CN-122022451-A - Contract checking method, system, equipment and storage medium

CN122022451ACN 122022451 ACN122022451 ACN 122022451ACN-122022451-A

Abstract

Marking first type risk clauses in contract clauses and second type risk clauses with strong logic dependency relationship with the first type risk clauses; and carrying out similarity matching on risk clauses in the contract to be inspected and risk clauses in case information, and generating a modification suggestion of the contract to be inspected by combining the case evolution track. According to the technical scheme provided by the embodiment of the invention, the accuracy of the contract checking task can be improved.

Inventors

WU LIN
DUAN GUOHUA
LIU JIANYU
Hu Renbing
WANG SIYU
JIANG TAO
FENG WEI

Assignees

武汉数众科技有限公司

Dates

Publication Date: 20260512
Application Date: 20251229

Claims (10)

1. A contract review method, comprising: identifying contract terms in the standardized structure text, marking first type risk terms in the contract terms, and constructing an initial knowledge graph; Carrying out fine-granularity risk identification on contract clauses, marking a second type of risk clause with strong logic dependency relationship with the first type of risk clause after a certain number of contract clauses are identified, and updating an initial knowledge graph as a new node; based on the first type risk clause and the second type risk clause, screening a plurality of similar cases of the contract to be checked from the existing cases, recording case information of each similar case, and constructing a case evolution track based on a processing scheme in each case information; Performing similarity matching on the risk clauses in the contract to be checked and the risk clauses in the case information, and obtaining the risk occurrence probability and influence degree of the risk clauses in the contract to be checked based on a matching result; and generating modification suggestions of the to-be-inspected contract based on the risk occurrence probability, the influence degree and the case evolution track.
2. The method according to claim 1, wherein the process of constructing the initial knowledge-graph specifically comprises: Identifying contract clauses in the standardized structure text, and performing coarse-grained risk identification on the contract clauses to obtain risk scores of the contract clauses and preliminary risk scores of the contracts to be checked; setting the number of nodes in an initial knowledge graph based on the total number of contract terms and the preliminary risk score of the contract to be checked; marking first-class risk clauses as nodes of the initial knowledge graph based on the risk scores of the contract clauses and the number of nodes in the initial knowledge graph; And based on the logic relationship among the nodes in the initial knowledge graph, connecting the nodes, and marking the risk score and the logic relationship of the nodes.
3. The method of claim 1, further comprising, after the initial knowledge-graph update is completed: Based on the random access probability of each node and the importance probability transmitted by the upstream node, PR values of each node are obtained, and if the PR value of any node is larger than a preset PR threshold value, marking any node as a core node; and marking the two nodes as contradictory nodes if the contradictory degree between the two nodes is greater than a preset contradictory threshold value.
4. The method according to claim 1, characterized in that the process of obtaining a plurality of similar cases of the contract under examination comprises in particular: taking the first type risk clause and the second type risk clause as risk feature vectors; constructing a case retrieval vector based on the risk feature vector, the contract type vector, the participant feature vector and the transaction amount normalization vector; And (3) carrying out similarity calculation based on the case retrieval vector and the existing cases, screening similar cases from the existing cases, and sorting the similar cases according to time.
5. The method according to claim 1, wherein the process of constructing the case evolution trajectory specifically comprises: Recording case basic information, risk clauses, processing schemes and legal basis of each similar case as case information of each similar case; analyzing the processing scheme of the similar case by a sliding window method to obtain the evolution trend of the case processing standard; Based on the evolution trend of the case processing standard, predicting the future trend of the processing scheme through a linear regression model to obtain a case evolution track.
6. The method according to claim 1, wherein the similarity matching process is performed, specifically comprising: Constructing a current term vector based on risk terms and related legal bases in the contract to be checked; Calculating the risk similarity between the risk clauses in the contract to be checked and the risk clauses in the case information by adopting the improved cosine similarity; and acquiring the maximum risk similarity, and if the maximum risk similarity is larger than a risk similarity threshold, marking the risk clause in the case information corresponding to the maximum risk similarity as the risk similar clause of the risk clause in the contract to be checked.
7. The method according to claim 1, characterized in that the process of generating a modification suggestion of the contract under examination, in particular comprises: Obtaining the risk grade of the risk clause in the contract to be checked based on the risk occurrence probability and the influence degree; Preferentially generating a plurality of different modification schemes for risk clauses of high risk grades, and marking effective modification schemes; and based on the case evolution track, carrying out overall adjustment on the effective modification scheme generated by each risk clause to obtain overall modification suggestion of the to-be-inspected contract.
8. A contract review system, characterized in that the system is configured for implementing the method of any one of claims 1-7, the system comprising: the first marking module is used for extracting text contents in the contract to be checked and converting the extracted text contents into a standardized structure text; the second marking module is used for carrying out fine-granularity risk identification on contract clauses, marking a second type of risk clause with strong logic dependency relationship with the first type of risk clause after a certain number of contract clauses are identified, and updating the initial knowledge graph as a new node; The case construction module is used for screening a plurality of similar cases of the contract to be checked from the existing cases based on the first type risk clauses and the second type risk clauses, recording case information of each similar case and constructing a case evolution track based on a processing scheme in each case information; The suggestion generation module is used for carrying out similarity matching on the risk clauses in the contract to be checked and the risk clauses in the case information, obtaining the risk occurrence probability and the influence degree of the risk clauses in the contract to be checked based on the matching result, and generating a modification suggestion of the contract to be checked based on the risk occurrence probability, the influence degree and the case evolution track.
9. An electronic device, the electronic device comprising: And a memory communicatively coupled to the at least one processor; the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the steps of the method of any one of claims 1-7.
10. A computer readable storage medium storing computer instructions for causing a processor to perform the steps of the method of any one of claims 1-7.

