Contract auditing method and device based on large model, equipment and medium

Abstract

The invention discloses a contract auditing method and device based on a large model, equipment and a medium. The contract auditing method comprises the steps of creating a standard auditing model; obtaining a to-be-examined contract, and performing content extraction on the to-be-examined contract to obtain a first extraction result of the to-be-examined contract; based on a preset cue word structure framework and a preset semantic model, the first extraction result is expanded, a second extraction result is obtained, and the preset cue word structure framework comprises industry background introduction, auxiliary information and a filling format; filling a standard review model based on the second extraction result to obtain a to-be-reviewed model; and constructing an examination process, and examining the to-be-examined model according to the examination process to obtain an examination result corresponding to the to-be-examined contract. The invention belongs to the field of contract auditing. The contract fraud risk can be reduced.

Description

Technical Field

[0001] The present invention relates to the field of contract review, and particularly to a contract review method, device, equipment, and medium based on a large model. Background Art

[0002] Contract review is a process in which professionals comprehensively review and evaluate the legality, integrity, accuracy, and enforceability of a contract text, aiming to clarify the rights and obligations of all parties to the contract, prevent potential risks, protect the legitimate rights and interests of the parties, and promote the smooth progress of transactions. It is an important link in maintaining the validity of the contract and the interests of all parties.

[0003] In contract review, some terms are too professional and not conducive to non-professionals' reading and understanding. Moreover, there may be situations where terms are similar but have completely different understandings in different industries. Therefore, in this context, there is a risk of contract fraud. The present invention provides a contract review method based on a large model to reduce the risk of contract fraud. Summary of the Invention

[0004] The present invention provides a contract review method, device, equipment, and medium based on a large model, which solves the technical problem in the prior art that some terms in contract review are too professional and not conducive to review, and are prone to the risk of contract fraud, and achieves the technical effect of reducing the risk of contract fraud.

[0005] In a first aspect, the present invention provides a contract review method based on a large model, and the method includes: Create a standard review model, where the standard review model includes several review modules. Among them, the review module includes fixed contract rules, protocol treaty rules, and amounts; Obtain the contract to be reviewed, and perform content extraction on the contract to be reviewed to obtain a first extraction result of the contract to be reviewed; Based on a preset prompt word structure framework and a preset semantic model, expand the first extraction result to obtain a second extraction result, where the preset prompt word structure framework includes industry background introduction, auxiliary information, and filling format; Fill the standard review model based on the second extraction result to obtain a review model to be reviewed; Construct a review process, and review the review model to be reviewed with the review process to obtain a review result corresponding to the contract to be reviewed.

[0006] Further, expanding the first extraction result based on a preset prompt word structure framework and a preset semantic model to obtain a second extraction result includes: Input the first extraction result into the preset prompt word structure framework to perform a first supplement to the first extraction result; Input the first extraction result after the first supplementation into a preset semantic model to obtain a second extraction result.

[0007] Furthermore, it also includes: Package the keywords, the fixed contract rules corresponding to the keywords, and the directly associated documents corresponding to the fixed contract rules to obtain several data groups; Input the data groups into the neural network to be trained to train the neural network to be trained, where the neural network to be trained is used to predict the directly associated documents corresponding to the fixed contract rules; When the preset training requirements are met, save the latest neural parameters and use the neural network to be trained as the preset semantic model.

[0008] Furthermore, the loss function of the neural network to be trained includes:

[0009] Among them, is the loss function of the neural network to be trained, is the number of data groups, represents is related to or not, is the embedded representation of the fixed contract rule, is the embedded representation of the directly associated document, and the embedded representations are all represented by vectors, is the cosine distance between the embedded representations, is the preset boundary value, and are both preset weights, is the true number of words of the directly associated document in the th data group,

[0010] Furthermore, perform content extraction on the contract to be reviewed to obtain the first extraction result of the contract to be reviewed, including: Convert the contract to be reviewed into a picture; Obtain the starting coordinates and ending coordinates of the paragraphs of the contract to be reviewed; According to the starting coordinates and ending coordinates of the paragraphs of the contract to be reviewed, split the picture to obtain the first extraction result of the content extraction of the contract to be reviewed.

[0011] Furthermore, construct a review process, including: Construct the first review point according to the agreement treaty rules and the amount; Construct the second review point according to the fixed contract rules and the amount; Construct a review process based on the first review point, the time weight of the first review point, the second review point, and the time weight of the second review point.

