Public benefit litigation case key element extraction method and system based on large model

Abstract

The invention provides a public benefit litigation case key element extraction method and system based on a large model, and belongs to the technical field of intelligent judicial, and the method comprises the steps: obtaining multi-modal data of a public benefit litigation case; matching the multi-modal data based on a preset time interval to obtain a time matching data set of the time sequence data set; and inputting the time matching data set into a preset large model, extracting key elements of the public benefit litigation case, and generating the key elements into output data of a preset structure. A time series data set for understanding time development veins of public benefit litigation cases is provided for a preset large model, so that context information of multi-modal data has a long-distance dependency relationship in a process of extracting key elements through the preset large model; the preset large model can accurately judge the relation between the key elements according to the time sequence, and the accuracy of key element extraction is further improved.

Description

Technical Field

[0001] This application relates to the technical field of intelligent justice, and specifically to a method and system for extracting key elements of public interest litigation cases based on large models. Background Art

[0002] A public interest litigation case refers to a litigation activity initiated by a specific organ or social organization stipulated by law, aiming to safeguard national interests, social public interests, or the legitimate rights and interests of an indefinite number of people. Its core lies in providing judicial relief for acts that damage public interests, rather than protecting individual private rights.

[0003] Chinese Patent CN113220888B discloses a method and system for extracting case clue elements based on the Ernie model. The solution of this patent is as follows: Input the clue text into the clue classification unit to obtain the clue type of the clue text; the clue classification unit consists of the Ernie model and keyword matching, adjust the weight output by the Ernie model, and classify according to the weight; divide the clue text into a single sentence set S1 and input it into the named entity recognition unit in sequence to identify the entities therein; input the elements in the single sentence set S1 into the illegal act and illegal consequence extraction unit in sequence to obtain the illegal act elements and illegal consequence elements; organize and integrate the information according to the elements, and organize and integrate the clue type, entities, illegal act elements, and illegal consequence elements to obtain the element extraction result.

[0004] The above-mentioned existing technologies have the following problems: The Ernie model only focuses on the information of single sentences in the case, lacks the ability to understand the overall context of the case, and thus reduces the accuracy of extracting key elements of the case; the Ernie model processes the content of the clue text and lacks the ability to process multi-modal information, making the extraction of key elements of the case not comprehensive enough. Therefore, how to overcome the above-mentioned technical problems and defects has become a problem that needs to be solved. Summary of the Invention

[0005] In order to overcome the above problems existing in the prior art, this application provides a method and system for extracting key elements of public interest litigation cases based on large models, and adopts the following technical solutions:

[0006] In the first aspect, this application provides a method for extracting key elements of public interest litigation cases based on large models, including:

[0007] Obtain multi-modal data of public interest litigation cases;

[0008] Match the multi-modal data based on a preset time interval to obtain a time matching data set of the time series data set;

[0009] Input the time-matched dataset into a pre-set large model to extract the key elements of public interest litigation cases, and generate output data in a pre-set structure based on the key elements.

[0010] Furthermore, obtain the multi-modal data information of public interest litigation cases, where the multi-modal data information includes the text information, image information, video information, and audio information of public interest litigation cases.

[0011] Furthermore, before matching the multi-modal data based on a pre-set time interval, it is necessary to preprocess the text information, image information, video information, and audio information.

[0012] Furthermore, during the preprocessing of multi-modal data, when obtaining the time information of multi-modal data, annotate the time information in the multi-modal data information.

[0013] Furthermore, match the multi-modal data based on a pre-set time interval to obtain the time-matched dataset of the time series dataset, including:

[0014] Extract time-related features from the preprocessed time series dataset;

[0015] Convert the time-related features extracted from the time series dataset into a unified format of time data representation;

[0016] Based on the unified format of time data representation, match the time series dataset based on a pre-set time interval to obtain the time-matched dataset of the time series dataset;

[0017] Construct the time series data structure of the time-matched dataset based on the time order.

