Contract risk management method and system based on large model

Abstract

The invention discloses a contract risk management method and system based on a large model, belongs to the technical field of artificial intelligence, and aims to solve the technical problem of how to quickly and accurately analyze an original text of a contract and identify a contract risk. Comprising the following steps: performing document analysis on a contract document uploaded by business personnel; according to roles of clients in a contract, risk analysis is carried out on contract document information through a large model, contract risk data related to the client roles are identified, the large model is called to carry out contract risk suggestions, risk suggestions of each client role are obtained, and the risk suggestions of each client role are judged and audited; constructing a message body based on the contract risk data and the approved risk suggestion, and storing the message body in an object storage; and performing model fine adjustment on the large model regularly according to the message body in the object storage.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and more specifically, to a contract risk management method and system based on a large model. Background Art

[0002] With the continuous increase in the number and types of enterprise contracts, the original method of manually reading contract texts requires relevant personnel to have a profound background in finance and law and understand the relevant characteristics of contract risks in order to extract relevant risks. Contracts contain complex legal clauses, and manual contract review is time-consuming and laborious, and prone to omissions, affecting the decision-making efficiency of enterprises. Once an omission occurs and the corresponding risk points are not identified, it can cause significant losses to the enterprise. Therefore, it has become increasingly important to develop an intelligent contract risk management method.

[0003] How to quickly and accurately parse the original contract text and identify contract risks is a technical problem that needs to be solved. Summary of the Invention

[0004] The technical task of the present invention is to address the above deficiencies and provide a contract risk management method and system based on a large model to solve the technical problem of how to quickly and accurately parse the original contract text and identify contract risks.

[0005] In a first aspect, a contract risk management method based on a large model according to the present invention includes the following steps:

[0006] Contract parsing: Parse the contract document uploaded by the business personnel to obtain contract document information;

[0007] Risk identification and analysis: Based on the role of the customer in the contract, perform risk analysis on the contract document information through a large model to identify contract risk data related to the customer role. For each customer role, construct a Prompt based on the corresponding contract risk data and call the large model for contract risk suggestions, obtain risk suggestions for each customer role, and judge and review the risk suggestions for each customer role. Based on the contract risk data and the approved risk suggestions, construct a message body and save the message body to an object storage;

[0008] Model iteration: Regularly fine-tune the large model according to the message body in the object storage.

[0009] Preferably, during contract parsing, parse the contract document uploaded by the business personnel, identify the key elements of the contract, and obtain contract document information including contract number, signing date, both parties to the contract, contract amount, service or product description, payment deadline, and liability for breach of contract.

[0010] Preferably, the risk identification and analysis includes the following steps:

[0011] Analyze the contract document information, including the payment terms, liability limitations, intellectual property ownership, keywords and context in the contract document information of the contract, to determine the role of the customer in the contract;

[0012] According to the role of the customer in the contract, identify the risk terms for different customer roles through the Bert model to obtain contract risk data. The contract risk data includes liability defaults, dispute resolution mechanisms, contract changes and termination conditions. Among them, the customer roles include Party A and Party B. Party A is the principal or purchaser in the contract, and Party B is the executor or service provider in the contract;

[0013] For various customer roles, construct corresponding Prompts based on the contract risk data of the customer role. Using the Prompt as the input, call the large model to ask questions about the contract risk category, contract risk description, and contract risk suggestions, and obtain the contract risk category, contract risk description, and contract risk suggestions returned by the large model. Take the contract risk data, contract risk category, contract risk description, and contract risk suggestions as contract risk information;

[0014] Judge the correctness of the contract risk information returned by the large model. If it passes the judgment, assemble it into a message body including contract risk data, contract risk category, contract risk description, and contract risk suggestions according to the format requirements. If it fails to pass the judgment, annotate the contract risk information, and have business experts review the annotated contract risk information. After passing the review, assemble it into a message body including contract risk data, contract risk category, contract risk description, and contract risk suggestions according to the format requirements;

[0015] Save the message body to the object storage.

