Intelligent contract examination and generation system based on AI large model

Abstract

The invention provides an intelligent contract examination and generation system based on an AI large model. The intelligent contract examination and generation system comprises the following steps: S1, data processing; s2, constructing a knowledge base; and S3, model training architecture. According to the intelligent contract examination and generation system based on the AI large model, the seriousness requirement of laws and regulations is met, the accuracy of the AI model in the law field can be greatly improved through close cooperation with a law expert team and fine adjustment of mass data, errors and illusion are reduced, the law compliance of the contract is concerned, and the law compliance of the contract is improved. More attention is paid to specific requirements and sites of users, customer interests are protected, contract terms better conforming to the user interests are generated by a tool by analyzing service backgrounds and requirements of the users, user interface design is simple and visual, the users can easily use the contract terms without professional legal knowledge, the users only need to describe the requirements through natural languages, and user experience is improved. Therefore, the contract can be examined and generated, the use threshold is greatly reduced, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the field of contract review and generation, and in particular to an intelligent contract review and generation system based on an AI big model. Background Art

[0002] Smart contract generation technology has undergone many years of development and is currently generally implemented using the template keyword replacement method. This method requires users to first enter keywords and list specific requirements in detail. Then, the system will match the corresponding template based on these inputs and generate the contract text in an automated manner.

[0003] In the past two years, smart contract review technology has made significant progress, especially thanks to the development of artificial intelligence big models, which has spawned many contract review technologies based on AI big models. These technologies use AI-driven legal big models to conduct contract reviews, aiming to improve the efficiency and accuracy of legal services. By quickly understanding the content of the contract, identifying potential risks, and providing corresponding modification suggestions, these technologies have greatly optimized the contract review process.

[0004] The contract generation technology using the template replacement method has several limitations. First, this method requires users to fill in keywords and core elements of the contract one by one. For those users who are not familiar with the key points of contract filling, it will be difficult to use legal professional terms to accurately describe them, which will cause the generated contract text to have legal compliance risks and will not be able to fully meet the actual needs of customers. Secondly, with the development of society and the continuous changes in laws and regulations, templates cannot cover all types of contracts. Therefore, the contract generation method that relies on template replacement technology has certain limitations and requires a large number of manual review processes to ensure the timely update and accuracy of contract templates.

[0005] The current contract review technology based on artificial intelligence big models has several shortcomings. First, AI big models are prone to the so-called "hallucination" problem, that is, they produce results that are inconsistent with laws and regulations or are wrong during the review process, which will mislead users. Especially in some special or segmented legal fields, AI models will not be able to provide accurate and reliable review opinions. Secondly, the existing AI intelligent review technology mainly focuses on the legal compliance of contracts, but there are deficiencies in the choice of review positions, and it cannot fully meet the user's personalized needs for contract positions. These problems limit the scope of application of AI contract review technology and will affect its effectiveness and practicality in actual legal services.

[0006] Therefore, it is necessary to provide an intelligent contract review and generation system based on AI big model to solve the above technical problems. Summary of the invention

[0007] The present invention provides an intelligent contract review and generation system based on an AI big model, which solves the challenges faced by non-legal professionals in the contract review and generation process.

[0008] In order to solve the above technical problems, the present invention provides an intelligent contract review and generation system based on an AI big model, comprising the following steps:

[0009] S1. Data processing;

[0010] S11, data preprocessing process;

[0011] S111. First, a round of format normalization is carried out to uniformly process data from different sources into plain text format to remove the file format. Among them, scanned content needs to be extracted with the help of OCR technology;

[0012] S112, pre-processing the text content to remove advertisements, irrelevant information, and copyright information;

[0013] S113, performing a round of data cleaning to remove duplicate data through text comparison;

[0014] S114, perform text error correction and manual calibration. For OCR recognition errors, outdated information, and typos in massive texts, collect and organize them through a pre-trained large model and enter the manual calibration process;