[0018] Furthermore, before matching the time series dataset based on a pre-set time interval, it also includes sorting the unified format of time data representation according to the development context of time to obtain the time series dataset of multi-modal data.

[0019] Furthermore, input the time-matched dataset into a pre-set large model to extract the key elements of public interest litigation cases, and generate output data in a pre-set structure based on the key elements. The specific content includes:

[0020] Obtain the feature vectors of each modality data in the time-matched dataset, fuse the feature vectors of each modality data through the attention mechanism, and integrate the time information of each modality data into the feature vectors based on the position encoding to obtain the comprehensive feature vector sequence of the time-matched dataset;

[0021] Use the comprehensive feature vector sequence as the input data of the preset large model, and through the multi-layer neural network of the preset large model, perform non-linear transformation on the input data to map low-level features to high-level semantic features;

[0022] Perform key element prediction on the high-level semantic features through the output layer of the preset large model, and use the key elements that meet the preset threshold as the key element extraction results of the public interest litigation cases.

[0023] Convert the key element extraction results into output data of a preset structure.

[0024] In a second aspect, the present application also provides a key element extraction system for public interest litigation cases based on a large model, including:

[0025] A multi-modal data acquisition module for acquiring multi-modal data of public interest litigation cases;

[0026] A time series data acquisition module for matching the multi-modal data based on a preset time interval to obtain a time matching data set of the time series data set;

[0027] A key element extraction module for inputting the time matching data set into the preset large model to extract the key elements of the public interest litigation cases and generating output data of a preset structure from the key elements.

[0028] In a third aspect, the present application provides an electronic device, including:

[0029] One or more processors; a memory; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions that, when executed by the device, cause the device to execute the method described in the first aspect.

[0030] In a fourth aspect, the present application provides a computer-readable storage medium, in which a computer program is stored, and when it runs on a computer, it causes the computer to execute the method described in the first aspect.

[0031] In a fifth aspect, the present application provides a computer program that, when executed by a computer, is used to execute the method described in the first aspect.

[0032] In a possible design, the program in the fifth aspect can be stored in whole or in part on a storage medium packaged together with the processor, or can be stored in whole or in part on a memory not packaged together with the processor.

[0033] The present application has the following beneficial effects:

[0034] 1. This application obtains multi-modal data of public interest litigation cases, so that in the process of extracting key elements of public interest litigation cases, it no longer only relies on text information, but extracts key elements through more comprehensive multi-modal data, improving the accuracy of key element extraction.

[0035] 2. This application matches multi-modal data at preset time intervals to obtain a time-matched data set of the time series data set, inputs the time-matched data set into a preset large model, extracts the key elements of public interest litigation cases, and generates output data of a preset structure for the key elements. By matching multi-modal data at preset time intervals, this application provides a time series data set for the preset large model to understand the time development context of public interest litigation cases, enabling the preset large model to have a long-distance dependence on the context information of multi-modal data during the process of extracting key elements, so that the preset large model can accurately judge the relationship between key elements according to the chronological order, further improving the accuracy of key element extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is an exemplary system architecture diagram to which the embodiments of this application can be applied;

[0037] Figure 2 It is a flowchart of the method for extracting key elements of public interest litigation cases of the large model in the embodiments of this application;

[0038] Figure 3 It is a flowchart of the preprocessing of text information of the large model in the embodiments of this application;

[0039] Figure 4 It is a flowchart of obtaining the time series data set in the embodiments of this application;

[0040] Figure 5 It is a flowchart of obtaining the time-matched data set of multi-modal data in the embodiments of this application;

[0041] Figure 6 It is a system flowchart of the embodiments of this application;

[0042] Figure 7 It is a schematic diagram of a computer device in the embodiments of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application belongs; the terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above drawings are intended to cover non-exclusive inclusion. The terms "first", "second", etc. in the specification and claims of this application or the above drawings are used to distinguish different objects and not to describe a specific order.

[0044] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The phrase does not necessarily refer to the same embodiment at every occurrence in the specification, nor is it an independent or alternative embodiment mutually exclusive of other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0045] In order to enable those skilled in the art of this technology to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0046] As Figure 1 shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0047] Users can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0048] The terminal devices 101, 102, 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, and desktop computers, etc.

