Legal contract risk monitoring and early warning method

Abstract

The invention discloses a legal contract risk monitoring and early warning method, and particularly relates to the technical field of risk early warning. The method comprises the following steps: constructing a multi-contract risk propagation network by identifying transmissible terms in a plurality of legal contract texts, and generating a multi-contract propagation path topology network model by adopting a graph theory algorithm; identifying a cascade trigger mechanism of risk propagation terms between contracts by using time sequence causal analysis, and outputting a cascade risk propagation link; performing sensitivity grading on each risk propagation path to obtain propagation path sensitivity grading data; integrating the cascaded risk propagation link and propagation path sensitivity grading data, constructing a contract network risk propagation comprehensive evaluation model, and outputting a comprehensive risk evaluation result; comparing the comprehensive risk assessment result with a preset risk early warning threshold rule to generate corresponding risk prompt information; according to the method, the propagation trend and the diffusion path of the legal contract risk can be effectively predicted, and the timeliness and the accuracy of contract risk management and control are improved.

Description

Technical Field

[0001] This invention relates to the field of risk warning technology, and more specifically, to a method for monitoring and warning of legal contract risks. Background Technology

[0002] In the current business operations, the number and complexity of legal contracts are increasing day by day. Companies often sign a large number of contracts, and there are often certain connections between these contracts.

[0003] Existing legal contract risk management methods are usually limited to the isolated monitoring and management of individual contract risks. They do not systematically consider and effectively monitor the risk diffusion patterns and cascading effects caused by the propagation of clauses between contracts, making it difficult to achieve timely perception and dynamic prevention and control of contract risk diffusion. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a legal contract risk monitoring and early warning method to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for monitoring and early warning of legal contract risks includes the following steps: S1: Acquire and identify dissemination clauses in multiple legal contracts, extract the type characteristics, triggering conditions and scope of effect of dissemination clauses, and output the characteristic data of dissemination clauses; S2, construct a multi-contract risk propagation network based on the characteristic data of propagation clauses, establish the propagation association path between contract nodes and clauses through graph theory algorithm, and output a multi-contract propagation path topology network model; S3, based on the multi-contract propagation path topology network model, identifies the cascading triggering mechanism of propagation clauses among multiple contracts through temporal causal analysis, and outputs the cascading risk propagation link; S4, based on the multi-contract propagation path topology network model, determines the risk propagation sensitivity level of each propagation path through risk sensitivity analysis and outputs propagation path sensitivity classification data; S5 integrates cascading risk propagation links and propagation path sensitivity classification data to establish a comprehensive assessment model for contract network risk propagation and output comprehensive risk assessment results. S6, based on the comprehensive risk assessment results, triggers cross-contract cascading risk warnings based on risk warning threshold rules, and generates corresponding risk alerts.

[0006] In a preferred embodiment, S1 specifically refers to: Batch read multiple legal contract texts; Each contract is segmented into paragraphs and its clause numbers are corrected to obtain the clause text segments; Candidate propagation clauses are detected in the clause text segments, and non-candidate clauses are filtered out based on a pre-set propagation clause dictionary and syntactic dependency rules. Classify candidate propagation clauses using multiple labels and output the propagation clause type; Based on regular rules and event extraction models, trigger condition fields and scope fields are extracted from candidate propagation clause texts; The propagation clause type, trigger condition field, and scope field are combined to generate the feature data of the propagation clause.

[0007] In a preferred embodiment, S2 specifically refers to: Based on the characteristic data of propagation clauses, a multi-contract risk propagation network node is established. Each legal contract text is defined as a multi-contract risk propagation network node, and each propagation clause is defined as an edge of the propagation relationship between the multi-contract risk propagation network nodes. Based on the trigger condition field and scope field of the propagation clause, the direction of risk propagation between nodes in the multi-contract risk propagation network is determined; Graph theory algorithms are used to traverse all nodes in the multi-contract risk propagation network to determine the direct and indirect propagation relationships between the nodes. Based on the established direct and indirect propagation relationships, a multi-contract risk propagation path is constructed, generating a multi-contract propagation path topology network model.

[0008] In a preferred embodiment, S3 specifically refers to: Based on the multi-contract propagation path topology network model, a propagation clause triggering time sequence is defined. The event occurrence time corresponding to the triggering condition field of each propagation clause is defined as the starting time point of clause triggering, and the starting time of the effect or liability agreed in the scope field of the propagation clause is defined as the ending time point of clause triggering. Based on the start and end times of the triggering of the propagation clause, determine the order of risk cascading propagation among nodes in the multi-contract risk propagation network. The causal relationship of all risk cascading propagation is identified to determine the cascading triggering order between propagation clauses, thus forming a cascading risk propagation chain.