[0015] S115. Processing structured information of data, extracting structured information from different types of legal documents or other data;

[0016] S12, generation of vector database;

[0017] S2, knowledge base construction;

[0018] S21, generation of knowledge graph;

[0019] S211. First, relevant teams will extract and construct entity relationships for basic and common legal terms;

[0020] S212, then, the preprocessed data document is combined with an advanced large language model to further construct and extract a deeper lexical graph and definition graph;

[0021] S22, large model fine-tuning training;

[0022] S221. In order to expand the data scale, three methods, namely behavior shaping, knowledge expansion and thinking development, were used to reconstruct the data during the fine-tuning process;

[0023] S23, pre-generation of static content library;

[0024] S231. A series of static content libraries are pre-generated in the background system;

[0025] S232. The pre-generated content will be combined with the fine-tuned large language model and the knowledge base content including the vector database and knowledge graph, and the RAG search enhancement method will be used to pre-generate the review rules, review points, contract templates, and Q&A content of different types of contract documents;

[0026] S233. Pre-generated content needs to be manually calibrated to ensure its accuracy, and regular calibration by legal experts is required to ensure that it does not contain outdated, erroneous, or ambiguous information;

[0027] S3, model training architecture;

[0028] S31. Technical architecture of practical applications;

[0029] S311. When actually conducting a contract review or other process, the structured information of the input content will first be extracted;

[0030] S312. Using vector database technology, key excerpts from the contract are converted into vector form for storage, so as to facilitate efficient similarity retrieval;

[0031] S313. When the user raises a demand for contract review or generation, the system first retrieves relevant contract information through the knowledge graph and vector database;

[0032] S314, the retrieved information is then used to guide the generation process of the large language model;

[0033] S315. Ultimately, the contract text or review results output by the system not only include the user's customized needs, but also comply with the normative nature of the legal text, thus realizing the automation and intelligence of contract review and generation.

[0034] Preferably, the generation of the vector database in S12 comprises the following steps:

[0035] S121, the vector database is mainly used to store structured legal text information, and the user's subsequent intelligent retrieval and RAG modules;

[0036] S122. For the vectorization of legal documents, different processing methods are used for different types of source data.

[0037] Preferably, for the preprocessed data in S121, the system uses a customized AI model to convert the text into vector data and stores it in a vector database. Different processing methods such as a customized word segmentation and segmentation method and an adaptive context selection method are used in S122.

[0038] Preferably, the vector database in S12 is mainly used in Chinese scenarios, and the Chinese word segmentation of the vector database is specially optimized for Chinese scenarios to ensure the rigor of the knowledge base in legal scenarios.

[0039] Preferably, in the field of legal text processing in S21, in addition to converting text knowledge into vectorized form for machine understanding, constructing a knowledge graph of legal terms and lexicon can provide a more accurate and legally logical interpretation of the text.

[0040] Preferably, the process in S212 not only involves the identification of entities, but also includes the extraction of relationships between entities, such as references and dependencies between legal clauses, and logical connections between legal concepts.

[0041] Preferably, the generation of the knowledge graph of S21 adopts an attribute graph model, in which nodes represent legal entities, edges represent relationships between entities, and both nodes and edges can be accompanied by attributes. This model is intuitive and easy to implement, and is suitable for large-scale graph data processing. In addition, the attribute graph model is also an important part of knowledge graph construction. It defines the entity categories, attributes, and relationship types in the knowledge graph, ensuring the accuracy and comprehensiveness of the knowledge graph.

[0042] Preferably, the behavior shaping in S221 refers to rewriting the data in the format of the legal syllogism to complete the reconstruction of the behavior shaping. The knowledge expansion is aimed at case-type data, using the basic large model to expand the legal knowledge in the output to provide more reasoning details. The thinking development refers to the design of the legal thinking chain by referring to the thinking chain method to require the model to reason the answer according to the legal three-stage thinking model.