[0049] Server 105 can be a server that provides various services, such as a background server that supports the pages displayed on the terminal devices 101, 102, and 103.

[0050] It should be noted that the method for extracting key elements of public interest litigation cases provided by the embodiments of the present application is generally executed by a server / terminal device. Correspondingly, the system for extracting key elements of public interest litigation cases based on a large model is generally set in the server / terminal device.

[0051] It should be understood that Figure 1 the numbers of terminal devices, networks, and servers in are merely illustrative. According to actual needs, there can be any number of terminal devices, networks, and servers.

[0052] Continuing to refer to Figure 2 , the figure shows a flowchart of a method for extracting key elements of public interest litigation cases based on a large model according to the present application. The method includes the following steps:

[0053] Step 201, obtain multi-modal data information of a public interest litigation case.

[0054] In a possible implementation, obtain multi-modal data information of a public interest litigation case, where the multi-modal data information includes text information, image information, video information, and audio information of the public interest litigation case. The text information includes case descriptions, indictments, judgments, tip-off letters, news reports, social media posts, etc.; the image information includes photos of the scene, evidence pictures, etc.; the video information includes on-site investigation videos, etc.; the audio includes witness testimonies, etc.

[0055] Step 202, match the multi-modal data based on a preset time interval to obtain a time-matched data set of the time series data set.

[0056] In the process of extracting key elements from current public interest litigation cases, feature extraction is performed on text information, or keyword matching is performed based on single-sentence information of the case. However, this method ignores the time development context of public interest litigation cases, resulting in a reduction in the accuracy of extracting clue features of public interest litigation cases during the feature extraction process. Therefore, the present application matches the multi-modal data based on a preset time interval to obtain a time-matched data set of the time series data set, improving the accuracy of extracting key elements of public interest litigation cases.

[0057] If the multimodal data is directly matched based on a preset time interval without preprocessing the multimodal data, then during the process of matching the multimodal data based on the preset time interval, the processed data is relatively messy, and there will be a lot of noise. These problems bring a great deal of workload to the matching of multimodal data based on the preset time interval, reducing the time sorting efficiency of the multimodal data. Therefore, before matching the multimodal data based on the preset time interval, it is necessary to preprocess the text information, image information, video information, and audio information. By removing the noise, redundant information, and duplicate information in the multimodal data, the amount of data is reduced during the process of sorting the time series of the multimodal data, improving the efficiency of time matching.

[0058] In a possible implementation manner, before matching the multimodal data based on a preset time interval, it is necessary to preprocess the text information, image information, video information, and audio information. For the preprocessing of the text information, please refer to Figure 3 , and the specific content includes:

[0059] Step 31, remove the noise in the text information to obtain the first text information, and remove the noise in the text such as HTML tags, special characters, javascript code, etc.

[0060] Step 32, convert the first text information into text in a unified format to obtain the second text information.

[0061] Step 33, obtain the timestamp or time information describing the occurrence of time in the second text information, and format it into a unified time format to obtain the third text information.

[0062] In a possible implementation manner, before matching the multimodal data based on a preset time interval, it is necessary to preprocess the text information, image information, video information, and audio information. The preprocessing of the image information includes:

[0063] Perform denoising and sharpening processing on the image information, adjust the size ratio of the image, and obtain a standardized image.

[0064] In a possible implementation manner, before matching the multimodal data based on a preset time interval, it is necessary to preprocess the text information, image information, video information, and audio information. The preprocessing of the video information includes: performing key frame segmentation on the video, performing standardized processing on each frame of the segmented image, and adding a timestamp to each frame;

[0065] In a possible implementation, before matching multimodal data based on a preset time interval, it is necessary to preprocess text information, image information, video information, and audio information. Preprocessing the audio information includes: removing background noise from the audio information, enhancing the audio information, and recording the recording time of each audio segment.