[0016] Preferably, the model iteration includes the following steps:

[0017] Authorized managers pull the contract risk information from the storage object;

[0018] Divide the pulled contract risk information according to the requirements to construct a training set, a validation set, and a test set;

[0019] Set the training strategy, including the learning rate and batch size;

[0020] Set the training set to iteratively update the model parameters and perform model fine-tuning training operations based on the training set and training strategy;

[0021] Based on the validation set and the test set, evaluate the fine-tuned large model. If the functions of the fine-tuned large model meet the expectations, deploy the fine-tuned large model to the production environment. If the functions of the fine-tuned large model do not meet the expectations, perform model fine-tuning training operations again based on the training set and the training strategy until the expected model fine-tuning goal is met.

[0022] In a second aspect, a contract risk management system based on a large model of the present invention is used to perform contract risk management through a contract risk management method based on a large model as described in any item of the first aspect. The system includes a contract parsing module, a risk identification and analysis module, and a model iteration module;

[0023] The contract parsing module is used to perform the following: parse the contract document uploaded by the business personnel to obtain contract document information;

[0024] The risk identification and analysis module is used to perform the following: according to the role of the customer in the contract, perform risk analysis on the contract document information through the large model, identify contract risk data related to the customer role. For each customer role, construct a Prompt based on the corresponding contract risk data and call the large model for contract risk suggestions, obtain risk suggestions for each customer role, and judge and review the risk suggestions for each customer role. Based on the contract risk data and the risk suggestions that pass the review, construct a message body and save the message body to the object storage;

[0025] The model iteration module is used to perform the following: regularly fine-tune the large model according to the message body in the object storage.

[0026] Preferably, the contract parsing module is used to parse the contract document uploaded by the business personnel, identify the key elements of the contract, and obtain contract document information including the contract number, signing date, both parties to the contract, contract amount, service or product description, payment period, and liability for violation.

[0027] Preferably, the risk identification and analysis module is used to perform the following operations:

[0028] Analyze the contract document information, analyze the payment conditions, liability limitations, intellectual property rights ownership, keywords and context in the contract document information to determine the role of the customer in the contract;

[0029] According to the role of the customer in the contract, identify the risk terms of different customer roles through the Bert model to obtain contract risk data. The contract risk data includes liability for breach, dispute resolution mechanism, contract change and termination conditions. Among them, the customer roles include Party A and Party B. Party A is the principal or purchaser in the contract, and Party B is the executor or service provider in the contract;

[0030] For various customer roles, construct corresponding Prompts based on the contract risk data of the customer roles. Using the Prompts as input, call the large model to ask questions about the contract risk category, contract risk description, and contract risk suggestions, and obtain the contract risk category, contract risk description, and contract risk suggestions returned by the large model. Take the contract risk data, contract risk category, contract risk description, and contract risk suggestions as contract risk information;

[0031] Judge the correctness of the contract risk information returned by the large model. If it passes the judgment, assemble it into a message body including contract risk data, contract risk category, contract risk description, and contract risk suggestions according to the format requirements. If it fails to pass the judgment, annotate the contract risk information, and have the business experts review the annotated contract risk information. After passing the review, assemble it into a message body including contract risk data, contract risk category, contract risk description, and contract risk suggestions according to the format requirements;

[0032] Save the message body to the object storage.

[0033] Preferably, the model iteration module is used to perform the following operations:

[0034] Authorized managers pull contract risk information from the storage object;

[0035] Divide the pulled contract risk information according to requirements to construct a training set, a validation set, and a test set;

[0036] Set training strategies, including learning rate and batch size;

[0037] Set the training set to iteratively update the model parameters, and perform model fine-tuning training operations based on the training set and training strategies;

[0038] Evaluate the fine-tuned large model based on the validation set and the test set. If the functions of the fine-tuned large model meet the expectations, deploy the fine-tuned large model to the production environment. If the functions of the fine-tuned large model do not meet the expectations, perform model fine-tuning training operations again based on the training set and training strategies until the expected model fine-tuning goal is met.