[0043] Preferably, in the field of contract review and generation in S231, these pre-generated contents will significantly improve the response speed of actual application and the standardization of output.

[0044] Preferably, the information in S311 is extracted, such as contract terms, party information and applicable laws, and this information is structured into nodes and relationships in a knowledge graph. In the information extraction step, various components of the contract, such as clause types, excerpts and party roles, are converted into entities and relationships in the graph for subsequent retrieval and analysis. The retrieval process in S313 may involve queries for specific clause types, searches for contracts involving specific organizations, or matching of contract clauses containing specific text content. The model in S314 generates customized contract text or review opinions based on the retrieved content and the user's query intent. In this process, previously pre-generated static knowledge base information or template files are also used to ensure that the format and structure of the generated content conform to the standards and specifications of legal documents.

[0045] Compared with related technologies, the intelligent contract review and generation system based on AI big model provided by the present invention has the following beneficial effects:

[0046] The present invention provides an intelligent contract review and generation system based on AI big model, which meets the seriousness requirements of laws and regulations: through close cooperation with a team of legal experts and fine-tuning of massive data, it can greatly improve the accuracy of AI models in the legal field and reduce errors and illusions.

[0047] Focus on customer demands and review positions: Not only does the tool focus on the legal compliance of the contract, but it also pays more attention to the specific needs and positions of the user, with the protection of customer interests as the core. By analyzing the user's business background and needs, the tool will generate contract terms that are more in line with the user's interests.

[0048] Convenience of interaction: The user interface is simple and intuitive, and users can use it easily without having legal expertise. Users only need to describe their needs in natural language to review and generate contracts, which greatly reduces the usage threshold and improves the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A schematic diagram of a preferred embodiment of an intelligent contract review and generation system based on an AI big model provided by the present invention;

[0050] Figure 2 A schematic diagram of the framework structure for smart contract review and generation;

[0051] Figure 3 A schematic diagram of the first step of review and generation of smart contracts;

[0052] Figure 4 A schematic diagram of the second step of review and generation of smart contracts;

[0053] Figure 5 A schematic diagram of the third step of review and generation of smart contracts;

[0054] Figure 6 A schematic diagram of the fourth step of review and generation of smart contracts;

[0055] Figure 7 Schematic diagram of the specific implementation operations for smart contract review and generation. DETAILED DESCRIPTION

[0056] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0057] Please refer to Figure 1 , Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 and Figure 7 ,in, Figure 1 A schematic diagram of a preferred embodiment of an intelligent contract review and generation system based on an AI big model provided by the present invention; Figure 2 A schematic diagram of the framework structure for smart contract review and generation; Figure 3 A schematic diagram of the first step of review and generation of smart contracts; Figure 4 A schematic diagram of the second step of review and generation of smart contracts; Figure 5 A schematic diagram of the third step of review and generation of smart contracts; Figure 6 A schematic diagram of the fourth step of review and generation of smart contracts; Figure 7 A schematic diagram of the specific implementation of smart contract review and generation. A smart contract review and generation system based on an AI big model includes the following steps:

[0058] S1. Data processing;

[0059] S11, data preprocessing process;

[0060] S111. First, a round of format normalization is carried out to uniformly process data from different sources into plain text format to remove the file format. Among them, scanned content needs to be extracted with the help of OCR technology;

[0061] S112, pre-processing the text content to remove advertisements, irrelevant information, and copyright information;

[0062] S113, performing a round of data cleaning to remove duplicate data through text comparison;

[0063] S114, perform text error correction and manual calibration. For OCR recognition errors, outdated information, and typos in massive texts, collect and organize them through a pre-trained large model and enter the manual calibration process;

[0064] S115. Processing structured information of data, extracting structured information from different types of legal documents or other data;

[0065] S12, generation of vector database;

[0066] S2, knowledge base construction;

[0067] S21, generation of knowledge graph;

[0068] S211. First, relevant teams will extract and construct entity relationships for basic and common legal terms;