[0012] Further, fill the standard review model based on the second extraction result to obtain a model to be reviewed, including: Fill the second extraction result according to the review modules in the standard review model to obtain a model to be reviewed.

[0013] In a second aspect, the present invention provides a contract review device based on a large model. The device includes: A standard creation module for creating a standard review model, which includes several review modules. Among them, the review modules include fixed contract rules, protocol treaty rules, and amounts; An acquisition module for acquiring a contract to be reviewed and performing content extraction on the contract to be reviewed to obtain a first extraction result of the contract to be reviewed; An expansion module for expanding the first extraction result based on a preset prompt word structure framework and a preset semantic model to obtain a second extraction result. Among them, the preset prompt word structure framework includes industry background introduction, auxiliary information, and filling format; A filling module for filling the standard review model based on the second extraction result to obtain a model to be reviewed; Construct a review process and review the model to be reviewed according to the review process to obtain a review result corresponding to the contract to be reviewed.

[0014] In a third aspect, the present invention provides an electronic device, including: A processor; A memory for storing instructions executable by the processor; Among them, the processor is configured to execute to implement the contract review method based on a large model provided in the first aspect.

[0015] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium. When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device can execute to implement the contract review method based on a large model provided in the first aspect.

[0016] One or more technical solutions provided in the present invention have at least the following technical effects or advantages: The present invention provides a comprehensive and standardized framework for contract review by creating a standard review model and setting specific review modules; content extraction and information expansion based on preset structures and models can completely and deeply dig out the key information of the contract; filling the expansion results into the standard review model to obtain the to-be-reviewed model, and then combining the constructed review process for review can ensure accurate review of the contract from multiple dimensions and according to a systematic process, effectively improving the efficiency and accuracy of contract review, reducing the potential risks of the contract, and since the content to be reviewed is supplemented, it can help non-professionals better understand the contract and reduce the risk of being defrauded by the contract.

[0017] The present invention can guide the neural network to learn the correct correlation relationship between fixed contract rules and directly related documents by designing different penalty terms for relevant and non-relevant situations respectively. The introduction of the logarithmic error term makes the neural network pay more attention to the prediction accuracy of the number of words in directly related documents. In contract review, the number of words in a document may be related to factors such as the importance and complexity of the contract. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 It is a schematic flowchart of the contract review method based on a large model provided by the present invention; Figure 2 It is a schematic structural diagram of the contract review device based on a large model provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] The embodiments of the present invention solve the technical problems in the prior art that some terms of contract review are too professional, not conducive to review, and prone to the risk of contract fraud by providing a contract review method based on a large model.

[0021] The technical solution of the present invention to solve the above technical problems is generally as follows: Contract review method based on a large model, the method comprising: creating a standard review model, the standard review model including a number of review modules, wherein the review modules include fixed contract rules, agreement treaty rules, and amounts; obtaining a contract to be reviewed, and performing content extraction on the contract to be reviewed to obtain a first extraction result of the contract to be reviewed; based on a preset prompt structure framework and a preset semantic model, expanding the first extraction result to obtain a second extraction result, wherein the preset prompt structure framework includes an industry background introduction, auxiliary information, and a filling format; filling the standard review model based on the second extraction result to obtain a to-be-reviewed model; constructing a review process, and reviewing the to-be-reviewed model according to the review process to obtain a review result corresponding to the contract to be reviewed.

[0022] To better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings of the specification and specific embodiments.

[0023] First, it should be noted that the term "and / or" appearing in this article is merely a description of the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after.

[0024] The large model in the present invention can be an expert model in the field of contract review. The expert model in the field of contract review can be a system that automates or assists the contract review process based on natural language processing (NLP) and machine learning algorithms. The expert model in the field of contract review can help lawyers, corporate legal departments, etc. quickly identify key terms, potential risk points, and non-compliant content in contracts.

[0025] Natural language processing (NLP): One of the core technologies of the contract review expert model. Through NLP, the model can understand and analyze the content of contract texts, and automatically extract key information, such as the names of all parties, dates, amounts, liability and obligation clauses, etc.

[0026] Machine learning algorithms: The expert model will use supervised learning or unsupervised learning methods for training. In supervised learning, the model is trained through a large number of labeled contract data sets to learn how to correctly identify and classify different elements in contracts. In unsupervised learning, the model tries to discover patterns and rules from unlabeled data.