[0039] The contract risk management method and system based on the large model of the present invention have the following advantages:

[0040] 1. Improve the efficiency of contract risk data identification: The traditional process of identifying and discovering contract risk data relies on the professional knowledge and experience of business personnel, which is time-consuming and laborious. This method can automatically identify potential contract risk clause information from the contract text, greatly accelerating the speed of contract risk identification;

[0041] 2. Provide prediction accuracy: Identify and warn of potential risks in advance, take targeted measures to reduce the probability of contract risks occurring;

[0042] 3. Reduce the cost of contract disputes: Discover potential problems in the contract in advance, avoid the occurrence of contract disputes, and reduce the cost of dispute resolution;

[0043] 4. Enhance the enterprise's compliance ability: Continuously enhance the enterprise's compliance ability through model training and iteration, and reduce the risk of violations;

[0044] 5. Improve contract execution: Optimize the contract text to make it more compliant with laws, regulations, and market rules, and improve contract execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] 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 use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0046] The present invention will be further described below in conjunction with the drawings.

[0047] Figure 1 FIG. is a flowchart of a contract risk management method based on a large model for Embodiment 1;

[0048] Figure 2 FIG. is a flowchart of risk identification and analysis in a contract risk management method based on a large model for Embodiment 1;

[0049] Figure 3 FIG. is a flowchart of model iteration in a contract risk management method based on a large model for Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] The present invention will be further described below in conjunction with the drawings and specific embodiments, so that those skilled in the art can better understand the present invention and be able to implement it. However, the specific embodiments cited are not intended to limit the present invention. Without conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0051] The embodiments of the present invention provide a contract risk management method and system based on a large model, which are used to solve the technical problem of how to quickly and accurately parse the original contract text and identify contract risks.

[0052] Embodiment 1:

[0053] A contract risk management method based on a large model of the present invention includes three steps: contract parsing, risk identification and analysis, and model iteration.

[0054] Step S100 Contract Parsing: Parse the contract document uploaded by the business personnel to obtain contract document information.

[0055] As a specific implementation of contract parsing, parse the contract document uploaded by the business personnel, identify the key elements of the contract, and obtain contract document information including contract number, signing date, both parties to the contract, contract amount, service or product description, payment period, and liability for violation.

[0056] Step S200 Risk Identification and Analysis: Based on the role of the customer in the contract, conduct risk analysis on the contract document information through a large model, identify contract risk data related to the customer role. For each customer role, construct a Prompt according to the corresponding contract risk data, call the large model for contract risk suggestions, obtain risk suggestions for each customer role, judge and review the risk suggestions for each customer role, construct a message body based on the contract risk data and the approved risk suggestions, and save the message body to the object storage.

[0057] As a specific implementation of risk identification and analysis, this step includes the following operations:

[0058] (1) Analyze the contract document information, analyze the payment terms, liability limitations, intellectual property ownership, keywords and context in the contract document information to determine the role of the customer in the contract;

[0059] (2) According to the role of the customer in the contract, identify risk terms for different customer roles through the Bert model to obtain contract risk data. The contract risk data includes liability for breach, dispute resolution mechanism, contract change and termination conditions. Among them, the customer roles include Party A and Party B. Party A is the principal or purchaser in the contract, and Party B is the executor or service provider in the contract;

[0060] (3) For various customer roles, construct corresponding Prompts according to the contract risk data of the customer role. Use the Prompt as the input, call the large model to ask questions about contract risk categories, contract risk descriptions, and contract risk suggestions, obtain the contract risk categories, contract risk descriptions, and contract risk suggestions returned by the large model, and regard the contract risk data, contract risk categories, contract risk descriptions, and contract risk suggestions as contract risk information;

[0061] (4) Judge the correctness of the contract risk information returned by the large model. If the judgment is passed, assemble it into a message body including contract risk data, contract risk category, contract risk description, and contract risk suggestions according to the format requirements. If the judgment fails, annotate the contract risk information and have the business experts review the annotated contract risk information. After the review is passed, assemble it into a message body including contract risk data, contract risk category, contract risk description, and contract risk suggestions according to the format requirements;

[0062] (5) Save the message body to the object storage.

[0063] Step S300 Model Iteration: Regularly fine-tune the large model based on the message body in the object storage.