[0069] S212, then, the preprocessed data document is combined with an advanced large language model to further construct and extract a deeper lexical graph and definition graph;

[0070] S22, large model fine-tuning training;

[0071] S221. In order to expand the data scale, three methods, namely behavior shaping, knowledge expansion and thinking development, were used to reconstruct the data during the fine-tuning process;

[0072] S23, pre-generation of static content library;

[0073] S231. A series of static content libraries are pre-generated in the background system;

[0074] S232. The pre-generated content will be combined with the fine-tuned large language model and the knowledge base content including the vector database and knowledge graph, and the RAG search enhancement method will be used to pre-generate the review rules, review points, contract templates, and Q&A content of different types of contract documents;

[0075] S233. Pre-generated content needs to be manually calibrated to ensure its accuracy, and regular calibration by legal experts is required to ensure that it does not contain outdated, erroneous, or ambiguous information;

[0076] S3, model training architecture;

[0077] S31. Technical architecture of practical applications;

[0078] S311. When actually conducting a contract review or other process, the structured information of the input content will first be extracted;

[0079] S312. Using vector database technology, key excerpts from the contract are converted into vector form for storage, so as to facilitate efficient similarity retrieval;

[0080] S313. When the user raises a demand for contract review or generation, the system first retrieves relevant contract information through the knowledge graph and vector database;

[0081] S314, the retrieved information is then used to guide the generation process of the large language model;

[0082] S315. Ultimately, the contract text or review results output by the system not only include the user's customized needs, but also comply with the normative nature of the legal text, thus realizing the automation and intelligence of contract review and generation.

[0083] The generation of the vector database in S12 comprises the following steps:

[0084] S121, the vector database is mainly used to store structured legal text information, and the user's subsequent intelligent retrieval and RAG modules;

[0085] S122. For the vectorization of legal documents, different processing methods are used for different types of source data.

[0086] For the preprocessed data in S121, the system uses a customized AI model to convert the text into vector data and stores it in a vector database. Different processing methods are used in S122, such as a customized word segmentation and segmentation method and an adaptive context selection method.

[0087] The vector database in S12 is mainly used in Chinese scenarios, and the Chinese word segmentation of the vector database is specially optimized for Chinese scenarios to ensure the rigor of the knowledge base in legal scenarios.

[0088] In the field of legal text processing in S21, in addition to converting text knowledge into vectorized form for machine understanding, constructing a knowledge graph of legal terms and lexicon can provide a more accurate and legally logical interpretation of the text.

[0089] The process in S212 not only involves the identification of entities, but also includes the extraction of relationships between entities, such as references and dependencies between legal clauses, and logical connections between legal concepts.

[0090] The generation of the knowledge graph of S21 adopts an attribute graph model, in which nodes represent legal entities, edges represent the relationships between entities, and both nodes and edges can be accompanied by attributes. This model is intuitive and easy to implement, and is suitable for large-scale graph data processing. In addition, the attribute graph model is also an important part of knowledge graph construction. It defines the entity categories, attributes and relationship types in the knowledge graph, ensuring the accuracy and comprehensiveness of the knowledge graph.

[0091] The behavior shaping in S221 refers to rewriting the data in the format of the legal syllogism to complete the reconstruction of the behavior shaping. The knowledge expansion is aimed at case-type data, using the basic large model to expand the legal knowledge in the output to provide more reasoning details. The thinking development refers to the thinking chain method to design the legal thinking chain to require the model to infer the answer according to the legal three-stage thinking model.

[0092] In the field of contract review and generation in S231, these pre-generated contents will significantly improve the response speed of actual applications and the standardization of outputs.