[0009] In a preferred embodiment, S4 specifically refers to: Based on the multi-contract propagation path topology network model, the propagation clause type, trigger condition field, and scope field of each propagation relationship are read; According to the preset risk weight mapping rules, the combination of the propagation clause type and the triggering condition field is mapped to the propagation relationship risk weight value; Traverse each multi-contract risk propagation path, sum up the risk weight values ​​of all propagation relationships, and obtain the path risk propagation value; By comparing the path risk propagation value with the preset risk sensitivity threshold within a range, propagation path sensitivity classification data is generated.

[0010] In a preferred embodiment, S5 specifically refers to: Based on the cascading risk propagation chain, obtain the network nodes of multi-contract risk propagation and the triggering order of propagation clauses in each cascading risk propagation chain; Based on the sensitivity classification data of the propagation path, obtain the path risk propagation value and sensitivity level corresponding to each multi-contract risk propagation path; A comprehensive risk assessment model for contract network propagation is established. Based on the multiple contract risk propagation network nodes and the triggering order of propagating clauses in the cascading risk propagation chain, the path risk propagation value and sensitivity level of each propagation path are combined and judged, and the comprehensive risk assessment result of each cascading risk propagation chain is output.

[0011] In a preferred embodiment, S6 specifically refers to: The comprehensive risk assessment results of each cascading risk propagation link are compared with the preset high-risk warning threshold and medium-risk warning threshold; When the comprehensive risk assessment result is greater than or equal to the high-risk warning threshold, the corresponding cascading risk propagation link is marked as high-risk, and a high-risk cross-contract cascading risk warning is generated. When the comprehensive risk assessment result is greater than the medium risk warning threshold but less than the high risk warning threshold, the corresponding cascading risk propagation link is marked as medium risk level, and a cross-contract cascading risk warning of medium risk level is generated. When the comprehensive risk assessment result is less than or equal to the medium risk warning threshold, the corresponding cascading risk propagation link is marked as low risk level, and a low-risk cross-contract cascading risk warning is generated.

[0012] The technical effects and advantages of the legal contract risk monitoring and early warning method of the present invention are as follows: By extracting and structuring dissemination clauses from multiple legal contract texts, this system accurately identifies potential dissemination relationships between contracts. By constructing a multi-contract risk dissemination network and generating a dissemination path topology model, it visually reveals the risk transmission pathways between contract nodes. Temporal causal analysis identifies cross-contract cascading triggering mechanisms, effectively determining the order of risk diffusion. By classifying and evaluating each dissemination path according to preset risk weights and sensitivity thresholds, it quantifies the impact of risks along different paths. By integrating cascading link and sensitivity classification data, a comprehensive evaluation model is constructed, enabling accurate assessment of the overall risk of the contract network. By dynamically comparing the comprehensive evaluation results with early warning threshold rules, risk alerts are output, significantly improving the timeliness and accuracy of risk warnings. This system achieves monitoring and early warning of contract risks from the clause level to the network level, effectively enhancing the early identification and control efficiency of cross-contract cascading risks and reducing corporate contract performance risks. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of a legal contract risk monitoring and early warning method according to the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0015] Example Figure 1 This invention provides a method for monitoring and early warning of legal contract risks, which includes the following steps: S1: Acquire and identify dissemination clauses in multiple legal contracts, extract the type characteristics, triggering conditions and scope of effect of dissemination clauses, and output the characteristic data of dissemination clauses; S2, construct a multi-contract risk propagation network based on the characteristic data of propagation clauses, establish the propagation association path between contract nodes and clauses through graph theory algorithm, and output a multi-contract propagation path topology network model; S3, based on the multi-contract propagation path topology network model, identifies the cascading triggering mechanism of propagation clauses among multiple contracts through temporal causal analysis, and outputs the cascading risk propagation link; S4, based on the multi-contract propagation path topology network model, determines the risk propagation sensitivity level of each propagation path through risk sensitivity analysis and outputs propagation path sensitivity classification data; S5 integrates cascading risk propagation links and propagation path sensitivity classification data to establish a comprehensive assessment model for contract network risk propagation and output comprehensive risk assessment results. S6, based on the comprehensive risk assessment results, triggers cross-contract cascading risk warnings based on risk warning threshold rules, and generates corresponding risk alerts.