[0064] As a specific implementation of model iteration, this step includes the following operations:

[0065] (1) Authorized managers pull the contract risk information from the storage object;

[0066] (2) Divide the pulled contract risk information according to requirements to construct a training set, a validation set, and a test set;

[0067] (3) Set the training strategy, including the learning rate and batch size;

[0068] (4) Set the training set to iteratively update the model parameters and perform model fine-tuning training operations based on the training set and training strategy;

[0069] (5) Evaluate the fine-tuned large model based on the validation set and the test set. If the functions of the fine-tuned large model meet the expectations, deploy the fine-tuned large model to the production environment. If the functions of the fine-tuned large model do not meet the expectations, perform model fine-tuning training operations again based on the training set and training strategy until the expected model fine-tuning goal is met.

[0070] Based on the method disclosed in this embodiment, a detailed operation process for contract risk identification and management is given:

[0071] (1) Deploy this method to the system and deploy a large model in the system.

[0072] (2) Business personnel upload the contract original text and perform document parsing operations on the contract original text;

[0073] If the document parsing fails, the process ends;

[0074] If the document parsing is successful, continue to the next step.

[0075] (3) The large model automatically identifies and extracts the implicit contract risk clause information from the parsed text fragments, and displays the progress information of the contract risk clauses through the front end, with the status being "analyzing".

[0076] (4) After the system analysis is completed, the page progress status is displayed as "analysis successful".

[0077] (5) The business personnel check the contract risk clause information to confirm the accuracy of the data.

[0078] If the contract risk analysis and extraction are correct, the risk data information is saved to the object storage.

[0079] If the contract risk identification is incorrect, after making annotations, it is submitted to the business expert for review and an email notification is sent.

[0080] (6) After receiving the notification email, the business expert logs in to the system to perform the contract risk identification review operation.

[0081] If the review is passed, the contract risk information is saved to the object storage.

[0082] If the review is not passed, the corresponding opinions are input and the process returns to step 5.

[0083] Among them, the contract risk identification and analysis process is as follows: The role identification in the contract is the key to risk management. Different roles bear different responsibilities and risks in the contract, so it is necessary to identify the risk clauses separately according to the roles. It is necessary to conduct contract risk identification and analysis in different contracts according to the customer's role, so that the analysis results are more in line with the actual requirements of the enterprise. The following is the corresponding process information.

[0084] (1) The system obtains the parsed contract document information and identifies the key elements of the contract, such as contract number, signing date, both parties to the contract, contract amount, service or product description, delivery period, liability for breach of contract, etc.

[0085] (2) The system analyzes the contract terms in depth, such as payment terms, liability limitations, intellectual property ownership, analyzes the keywords and context in the contract text to determine the role the customer plays in the contract.

[0086] (3) According to the role of the customer in the contract, the system will use the Bert small model to identify the risk clauses for the risk of Party A and Party B respectively. These risk clauses may include but are not limited to liability for breach of contract, dispute resolution mechanism, contract change and termination conditions, etc. Among them, Party A usually refers to the principal or purchaser in the contract, while Party B is the executor or service provider.

[0087] (4) After the system identifies risk terms related to the customer role, it needs to assemble corresponding Prompts based on these terms. If the customer role in the contract is Party A, the system assembles the Prompt according to Party A and calls the large model to ask about the contract risk category: For the contract clause "XXX", give the risk category for Party A, maximizing the interests of Party A. Strictly use the format "Risk category:..." as the answer format. The given risk category only contains one category, and the number of characters in the risk category does not exceed 10 Chinese characters. The risk type should be as concise as possible and not contain punctuation marks.

[0088] (5) If the customer role in the contract is Party A, the system assembles the Prompt according to Party A and calls the large model to ask about the contract risk explanation: For the contract clause "XXX", give the risk explanation for Party A, maximizing the interests of Party A. Strictly use the format "Risk explanation:..." as the answer format, and the number of characters in the risk explanation does not exceed 40 Chinese characters.

[0089] (6) If the customer role in the contract is Party A, the system assembles the Prompt according to Party A and calls the large model to ask about the contract risk suggestion: For the contract clause "XXX", give the modification suggestion for Party A, maximizing the interests of Party A. Avoid content such as "extending Party A's payment period", "reducing Party B's liquidated damages ratio", "increasing Party A's liquidated damages ratio", and "adding restrictive clauses for Party A" in the suggestion. Strictly use the format "Modification suggestion:..." as the answer format, and the number of characters in the modification suggestion does not exceed 40 Chinese characters.