[0066] Match the multimodal data based on a preset time interval, and the data used is the preprocessed text data, image data, video data, and audio data.

[0067] In a possible implementation, during the preprocessing of multimodal data, when obtaining the time information of the multimodal data, annotate the time information in the multimodal data information.

[0068] In a possible implementation, before matching the time series data set based on a preset time interval, it also includes sorting the time data representation in a unified format according to the development context of time to obtain the time series data set of multimodal data. Since the multimodal data is sorted by time before matching the time series data set, data matching can be performed based on the process of time development during multimodal data matching, which can improve the efficiency of data matching.

[0069] In a possible implementation, match the multimodal data based on a preset time interval to obtain the time matching data set of the time series data set. Please refer to Figure 4 , and the specific content includes:

[0070] Step 41, extract time-related features from the preprocessed time series data set.

[0071] In a possible implementation, the preprocessed time series data set is annotated with time information, and the time information in the multimodal data can be obtained through the annotated time information field, that is, extract the time-related features in the time series data set.

[0072] Step 42, convert the time-related features extracted from the time series data set into a unified format of time data representation.

[0073] In a possible implementation, convert the time-related features extracted from the time series data set into a unified format of time data representation. For example, the extracted time features can be converted into a unified timestamp or ISO8601 format, where the ISO 8601 format is YYYY-MM-DDTHH:MM:ssZ.

[0074] Step 43: Based on the time data representation in a unified format, match the time series data set based on a preset time interval to obtain the time matching data set of the time series data set.

[0075] In a possible implementation manner, based on the time data representation in a unified format, match the time series data set based on a preset time interval to obtain the time matching data set of the time series data set. Please refer to Figure 5 , specifically manifested as:

[0076] For example, set a preset time interval, such as ±10s, ±20s, or 1 minute, 2 minutes, etc.; organize the multimodal data into a unified data structure, such as modal type (text, image, video, audio), content (specific information), and timestamp.

[0077] Step 51: Obtain the start time and end time of the time series data set. Based on the start time and the preset time interval, traverse the multimodal data in sequence based on the preset time interval until the end time.

[0078] Step 52: In each traversal process, obtain the data information of each modal data within the current time interval.

[0079] Step 53: Match the different modal data within the same time interval; when there are multiple data of a certain modal at a certain time interval, while there is only one data of another modal at a certain time interval, match the different modal data through the data closest to the center time of the interval.

[0080] Step 44: Construct the time series data structure of the time matching data set based on the time order.

[0081] In a possible implementation manner, the time series data structure includes an outer structure and an inner structure, where the outer structure is indexed by the time interval, and the inner structure contains the matched different modal data within that time interval.

[0082] Step 203: Input the time matching data set into a preset large model, extract the key elements of the public interest litigation case, and generate the output data in a preset structure from the key elements.

[0083] In a possible implementation manner, input the time matching data set into a preset large model, extract the key elements of the public interest litigation case, and generate the output data in a preset structure from the key elements. The specific content includes:

[0084] Obtain the feature vectors of each modality data in the time-matching dataset, fuse the feature vectors of each modality data through the attention mechanism, incorporate the time information of each modality data into the feature vectors based on positional encoding, and obtain the comprehensive feature vector sequence of the time-matching dataset.

[0085] In a possible implementation, obtaining the feature vectors of each modality data in the time-matching dataset includes: converting the text data in the time-matching dataset into a text vector representation based on word embedding operations; extracting the visual features of the image data and the key frames of the video data to obtain the feature vectors of the image data and the video data; extracting the Mel-frequency cepstral coefficients of the audio data to obtain the feature vector of the audio data.

[0086] Use the comprehensive feature vector sequence as the input data of a preset large model. Based on the multi-layer neural network of the preset large model, perform non-linear transformation on the input data to map the low-level features to high-level semantic features.

[0087] The preset large model in this application is based on the Transformer architecture. By converting the time data in the multi-modal data into positional encoding, the preset large model can have a clear understanding ability of the time series features of the multi-modal data.