[0027] The present invention provides a Figure 1 contract review method based on a large model as shown, including steps S11 - S15: Step S11: Create a standard review model. The standard review model includes several review modules. Among them, a review module includes fixed contract rules, agreement treaty rules, and amounts.

[0028] The purpose of creating the standard review model is to build a general and standardized framework to review contracts in various industries. The standard review model covers the core elements required for contract review, ensuring the comprehensiveness and accuracy of contract review work.

[0029] It can be understood that the fixed contract rules refer to the established contract rules in a certain industry, and the fixed contract rules are usually fixed and unchanged. It can also be understood that the fixed contract rules in different industries may be different. For example, the fixed contract rules in the construction industry are different from those in the transportation industry.

[0030] The agreement treaty rules refer to the specific terms negotiated and determined by both parties of the contract according to specific business needs. They vary depending on the nature of the contract, the content of the transaction, and the wishes of both parties. For example, in a software development contract, the terms regarding software functions, delivery time, acceptance criteria, etc. can belong to the agreement treaty rules.

[0031] It can be understood that both the fixed contract rules and the agreement treaty rules may be related to amounts.

[0032] Step S12: Obtain the contract to be reviewed, and perform content extraction on the contract to be reviewed to obtain the first extraction result of the contract to be reviewed.

[0033] Specifically, it includes: converting the contract to be reviewed into a picture; obtaining the starting coordinates and ending coordinates of the paragraphs of the contract to be reviewed; and splitting the picture according to the starting coordinates and ending coordinates of the paragraphs of the contract to be reviewed to obtain the first extraction result of the content extraction of the contract to be reviewed.

[0034] Specifically speaking: The contract document to be reviewed (which may be in common document formats such as Word, PDF, etc.) can be converted into a picture format.

[0035] It can be understood that contract documents in different formats may have differences in structure and layout. Converting them into a unified picture format can eliminate these differences and facilitate subsequent processing using a unified method. Moreover, the picture format can retain the original layout information of the contract, including the position of the text, the division of paragraphs, etc., providing a more accurate basis for subsequent content extraction.

[0036] In the converted image, determine the starting and ending positions of each paragraph, represented by coordinates. A contract usually consists of multiple paragraphs, and each paragraph may contain different topics or information. By obtaining the starting and ending coordinates of the paragraphs, the scope of each paragraph can be clarified, providing a basis for subsequent image splitting and content extraction. Ensure that the extracted content is divided by paragraphs, which is conducive to maintaining the integrity and logic of the information. (In a contract, usually one paragraph represents one contract rule). According to the paragraph coordinates obtained previously, split the image into multiple parts, with each part corresponding to a paragraph in the contract. After splitting the image by paragraphs, each paragraph can be processed individually, facilitating the subsequent use of OCR (Optical Character Recognition) technology to convert the text in the image into editable text.

[0037] Perform OCR recognition on the split image to convert the text in the image into text, obtaining the first extraction result.

[0038] Step S13: Based on a preset prompt structure framework and a preset semantic model, expand the first extraction result to obtain a second extraction result. The preset prompt structure framework includes an industry background introduction, auxiliary information, and a filling format.

[0039] Specifically, it includes: inputting the first extraction result into the preset prompt structure framework to make the first supplementary to the first extraction result; inputting the first extraction result after the first supplement into the preset semantic model to obtain the second extraction result.

[0040] The industry background introduction refers to information related to the industry to which the contract to be reviewed belongs, such as industry rules, practices, market conditions, etc. The background information helps to understand the rationality and compliance of the contract terms in the industry environment.

[0041] The auxiliary information includes other useful information other than the main body of the contract, such as the handling of previous similar contracts signed for the contract, etc. The auxiliary information can provide a wider reference for contract review.

[0042] The filling format stipulates the format and requirements for information supplementation, ensuring that the supplemented information is consistent and standardized, facilitating subsequent processing.

[0043] As mentioned in the preface, the fixed contract rules in different industries may be different, which means that in different industries, even if the fixed contract rules are relatively close and the keywords are the same, there may be different interpretations. Therefore, supplementation can be carried out through the industry background introduction and auxiliary information.