[0093] The information in S311 is extracted, such as contract terms, party information and applicable laws, and this information is structured into nodes and relationships in the knowledge graph. In the information extraction step, the various components of the contract, such as clause types, excerpts and party roles, are converted into entities and relationships in the graph for subsequent retrieval and analysis. The retrieval process in S313 may involve queries for specific clause types, searches for contracts involving specific organizations, or matching contract clauses containing specific text content. The model in S314 generates customized contract text or review opinions based on the retrieved content and the user's query intent. In this process, previously pre-generated static knowledge base information or template files are also used to ensure that the format and structure of the generated content conform to the standards and specifications of legal documents.

[0094] Functionality

[0095] 1. Smart Contract Review:

[0096] When users upload contract documents, AI tools will automatically identify contract elements, analyze contract content, and identify potential legal risks and compliance issues.

[0097] First, we use a lightweight model to quickly identify the basic information of the contract, such as the contract type, contract parties, contract date, etc. The key parts of this information can be used for UI display and can also be confirmed and modified through user interaction to ensure the accuracy of subsequent processes;

[0098] The specific process of contract review can be referred to in the above technical architecture. Through RAG technology, combined with the pre-built knowledge base, the automation and intelligence of contract review can be ensured;

[0099] Finally, the system will provide a detailed review report, including risk points, legal advice, and modification suggestions, to help users understand and optimize contract terms;

[0100] It supports users to modify contract documents online, modify them with one click directly according to the review results, and download the contract documents with annotations after the modification.

[0101] 2. Smart contract generation:

[0102] Users can freely describe their contract requirements in natural language, and AI tools will generate customized contract drafts based on the user's requirements and legal knowledge base;

[0103] Based on the information input by the user, the system will first perform preliminary intent recognition through a lightweight model. According to the user's demands with different amounts of information and purposes, the system will arrange different processes to optimize the generation efficiency and effect as much as possible;

[0104] For insufficient information or the need for standardized information, the contract template combined with the large language model can quickly generate the contract template required by the user;

[0105] When complex intent or demand scenarios are identified, the pre-intention recognition model determines whether it is necessary to call the vector database and knowledge graph to complete the generation process;

[0106] The draft contract generated by the system will comply with laws and regulations and protect the interests of users as much as possible;

[0107] Supports online modification and saving of generated contract documents.

[0108] Please refer to Figure 5 It is learned that the contract review process:

[0109] The user first uploads the contract document, and AI automatically identifies the contract type and key information such as Party A and Party B. The user proceeds to the next step after confirming or supplementing the information;

[0110] Through AI intelligent identification of risk points in contracts, it lists modification suggestions and recommended modification contents, locates the original text, supports users to modify and edit online, and supports one-click modification and replacement;

[0111] The contract document with annotations can be exported after user confirmation.

[0112] Please refer to Figure 6 It is learned that the contract generation process is: the user first selects the type of contract to be generated, and supplements the key information and generation requirements. The contract draft is quickly generated through AI, and can be edited, modified and exported online.

[0113] Please refer to the geometry Figure 7 The specific implementation method is as follows:

[0114] 1. System initialization and model training

[0115] S1. Data collection: Cooperate with a team of legal experts to collect a large amount of legal provisions, precedents, contract templates and other data to build a legal knowledge base. The data sources include a large number of paper book scans, e-books, Internet content, etc.;

[0116] S2. Model pre-training: Use a large amount of processed text data to pre-train the AI ​​model to enable it to have language understanding and generation capabilities;

[0117] S3. Model fine-tuning: Combine the pre-trained AI model with the legal knowledge base data and perform fine-tuning to adapt it to the specific needs of the legal field.

[0118] 2. Construction of Knowledge Base and Internet Retrieval System

[0119] S4. Knowledge graph design: Design the structure of the knowledge graph, including entity types (such as legal provisions, cases), relationship types (such as citations, relevance), etc.

[0120] S5. Vector database construction: vectorize the collected legal data and store it in the form of a vector database for retrieval and use by AI models;

[0121] S6. Internet search module: connects to the search engine API and screens and processes the results, filtering out results that do not meet the requirements or are of poor quality, ensuring the accuracy of cited Internet results in the form of a whitelist, and subsequently used in the search enhancement module of the AI ​​model.