[0088] In a possible implementation, using the comprehensive feature vector sequence as the input data of a preset large model, and based on the multi-layer neural network of the preset large model, performing non-linear transformation on the input data to map the low-level features to high-level semantic features, further includes: when the self-attention mechanism of the preset large model processes each positional encoding, it pays attention to the positional encoding information of the comprehensive feature vector sequence at other positions to capture the long-range dependencies of the multi-modal data, so that during the process of mapping high-level semantic features, it can pay attention to the context information, and thus can accurately judge the association relationships between various elements.

[0089] Perform key element prediction on the high-level semantic features through the output layer of the preset large model, and use the key elements that meet the preset threshold as the extraction results of the key elements in the public interest litigation case.

[0090] In a possible implementation, performing key element prediction on the high-level semantic features through the output layer of the preset large model includes: based on the high-level semantic features obtained by the preset large model, by multiplying the high-level semantic feature vector by the weight matrix of the output layer and adding the bias, obtain the score corresponding to each key element, and map the score through an activation function to obtain the probability distribution corresponding to each key element.

[0091] Convert the key element extraction result into the output data of a preset structure.

[0092] In a possible implementation, the preset structure of the output data is a hierarchical structure, which includes key element categories such as the overall information of public interest litigation cases, the subjects involved, illegal acts, damage consequences, legal bases, etc.

[0093] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a Read-Only Memory (ROM), etc., or a Random Access Memory (RAM), etc.

[0094] It should be understood that although the steps in the flowchart of the accompanying drawings are shown in sequence according to the indication of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0095] Continue to refer to Figure 5 , the key element extraction system for public interest litigation cases based on a large model described in this embodiment includes:

[0096] A multi-modal data acquisition module 601, configured to acquire multi-modal data of public interest litigation cases;

[0097] A time series data acquisition module 602, configured to match the multi-modal data based on a preset time interval to obtain a time matching data set of the time series data set;

[0098] A key element extraction module 603, configured to input the time matching data set into a preset large model, extract the key elements of public interest litigation cases, and generate output data with a preset structure for the key elements.

[0099] To solve the above technical problems, the embodiments of the present application also provide a computer device. For details, please refer to Figure 7 , Figure 7 This is the basic structural block diagram of the computer device in this embodiment.

[0100] The computer device 7 includes a memory 7a, a processor 7b, and a network interface 7c that are communicatively connected to each other via a system bus. It should be noted that only the computer device 7 with components 7a - 7c is shown in the figure. However, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Among them, those skilled in the art of the present technology can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0101] The computer device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The computer device can perform human-computer interaction with the user through means such as a keyboard, a mouse, a remote control, a touchpad, or a voice control device.

[0102] The memory 7a includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disks, multimedia cards, card-type memories (such as SD or DX memories, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memories, magnetic disks, optical discs, etc. In some embodiments, the memory 7a can be an internal storage unit of the computer device 7, such as the hard disk or memory of the computer device 7. In other embodiments, the memory 7a can also be an external storage device of the computer device 7, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc., equipped on the computer device 7. Of course, the memory 7a can also include both the internal storage unit and the external storage device of the computer device 7. In this embodiment, the memory 7a is generally used to store the operating system and various application software installed on the computer device 7, such as the program code of the method for extracting key elements of public interest litigation cases of large models. In addition, the memory 7a can also be used to temporarily store various data that have been output or will be output.

[0103] The processor 7b may be a Central Processing Unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 7b is generally used to control the overall operation of the computer device 7. In this embodiment, the processor 7b is used to run the program code stored in the memory 7a or process data, such as running the program code of the method for extracting key elements of public interest litigation cases of the large model.

[0104] The network interface 7c may include a wireless network interface or a wired network interface, and this network interface 7c is generally used to establish a communication connection between the computer device 7 and other electronic devices.

[0105] The present application also provides another implementation manner, that is, to provide a non-volatile computer-readable storage medium storing a program of the method for extracting key elements of public interest litigation cases of the large model, and the method for extracting key elements of public interest litigation cases of the large model can be executed by at least one processor, so that the at least one processor executes the steps of the method for extracting key elements of public interest litigation cases of the large model as described above.