[0044] The preset semantic model can understand and process natural language text. Its function is to mine the legal provisions and sources directly related to the fixed contract rules based on the input text content, so as to further expand and improve the text.

[0045] Specifically: the first extraction result can be filled in according to the requirements of the preset prompt word structure framework, industry background introduction, auxiliary information, etc. can be added to the first extraction result, and the first extraction result after the first supplement can be input into the preset semantic model to let the model analyze and expand it.

[0046] The training process of the preset semantic model includes: Packing the keywords, the fixed contract rules corresponding to the keywords, and the directly related documents corresponding to the fixed contract rules to obtain several data groups; Inputting the data set into the neural network to be trained to train the neural network, wherein the neural network to be trained is used to predict directly related documents corresponding to the fixed contract rules; When the preset training requirements are met, the latest neural parameters are saved and the neural network to be trained is used as the preset semantic model.

[0047] It is understandable that the preset training requirement may be the number of training times, the threshold of the loss function, etc., which are not limited here.

[0048] In addition, the present invention also provides a method for packaging keywords, fixed contract rules corresponding to the keywords, and directly related documents corresponding to the fixed contract rules to obtain a plurality of data groups, including: Build structured data groups and define data units, and abstract the knowledge in each business scenario into triple data groups: Keywords: core business concepts (such as "liability for breach of contract" and "force majeure"); fixed contract rules: contract terms directly related to the keywords (such as "If Party B breaches the contract, it must pay XX% liquidated damages"); directly related documents: legal provisions, industry standards or historical cases supporting the rules (such as Article XX of the Civil Code, the case number of a certain court).

[0049] Cleaning and structuring: remove duplicate data groups and unify text formats (such as punctuation and terminology consistency); separate long text rules and documents into sentences to facilitate subsequent vectorized decomposition.

[0050] Recursive vectorization: Extract semantic features hierarchically, including: Basic vectorization (single data set encoding) of keywords, fixed contract rules, and directly related documents; use pre-trained language models (such as BERT, Sentence-BERT) to convert each triple of text (keywords + rules + documents) into a fixed-dimensional semantic vector.

[0051] Recursive hierarchical vectorization (mining hierarchical associations), including: Group by keywords, calculate the mean / cluster center of the vectors within the group as the "parent vector" of the keyword; Perform cross-group associations on keywords, and construct a multi-granularity semantic vector pyramid by recursively merging vectors; Hierarchical clustering (constructing a knowledge network structure), including: Calculate semantic similarity (using cosine similarity to measure the semantic distance between vectors (such as the similarity between rule A and rule B).

[0052] Set a threshold to cluster data groups with high similarity into one class; Agglomerative hierarchical clustering: Start from a single data group, merge similar clusters layer by layer to form a tree-like clustering structure. Each leaf node is a specific triple data group; the intermediate nodes are clusters of similar rules (such as "breach of contract category", "termination of contract category"); the root node is the top-level abstract concept (such as "core contract terms").

[0053] Each cluster represents a semantic module (i.e., a data group), which can quickly retrieve similar rules (such as finding all rules and related literature on "liquidated damages calculation methods"). Identify potential associations through vector similarity (such as the semantic intersection between the "intellectual property" cluster and the "confidentiality clause" cluster).

[0054] The loss function of the neural network to be trained, including:

[0055] Among them, is the loss function of the neural network to be trained, is the number of data groups, indicates is related to or not, is the embedded representation of the fixed contract rule, is the embedded representation of the directly related literature, and the embedded representations are all represented by vectors, is the cosine distance between the embedded representations, is the preset boundary value, and are both preset weights, is the true word count of the directly related literature in the is the predicted word count of the directly related literature in the

[0056] The design purpose of the loss function provided by the present invention is to accurately predict the neural network of keywords and directly related documents corresponding to fixed contract rules. The optimization aspects are to judge the relevance between fixed contract rules and directly related documents and improve the prediction accuracy of the number of words in directly related documents.

[0057] By separately designing different penalty terms for relevant and non-relevant situations, the present invention can guide the neural network to learn the correct correlation relationship between fixed contract rules and directly related documents. The introduction of the logarithmic error term makes the neural network pay more attention to the prediction accuracy of the number of words in directly related documents. In contract review, the word count information of documents may be related to factors such as the importance and complexity of the contract.

[0058] Step S14: Fill the standard review model based on the second extraction result to obtain the model to be reviewed.