[0122] 3. Smart Contract Review and Generation

[0123] S7. User input processing: Users describe contract requirements in natural language, and the system analyzes user intentions through AI big models and natural language processing technology. The system will focus on identifying users' specific demands and positions;

[0124] S8. Contract review: Intelligently analyze uploaded contract documents, call AI big models, combine knowledge graphs and Internet search technology to identify legal risks and compliance issues in contracts;

[0125] S9, Risk identification and report generation: The system generates a detailed review report, including risk points, legal advice and modification suggestions, and provides a series of support such as original text indexing, online editing, quick modification, document export, etc.

[0126] S10. Contract generation: Based on user needs, combined with the legal knowledge base and AI big model, a contract draft that complies with laws and regulations is automatically generated. The generation algorithm will focus on identifying the user's specific demands and positions.

[0127] 4. User Feedback Loop Iteration

[0128] S11, User feedback loop: Users can provide modification feedback on the generated contract review results and contract draft. The system collects data based on the feedback, iterates and optimizes the algorithm to improve performance and user experience;

[0129] S12. Model and knowledge base updates: Regularly update AI models and knowledge bases to adapt to changes in laws and regulations and new legal practices.

[0130] Compared with related technologies, the intelligent contract review and generation system based on AI big model provided by the present invention has the following beneficial effects:

[0131] The present invention provides an intelligent contract review and generation system based on AI big model, which meets the seriousness requirements of laws and regulations: through close cooperation with a team of legal experts and fine-tuning of massive data, it can greatly improve the accuracy of AI models in the legal field and reduce errors and illusions.

[0132] Focus on customer demands and review positions: Not only does the tool focus on the legal compliance of the contract, but it also pays more attention to the specific needs and positions of the user, with the protection of customer interests as the core. By analyzing the user's business background and needs, the tool will generate contract terms that are more in line with the user's interests.

[0133] Convenience of interaction: The user interface is simple and intuitive, and users can use it easily without having legal expertise. Users only need to describe their needs in natural language to review and generate contracts, which greatly reduces the usage threshold and improves the user experience.

[0134] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An intelligent contract review and generation system based on AI big model, characterized in that: The following steps are involved: S1. Data processing; S11, data preprocessing process; S111. First, a round of format normalization is carried out to uniformly process data from different sources into plain text format to remove the file format. Among them, scanned content needs to be extracted with the help of OCR technology; S112, pre-processing the text content to remove advertisements, irrelevant information, and copyright information; S113, performing a round of data cleaning to remove duplicate data through text comparison; S114, perform text error correction and manual calibration. For OCR recognition errors, outdated information, and typos in massive texts, collect and organize them through a pre-trained large model and enter the manual calibration process; S115. Processing structured information of data, extracting structured information from different types of legal documents or other data; S12, generation of vector database; S2, knowledge base construction; S21, generation of knowledge graph; S211. First, relevant teams will extract and construct entity relationships for basic and common legal terms; S212, then, the preprocessed data document is combined with an advanced large language model to further construct and extract a deeper lexical graph and definition graph; S22, large model fine-tuning training; S221. In order to expand the data scale, three methods, namely behavior shaping, knowledge expansion and thinking development, were used to reconstruct the data during the fine-tuning process; S23, pre-generation of static content library; S231. A series of static content libraries are pre-generated in the background system; S232. The pre-generated content will be combined with the fine-tuned large language model and the knowledge base content including the vector database and knowledge graph, and the RAG search enhancement method will be used to pre-generate the review rules, review points, contract templates, and Q&A content of different types of contract documents; S233. Pre-generated content needs to be manually calibrated to ensure its accuracy, and regular calibration by legal experts is required to ensure that it does not contain outdated, erroneous, or ambiguous information; S3, model training architecture; S31. Technical architecture of practical applications; S311. When actually conducting a contract review or other process, the structured information of the input content will first be extracted; S312. Using vector database technology, key excerpts from the contract are converted into vector form for storage, so as to facilitate efficient similarity retrieval; S313. When the user raises a demand for contract review or generation, the system first retrieves relevant contract information through the knowledge graph and vector database; S314, the retrieved information is then used to guide the generation process of the large language model; S315. Ultimately, the contract text or review results output by the system not only include the user's customized needs, but also comply with the normative nature of the legal text, thus realizing the automation and intelligence of contract review and generation.