[0106] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation manner. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc), and includes several instructions to enable a terminal device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in various embodiments of the present application.

[0107] Obviously, the above-described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The preferred embodiments of the present application are shown in the drawings, but do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure content of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions described in the foregoing specific embodiments, or perform equivalent replacements on some of the technical features. Any equivalent structure directly or indirectly using the content of the specification and drawings of the present application in other related technical fields shall be similarly within the scope of the patent protection of the present application.

Claims

1. A method for extracting key elements of public interest litigation cases based on a large model, characterized in that: include: Obtain multimodal data on public interest litigation cases; Matching the multimodal data based on a preset time interval to obtain a time matching data set of the time series data set; The time-matching data set is input into the preset large model to extract the key elements of the public interest litigation case, and the key elements are used to generate output data with a preset structure.

2. According to the method for extracting key elements of public interest litigation cases based on a large model according to claim 1, it is characterized in that: Acquire multimodal data information of public interest litigation cases, wherein the multimodal data information includes text information, image information, video information, and audio information of the public interest litigation cases.

3. The method for extracting key elements of public interest litigation cases based on a large model according to claim 2 is characterized in that: Before matching multimodal data based on a preset time interval, the text information, image information, video information, and audio information need to be preprocessed.

4. The method for extracting key elements of public interest litigation cases based on a large model according to claim 3 is characterized in that: In the process of preprocessing the multimodal data, when the time information of the multimodal data is obtained, the time information in the multimodal data information is marked.

5. The method for extracting key elements of public interest litigation cases based on a large model according to claim 4 is characterized in that: Match the multimodal data based on the preset time interval to obtain the time matching data set of the time series data set, including: Extract time-related features from the preprocessed time series dataset; Convert the time-related features extracted from the time series dataset into a time data representation in a unified format; Based on the time data representation in a unified format, the time series data set will be matched based on the preset time interval to obtain the time matching data set of the time series data set; Construct a time series data structure based on the time order of the time matching dataset.

6. The method for extracting key elements of public interest litigation cases based on a large model according to claim 5 is characterized in that: Before matching the time series data set based on the preset time interval, the time data in a unified format is also represented and sorted according to the time development context to obtain a time series data set of multimodal data.

7. The method for extracting key elements of public interest litigation cases based on a large model according to claim 1 is characterized in that: The time matching data set is input into the preset large model to extract the key elements of the public interest litigation case, and the key elements are used to generate output data with a preset structure. The specific contents include: Obtain the feature vectors of each modality data in the time matching dataset, fuse the feature vectors of each modality data through the attention mechanism, integrate the time information of each modality data into the feature vector based on position encoding, and obtain the comprehensive feature vector sequence of the time matching dataset; The comprehensive feature vector sequence is used as the input data of the preset large model, and the input data is nonlinearly transformed based on the preset large model multi-layer neural network to map the low-level features into high-level semantic features; The key elements of high-level semantic features are predicted through the output layer of the preset large model, and the key elements that meet the preset threshold are used as the key element extraction results of the public interest litigation case; Convert key element extraction results into output data with preset structure.

8. A system for extracting key elements of public interest litigation cases based on a large model, used to implement the method for extracting key elements of public interest litigation cases based on a large model of claims 1-7, characterized in that: include: A multimodal data acquisition module, used to acquire multimodal data of public interest litigation cases; A time series data acquisition module is used to match the multimodal data based on a preset time interval to obtain a time matching data set of the time series data set; The key element extraction module is used to input the time matching data set into the preset large model, extract the key elements of the public interest litigation case, and generate output data of the preset structure from the key elements.

9. An electronic device, characterized in that: include: one or more processors; Memory; And one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, which, when executed by the device, enable the device to perform the steps of the method for extracting key elements of public interest litigation cases based on a large model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed on a computer, enables the computer to execute the steps of the method for extracting key elements of public interest litigation cases based on a large model as described in any one of claims 1 to 7.

Images