[0059] Specifically, it includes: filling the second extraction result according to the review modules in the standard review model to obtain the model to be reviewed.

[0060] Step S15: Construct a review process and review the model to be reviewed with the review process to obtain the review result corresponding to the contract to be reviewed.

[0061] Constructing the review process includes: constructing the first review point according to the agreement treaty rules and the amount; constructing the second review point according to the fixed contract rules and the amount; constructing the review process with the first review point, the time weight of the first review point, the second review point, and the time weight of the second review point.

[0062] The time weight refers to the time allocation for reviewing the contract to be reviewed, and the time weights of each review point can be determined according to the work experience of relevant personnel and the characteristics of each industry.

[0063] The first review point covers all agreement treaty rules and the amounts corresponding to the agreement treaty rules.

[0064] The second review point covers all fixed contract rules and the amounts corresponding to the fixed contract rules, and also covers the supplementary results to help non-relevant professionals understand.

[0065] In summary, the present invention provides a contract review method based on a large model. The method includes: creating a standard review model, which includes several review modules. Among them, the review modules include fixed contract rules, protocol treaty rules, and amounts; obtaining the contract to be reviewed, and performing content extraction on the contract to be reviewed to obtain the first extraction result of the contract to be reviewed; based on a preset prompt word structure framework and a preset semantic model, expanding the first extraction result to obtain a second extraction result, where the preset prompt word structure framework includes industry background introduction, auxiliary information, and filling formats; filling the standard review model based on the second extraction result to obtain the model to be reviewed; constructing a review process, and reviewing the model to be reviewed according to the review process to obtain the review result corresponding to the contract to be reviewed. The present invention provides a comprehensive and standardized framework for contract review by creating a standard review model and setting specific review modules; content extraction and information expansion based on a preset structure and model can completely and deeply mine the key information of the contract; filling the expansion result into the standard review model to obtain the model to be reviewed, and then combining the constructed review process for review can ensure accurate review of the contract from multiple dimensions and in a systematic process, effectively improving the efficiency and accuracy of contract review, reducing potential contract risks, and since the content to be reviewed is supplemented, it can help non-professionals better understand the contract and reduce the risk of being defrauded by the contract. The present invention can guide the neural network to learn the correct correlation relationship between the fixed contract rules and the directly related documents by designing different penalty terms for relevant and non-relevant situations respectively. The introduction of the logarithmic error term makes the neural network pay more attention to the prediction accuracy of the number of words in the directly related documents. In contract review, the number of words in the documents may be related to factors such as the importance and complexity of the contract.

[0066] Based on the same inventive concept, the present invention provides a contract review device based on a large model as shown in Figure 2 Figure, and the device includes: A standard creation module 21, configured to create a standard review model, where the standard review model includes several review modules. Among them, the review modules include fixed contract rules, protocol treaty rules, and amounts; An acquisition module 22, configured to obtain the contract to be reviewed, and perform content extraction on the contract to be reviewed to obtain the first extraction result of the contract to be reviewed; An expansion module 23, configured to expand the first extraction result based on a preset prompt word structure framework and a preset semantic model to obtain a second extraction result, where the preset prompt word structure framework includes industry background introduction, auxiliary information, and filling formats; A filling module 24, configured to fill the standard review model based on the second extraction result to obtain the model to be reviewed; Construct a review process 25, and review the review model according to the review process to obtain the review result corresponding to the contract to be reviewed.

[0067] Based on the same inventive concept, the present invention also provides an electronic device, including: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to execute to implement the contract review method based on the large model provided as described above.

[0068] Based on the same inventive concept, the present invention also provides a non-transitory computer-readable storage medium. When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device can execute to implement the contract review method based on the large model provided as described above.

[0069] Since the electronic device introduced in this embodiment is the electronic device used to implement the information processing method in the embodiment of the present invention, based on the information processing method introduced in the embodiment of the present invention, those skilled in the art can understand the specific implementation manners and various variations of the electronic device in this embodiment. Therefore, the specific implementation of how this electronic device implements the method in the embodiment of the present invention will not be described in detail here. As long as the electronic device used by those skilled in the art to implement the information processing method in the embodiment of the present invention belongs to the scope of protection of the present invention.

[0070] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0071] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0072] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means that implement the function specified in the flowchart(s) Figure 1 a flowchart or flowcharts and / or block(s) Figure 1 a block or blocks.