[0090] (7) If the customer role in the contract is Party B, the system assembles the Prompt according to Party B and calls the large model to ask about the contract risk category: For the contract clause "XXX", give the risk category for Party B, maximizing the interests of Party B. Strictly use the format "Risk category:..." as the answer format. The given risk category only contains one category, and the number of characters in the risk category does not exceed 10 Chinese characters. The risk type should be as concise as possible and not contain punctuation marks.

[0091] (8) If the customer role in the contract is Party B, the system assembles the Prompt according to Party B and calls the large model to ask about the contract risk explanation: For the contract clause "XXX", give the risk explanation for Party B, maximizing the interests of Party B. Strictly use the format "Risk explanation:..." as the answer format, and the number of characters in the risk explanation does not exceed 40 Chinese characters.

[0092] (9) If the customer role in the contract is Party B, the system assembles a prompt according to Party B and calls the large model to ask questions about contract risk suggestions: For the contract clause "XXX", give modification suggestions for Party B, taking the interests of Party B into full consideration, and avoid content such as "extending the payment period of Party A", "reducing the liquidated damages ratio of Party A", "increasing the liquidated damages ratio of Party B", and "adding restrictive clauses for Party B" in the suggestions. Strictly use the format of "Modification suggestions:... " as the answer, and the number of words in the modification suggestions does not exceed 40 Chinese characters.

[0093] (10) According to the results returned by the large model, uniformly assemble a message body containing the contract clause content, contract risk category, risk description, and risk suggestion according to the format requirements.

[0094] {

[0095] "clause": "Contract clause content",

[0096] "info": {

[0097] "Risk category": "Risk category",

[0098] "Risk description": "Risk description",

[0099] "Modification suggestions": "Modification suggestions"

[0100] }

[0101] }。

[0102] For adaptive optimization, according to the risk data in the object storage, the large model is fine-tuned regularly. The operations are as follows:

[0103] (1) Personnel with permissions pull contract risk data.

[0104] (2) Divide the pulled contract risk data according to requirements to construct a training set, a validation set, and a test set.

[0105] (3) Set training strategies, including but not limited to parameters such as learning rate and batch size

[0106] (4) Set the training set to iteratively update the model parameters and perform fine-tuning training operations on the model.

[0107] (5) Use the test set to evaluate the function of the fine-tuned model to ensure that the fine-tuned model achieves the expected effect on new tasks;

[0108] 1) If the model performance meets the expected goals, it indicates that the fine-tuning is successful, and the process can be continued to the next stage;

[0109] 2) If the model performance fails to meet the expectations, go to step 3 and perform the fine-tuning operation again until the expected optimization goal is met.

[0110] (6) Deploy the optimized model to the production environment.

[0111] The method of this embodiment realizes the rapid parsing of the original contract text and supports the goal of identifying contract risks according to the two parties (Party A and Party B) by constructing and training a large-scale model.

[0112] Embodiment 2:

[0113] A contract risk management system based on a large model of the present invention includes a contract parsing module, a risk identification and analysis module, and a model iteration module.

[0114] The contract parsing module is used to perform the following: parse the contract document uploaded by the business personnel to obtain contract document information.

[0115] In this embodiment, the contract parsing module is used to parse the contract document uploaded by the business personnel, identify the key elements of the contract, and obtain contract document information including contract number, signing date, both parties of the contract, contract amount, service or product description, payment period, and liability for violation.

[0116] The risk identification and analysis module is used to perform the following: according to the role of the customer in the contract, perform risk analysis on the contract document information through a large model, identify contract risk data related to the customer role, for each customer role, construct a Prompt based on the corresponding contract risk data and call the large model for contract risk suggestions, obtain risk suggestions for each customer role, judge and review the risk suggestions for each customer role, construct a message body based on the contract risk data and the risk suggestions that pass the review, and save the message body to the object storage.