2. The intelligent contract review and generation system based on AI big model according to claim 1 is characterized in that: The generation of the vector database in S12 comprises the following steps: S121, the vector database is mainly used to store structured legal text information, and the user's subsequent intelligent retrieval and RAG modules; S122. For the vectorization of legal documents, different processing methods are used for different types of source data.

3. The intelligent contract review and generation system based on AI big model according to claim 2 is characterized in that: For the preprocessed data in S121, the system uses a customized AI model to convert the text into vector data and stores it in a vector database. Different processing methods are used in S122, such as a customized word segmentation and segmentation method and an adaptive context selection method.

4. The intelligent contract review and generation system based on AI big model according to claim 2 is characterized in that: The vector database in S12 is mainly used in Chinese scenarios, and the Chinese word segmentation of the vector database is specially optimized for Chinese scenarios to ensure the rigor of the knowledge base in legal scenarios.

5. The intelligent contract review and generation system based on AI big model according to claim 1 is characterized in that: In the field of legal text processing in S21, in addition to converting text knowledge into vectorized form for machine understanding, constructing a knowledge graph of legal terms and lexicon can provide a more accurate and legally logical interpretation of the text.

6. The intelligent contract review and generation system based on AI big model according to claim 1 is characterized in that: The process in S212 not only involves the identification of entities, but also includes the extraction of relationships between entities, such as references and dependencies between legal clauses, and logical connections between legal concepts.

7. The intelligent contract review and generation system based on AI big model according to claim 1 is characterized in that: The generation of the knowledge graph of S21 adopts an attribute graph model, in which nodes represent legal entities, edges represent the relationships between entities, and both nodes and edges can be accompanied by attributes. This model is intuitive and easy to implement, and is suitable for large-scale graph data processing. In addition, the attribute graph model is also an important part of knowledge graph construction. It defines the entity categories, attributes and relationship types in the knowledge graph, ensuring the accuracy and comprehensiveness of the knowledge graph.

8. The intelligent contract review and generation system based on AI big model according to claim 1 is characterized in that: The behavior shaping in S221 refers to rewriting the data in the format of the legal syllogism to complete the reconstruction of the behavior shaping. The knowledge expansion is aimed at case-type data, using the basic large model to expand the legal knowledge in the output to provide more reasoning details. The thinking development refers to the thinking chain method to design the legal thinking chain to require the model to infer the answer according to the legal three-stage thinking model.

9. The intelligent contract review and generation system based on AI big model according to claim 1 is characterized in that: In the field of contract review and generation in S231, these pre-generated contents will significantly improve the response speed of actual applications and the standardization of outputs.

10. The intelligent contract review and generation system based on AI big model according to claim 1 is characterized in that: The information in S311 is extracted, such as contract terms, party information and applicable laws, and this information is structured into nodes and relationships in the knowledge graph. In the information extraction step, the various components of the contract, such as clause types, excerpts and party roles, are converted into entities and relationships in the graph for subsequent retrieval and analysis. The retrieval process in S313 may involve queries for specific clause types, searches for contracts involving specific organizations, or matching contract clauses containing specific text content. The model in S314 generates customized contract text or review opinions based on the retrieved content and the user's query intent. In this process, previously pre-generated static knowledge base information or template files are also used to ensure that the format and structure of the generated content conform to the standards and specifications of legal documents.

Images