[0073] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the function specified in the flowchart(s) Figure 1 a flowchart or flowcharts and / or block(s) Figure 1 a block or blocks.

[0074] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic inventive concept. Therefore, the appended claims are intended to be construed to cover the preferred embodiments as well as all changes and modifications that fall within the scope of the present invention.

[0075] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A contract review method based on a large model, characterized in that The method includes: Create a standard review model, which includes several review modules. Among them, the review module includes fixed contract rules, protocol treaty rules, and amounts; Obtain the contract to be reviewed, and perform content extraction on the contract to be reviewed to obtain the first extraction result of the contract to be reviewed; Based on a preset prompt word structure framework and a preset semantic model, expand the first extraction result to obtain a second extraction result. Among them, the preset prompt word structure framework includes industry background introduction, auxiliary information, and filling format; Fill the standard review model based on the second extraction result to obtain a review model to be reviewed; Construct a review process, and review the review model to be reviewed with the review process to obtain the review result corresponding to the contract to be reviewed.

2. The contract review method based on a large model according to claim 1, wherein, Based on a preset prompt word structure framework and a preset semantic model, expand the first extraction result to obtain a second extraction result, including: Input the first extraction result into the preset prompt word structure framework to perform the first supplement to the first extraction result; Input the first extraction result after the first supplement into the preset semantic model to obtain the second extraction result.

3. The contract review method based on a large model according to claim 2, wherein It also includes: Package keywords, the fixed contract rules corresponding to the keywords, and the directly associated documents corresponding to the fixed contract rules to obtain several data groups; Input the data groups into the neural network to be trained to train the neural network to be trained. Among them, the neural network to be trained is used to predict the directly associated documents corresponding to the fixed contract rules; When the preset training requirements are met, save the latest neural parameters, and use the neural network to be trained as the preset semantic model.

4. The contract review method based on a large model according to claim 3, wherein The loss function of the neural network to be trained includes: in, is the loss function of the neural network to be trained, is the number of data sets, express and Is it relevant? is the embedded representation of the fixed contract rules, is the embedding representation of the directly related documents, and the embedding representation is expressed as a vector. is the cosine distance between the embedded representations, is the preset boundary value, and All are preset weights. For the The actual number of words in the directly related documents in the array, For the The predicted number of words in the directly associated documents in the array.

5. The contract review method based on a large model according to claim 1, characterized in that Perform content extraction on the contract to be reviewed to obtain the first extraction result of the contract to be reviewed, including: Convert the contract to be reviewed into a picture; Obtain the starting coordinates and ending coordinates of the paragraphs of the contract to be reviewed; According to the starting coordinates and ending coordinates of the paragraphs of the contract to be reviewed, split the picture to obtain the first extraction result of the content extraction of the contract to be reviewed.

6. The contract review method based on the large model according to claim 1, wherein Construct a review process, including: Construct the first review point according to the protocol treaty rules and amounts; Construct the second review point according to the fixed contract rules and amounts; Construct a review process with the first review point, the time weight of the first review point, the second review point, and the time weight of the second review point.

7. The contract review method based on a large model according to claim 2, characterized in that Fill the standard review model based on the second extraction result to obtain a review model to be reviewed, including: Fill the second extraction result according to the review modules in the standard review model to obtain a review model to be reviewed.

8. The contract review device based on the large model is characterized in that The device includes: A standard creation module for creating a standard review model, which includes several review modules. Among them, the review module includes fixed contract rules, protocol treaty rules, and amounts; An acquisition module for acquiring the contract to be reviewed and performing content extraction on the contract to be reviewed to obtain the first extraction result of the contract to be reviewed; An expansion module, configured to expand the first extraction result based on a preset prompt structure framework and a preset semantic model to obtain a second extraction result, where the preset prompt structure framework includes an industry background introduction, auxiliary information, and a filling format; A filling module, configured to fill the standard review model based on the second extraction result to obtain a model to be reviewed; Construct a review process, and review the model to be reviewed according to the review process to obtain a review result corresponding to the contract to be reviewed.

9. An electronic device, characterized in that, Comprising: A processor; A memory for storing executable instructions of the processor; Wherein, the processor is configured to execute to implement the contract review method based on a large model according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute and implement the contract review method based on a large model according to any one of claims 1 to 7.

Images