[0117] As a specific implementation of the risk identification and analysis module, this module is used to perform the following operations:

[0118] (1) Analyze the contract document information, analyze the payment terms, liability limitations, intellectual property ownership, keywords and context in the contract document information to determine the role of the customer in the contract;

[0119] (2) According to the role of the customer in the contract, identify risk terms for different customer roles through the Bert model to obtain contract risk data, where the contract risk data includes liability for breach, dispute resolution mechanism, contract change and termination conditions. Among them, the customer roles include Party A and Party B. Party A is the principal or purchaser in the contract, and Party B is the executor or service provider in the contract;

[0120] (3) For various customer roles, construct corresponding Prompts based on the contract risk data of the customer roles. Using the Prompts as input, call the large model to ask questions about the contract risk category, contract risk description, and contract risk suggestions, and obtain the contract risk category, contract risk description, and contract risk suggestions returned by the large model. Take the contract risk data, contract risk category, contract risk description, and contract risk suggestions as contract risk information;

[0121] (4) Judge the correctness of the contract risk information returned by the large model. If it passes the judgment, assemble it into a message body including contract risk data, contract risk category, contract risk description, and contract risk suggestions according to the format requirements. If it fails to pass the judgment, annotate the contract risk information, and have the business experts review the annotated contract risk information. After passing the review, assemble it into a message body including contract risk data, contract risk category, contract risk description, and contract risk suggestions according to the format requirements;

[0122] (5) Save the message body to the object storage.

[0123] The model iteration module is used to perform the following: Regularly fine-tune the large model based on the message body in the object storage.

[0124] As a specific implementation of the model iteration module, this module is used for the following operations:

[0125] (1) Authorized managers pull contract risk information from the storage object;

[0126] (2) Divide the pulled contract risk information according to requirements to construct a training set, a validation set, and a test set;

[0127] (3) Set training strategies, including the learning rate and batch size;

[0128] (4) Set the training set to iteratively update the model parameters, and perform model fine-tuning training operations based on the training set and training strategies;

[0129] (5) Evaluate the fine-tuned large model based on the validation set and the test set. If the function of the fine-tuned large model meets the expectations, deploy the fine-tuned large model to the production environment. If the function of the fine-tuned large model does not meet the expectations, perform model fine-tuning training operations again based on the training set and training strategies until the expected model fine-tuning goal is met.

[0130] The system of this embodiment can implement contract risk management by executing the method disclosed in Embodiment 1.

[0131] The above has introduced in detail the contract risk management method and system based on large models provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation on the present invention.

Claims

1. A contract risk management method based on a large model, characterized in that: The steps include: Contract analysis: parse the contract documents uploaded by business personnel to obtain contract document information; Risk identification and analysis: Based on the customer's role in the contract, the contract document information is analyzed through the big model to identify the contract risk data related to the customer role. For each customer role, a prompt is built based on the corresponding contract risk data, and the big model is called to make contract risk recommendations. The risk recommendations for each customer role are obtained, and the risk recommendations for each customer role are judged and reviewed. The message body is built based on the contract risk data and the reviewed risk recommendations, and the message body is saved in the object storage. Model iteration: Regularly fine-tune the large model based on the message body in the object storage.

2. The contract risk management method based on a large model according to claim 1 is characterized in that: During contract analysis, the contract documents uploaded by the business personnel are analyzed to identify the key elements of the contract and obtain contract document information including the contract number, signing date, both parties to the contract, contract amount, service or product description, payment period and liability for violations.

3. The contract risk management method based on a large model according to claim 1 is characterized in that: Risk identification analysis includes the following steps: Analyze contract document information, including payment terms, liability limitations, intellectual property ownership, keywords and context in the contract document information to determine the role of the client in the contract; According to the role of the customer in the contract, the risk clauses of different customer roles are identified through the Bert model to obtain contract risk data. The contract risk data includes liability breach, dispute resolution mechanism, contract change and termination conditions. Among them, the customer roles include Party A and Party B. Party A is the entrusting party or purchaser in the contract, and Party B is the executor or service provider in the contract. For various customer roles, corresponding prompts are constructed based on the contract risk data of the customer roles. The prompts are used as input to call the big model to ask questions about the contract risk category, contract risk description, and contract risk suggestions. The contract risk category, contract risk description, and contract risk suggestions returned by the big model are obtained, and the contract risk data, contract risk category, contract risk description, and contract risk suggestions are used as contract risk information. The contract risk information returned by the big model is judged for correctness. If it passes the judgment, it is assembled into a message body including contract risk data, contract risk categories, contract risk descriptions, and contract risk suggestions in accordance with the format requirements. If it fails the judgment, the contract risk information is annotated and reviewed by business experts. After the review is passed, it is assembled into a message body including contract risk data, contract risk categories, contract risk descriptions, and contract risk suggestions in accordance with the format requirements. Save the message body to object storage.