Description

Contract checking method, system, equipment and storage medium Technical Field The invention relates to the technical field of artificial intelligence and legal science and technology intersection, in particular to a contract checking method, a system, equipment and a storage medium. Background With the development of computer technology, examination of contracts is increasingly dependent on intelligent compound co-examination systems. However, existing rule-based or search-enhanced generation (RAG) -based contract auditing systems lack deep reasoning capabilities, are difficult to accommodate for constant updates of legal regulations and judicial interpretations, and are also unable to effectively utilize historical case and similar contract processing experience, and lack mechanisms for learning and evolving from past cases, resulting in inconsistent auditing standards. Therefore, a set of more perfect contract checking mechanism needs to be designed, so that the accuracy of the contract checking task can be improved, and the trust of legal professionals can be obtained. Disclosure of Invention The invention provides a contract checking method, a contract checking system, contract checking equipment and a storage medium, so that the accuracy and automation level of a contract checking task are improved. In a first aspect, the present invention provides a contract review method, comprising: identifying contract terms in the standardized structure text, marking first type risk terms in the contract terms, and constructing an initial knowledge graph; Carrying out fine-granularity risk identification on contract clauses, marking a second type of risk clause with strong logic dependency relationship with the first type of risk clause after a certain number of contract clauses are identified, and updating an initial knowledge graph as a new node; based on the first type risk clause and the second type risk clause, screening a plurality of similar cases of the contract to be checked from the existing cases, recording case information of each similar case, and constructing a case evolution track based on a processing scheme in each case information; Performing similarity matching on the risk clauses in the contract to be checked and the risk clauses in the case information, and obtaining the risk occurrence probability and influence degree of the risk clauses in the contract to be checked based on a matching result; and generating modification suggestions of the to-be-inspected contract based on the risk occurrence probability, the influence degree and the case evolution track. Further, the process of constructing the initial knowledge graph specifically includes: Identifying contract clauses in the standardized structure text, and performing coarse-grained risk identification on the contract clauses to obtain risk scores of the contract clauses and preliminary risk scores of the contracts to be checked; setting the number of nodes in an initial knowledge graph based on the total number of contract terms and the preliminary risk score of the contract to be checked; marking first-class risk clauses as nodes of the initial knowledge graph based on the risk scores of the contract clauses and the number of nodes in the initial knowledge graph; And based on the logic relationship among the nodes in the initial knowledge graph, connecting the nodes, and marking the risk score and the logic relationship of the nodes. Further, after the initial knowledge graph updating is completed, the method further comprises: Based on the random access probability of each node and the importance probability transmitted by the upstream node, PR values of each node are obtained, and if the PR value of any node is larger than a preset PR threshold value, marking any node as a core node; and marking the two nodes as contradictory nodes if the contradictory degree between the two nodes is greater than a preset contradictory threshold value. Further, the process of obtaining the plurality of similar cases of the contract to be checked specifically includes: taking the first type risk clause and the second type risk clause as risk feature vectors; constructing a case retrieval vector based on the risk feature vector, the contract type vector, the participant feature vector and the transaction amount normalization vector; And (3) carrying out similarity calculation based on the case retrieval vector and the existing cases, screening similar cases from the existing cases, and sorting the similar cases according to time. Further, the process of constructing the case evolution track specifically includes: Recording case basic information, risk clauses, processing schemes and legal basis of each similar case as case information of each similar case; analyzing the processing scheme of the similar case by a sliding window method to obtain the evolution trend of the case processing standard; Based on the evolution trend of the case processing standa