the court's rulings. (e) The bodily injury classification information obtained in step (c) is correlated with the monetary information related to the court's rulings obtained in step (d) to obtain a structured data for statistical analysis, thereby providing a predicted range for the amount of mental distress compensation awarded in bodily injury cases.

2. The method as described in claim 1, wherein the significant variables comprise a plurality of primary variables and a plurality of secondary variables, wherein the primary variables comprise the group consisting of: compensation for emotional distress, description of bodily injury, and liability for the accident.

3. The method as described in claim 2, wherein the secondary variable comprises the group consisting of: increased living expenses, court location, and illegal acts.

4. The method as described in claim 1, wherein the feature information is selected from one or more feature information selected from a group consisting of the body part, degree of injury, liability for the accident, age of the parties, type of vehicle driven by the plaintiff, type of vehicle driven by the defendant, illegal acts of the plaintiff, illegal acts of the defendant, financial status of the plaintiff, and financial status of the defendant.

5. The method as described in claim 1, wherein the information related to the amount includes compensation for emotional distress, salary income, and medical expenses.

6. The method as described in claim 1, wherein the judgment data is greater than 5,000 characters and the plurality of short text data containing the feature information are each about 150-250 characters.

7. The method as described in claim 1, wherein prior to step (b) the steps of: processing a first training data set using a natural language processing model to obtain a plurality of important variables; and labeling a second training data set based on the important variables and inputting it into the composite text analysis model for analysis and training.

8. A physical injury determination analysis system, operating within a computer host, the computer host including a processor and a memory electrically connected to each other, wherein the memory stores a computer-readable instruction, which, when executed, can be used to perform the method described in any one of claims 1-7.

Images