4. The contract risk management method based on a large model according to claim 1 is characterized in that: Model iteration includes the following steps: Managers with permissions pull contract risk information from storage objects; Divide the retrieved contract risk information according to requirements and construct training sets, validation sets, and test sets; Set the training strategy, including learning rate and batch size; Set the training set to iteratively update the model parameters, and perform model fine-tuning training operations based on the training set and training strategy; The fine-tuned large model is evaluated based on the validation set and the test set. If the function of the fine-tuned large model meets expectations, the fine-tuned large model is deployed in the production environment. If the function of the fine-tuned large model does not meet expectations, the model is fine-tuned again based on the training set and training strategy until the expected model fine-tuning goals are met.

5. A contract risk management system based on a large model, characterized in that: Used to perform contract risk management through a large model-based contract risk management method as described in any one of claims 1 to 4, the system comprising a contract parsing module, a risk identification and analysis module, and a model iteration module; The contract parsing module is used to perform the following operations: parsing the contract document uploaded by the business personnel to obtain the contract document information; The risk identification and analysis module is used to perform the following: perform risk analysis on the contract document information based on the customer's role in the contract through the big model, identify the contract risk data related to the customer role, build a prompt for each customer role based on the corresponding contract risk data, call the big model to make contract risk recommendations, obtain risk recommendations for each customer role, and judge and review the risk recommendations for each customer role, build a message body based on the contract risk data and the reviewed risk recommendations, and save the message body to the object storage; The model iteration module is used to perform the following: regularly fine-tune the large model according to the message body in the object storage.

6. The large model-based contract risk management system according to claim 5 is characterized in that: The contract parsing module is used to parse the contract documents uploaded by business personnel, identify the key elements of the contract, and obtain contract document information including contract number, signing date, contract parties, contract amount, service or product description, payment period and violation liability.

7. The large model-based contract risk management system according to claim 5, characterized in that: The risk identification and analysis module is used to perform the following operations: Analyze contract document information, including payment terms, liability limitations, intellectual property ownership, keywords and context in the contract document information to determine the role of the client in the contract; According to the role of the customer in the contract, the risk clauses of different customer roles are identified through the Bert model to obtain contract risk data. The contract risk data includes liability breach, dispute resolution mechanism, contract change and termination conditions. Among them, the customer roles include Party A and Party B. Party A is the entrusting party or purchaser in the contract, and Party B is the executor or service provider in the contract. For various customer roles, corresponding prompts are constructed based on the contract risk data of the customer roles. The prompts are used as input to call the big model to ask questions about the contract risk category, contract risk description, and contract risk suggestions. The contract risk category, contract risk description, and contract risk suggestions returned by the big model are obtained, and the contract risk data, contract risk category, contract risk description, and contract risk suggestions are used as contract risk information. The contract risk information returned by the big model is judged for correctness. If it passes the judgment, it is assembled into a message body including contract risk data, contract risk categories, contract risk descriptions, and contract risk suggestions in accordance with the format requirements. If it fails the judgment, the contract risk information is annotated and reviewed by business experts. After the review is passed, it is assembled into a message body including contract risk data, contract risk categories, contract risk descriptions, and contract risk suggestions in accordance with the format requirements. Save the message body to object storage.

8. The large model-based contract risk management system according to claim 5, characterized in that: The model iteration module is used to perform the following operations: Managers with permissions pull contract risk information from storage objects; Divide the retrieved contract risk information according to requirements and construct training sets, validation sets, and test sets; Set the training strategy, including learning rate and batch size; Set the training set to iteratively update the model parameters, and perform model fine-tuning training operations based on the training set and training strategy; The fine-tuned large model is evaluated based on the validation set and the test set. If the function of the fine-tuned large model meets expectations, the fine-tuned large model is deployed in the production environment. If the function of the fine-tuned large model does not meet expectations, the model is fine-tuned again based on the training set and training strategy until the expected model fine-tuning goals are met.

Images