[0016] S1: Obtain and identify dissemination clauses in multiple legal contracts, extract the type characteristics, triggering conditions, and scope of application of the dissemination clauses, and output the characteristic data of the dissemination clauses, including: Batch read multiple legal contract texts; Specifically, multiple legal contract texts are read in batches through the legal contract management database interface to ensure sufficient original contract data is obtained. Electronic documents of multiple legal contract texts are read from a specified database by calling the standard application programming interface provided by the contract management database. The contract management database can be an internal contract data system already established and maintained by the enterprise or related unit. The electronic documents of the contract texts are generally stored in readable text formats, such as PDF, DOCX, TXT, or XML. Batch reading refers to simultaneously or sequentially acquiring the complete text data of multiple contracts, such as reading one hundred or more legal contract texts at once. These legal contract texts cover various business scenarios such as business cooperation, financing and lending, or project outsourcing, to meet the data richness requirements for identifying the dissemination clauses of multiple contracts.

[0017] Each contract is segmented into paragraphs and its clause numbers are corrected to obtain the clause text segments; Specifically, a segmentation algorithm is used to initially divide the natural paragraphs of the contract text. The segmentation algorithm divides the text into segments by recognizing line breaks, tabs, and specific paragraph spacing symbols in the contract text, and outputs multiple paragraph text units with clear structures. The paragraph text units are then corrected for clause numbers: regular expressions are used to identify standard clause numbers in the contract, such as the first clause, 1.1, (1), a), etc., and each paragraph of the contract is renumbered according to the identified numbering rules, so as to ensure that all paragraphs have consistent numbers. After paragraph segmentation and clause number correction, each legal contract text is divided into multiple clause text segments with a unified numbering format.

[0018] Candidate propagation clauses are detected in the clause text segments, and non-candidate clauses are filtered out based on a pre-set propagation clause dictionary and syntactic dependency rules. Specifically, the dictionaries of dissemination clauses are lexical resources pre-constructed manually or trained by machine learning based on practical experience in the legal field. They include legal terms and keywords with dissemination characteristics, such as cross-breach, cross-acceleration, liability transmission, and diffusion of consequences of breach. The syntactic dependency rules are pre-defined language rules using dependency parsing methods in natural language processing to determine the grammatical relationship between keywords and clause descriptions, and to determine the grammatical structural features of dissemination clauses in the contract text. Through a natural language analysis model, each clause text segment is scanned one by one according to the pre-defined dictionaries of dissemination clauses and syntactic dependency rules, filtering out clause texts that do not contain dissemination characteristics, and retaining candidate clauses that meet the dissemination characteristics.

[0019] Classify candidate propagation clauses using multiple labels and output the propagation clause type; Specifically, a multi-label classification model is used to classify candidate propagation clauses. This model is a supervised learning model, trained using real-world case data within the legal field. It categorizes each candidate propagation clause text. The input to the multi-label classification model is the text content of the candidate propagation clause, and the output is the propagation clause type corresponding to the candidate propagation clause. This includes, but is not limited to, multiple categories such as cross-breach clauses, cross-acceleration clauses, liability diffusion clauses, and joint liability clauses. Through this classification method, the propagation type of each candidate propagation clause is determined.

[0020] Based on regular rules and event extraction models, trigger condition fields and scope fields are extracted from candidate propagation clause texts; Specifically, regular expression rules refer to text matching rules pre-designed based on common expression patterns of triggering conditions and scope of effect in legal texts. Regular expressions are used to locate and extract sentence fragments with specific expressions in contract texts. The event extraction model, on the other hand, is a natural language processing model based on deep learning. After specific training, it can extract field information containing time, event triggering elements, and event outcome stipulations from the text of dissemination clauses, determining the event triggering conditions and scope of effect stipulated in the dissemination clauses. For example, a contract clause may stipulate that when one party commits a specific breach, the other party's contract clauses will be automatically activated. Through the synergistic effect of regular expression rules and the event extraction model, the triggering condition field and scope of effect field of each candidate dissemination clause are extracted. The triggering condition field of a dissemination clause refers to the conditions or events stipulated in the clause that can trigger the transmission of liability or risk to other contracts, such as the occurrence of a debtor's breach of contract or a contract payment delay reaching a certain number of days. The scope of effect field of a dissemination clause refers to the scope of contracts involved in the liability or risk stipulated in the clause text, such as the project name, amount, related contract number, and name of the affected party.

[0021] The propagation clause type, trigger condition field, and scope field are combined to generate the propagation clause's feature data; Specifically, the characteristic data of each propagation clause is recorded in the form of a structured data table. This includes the labeled propagation clause type, such as cross-default type or cross-acceleration type; the conditions and start time of the event specified in the triggering condition field; and the scope of effect, contractual objects to which liability takes effect, and termination time specified in the scope of effect field. The resulting characteristic data of the propagation clause expresses the key characteristic information of each propagation clause.

[0022] S2, construct a multi-contract risk propagation network based on the characteristic data of propagation clauses, establish propagation association paths between contract nodes and clauses using graph theory algorithms, and output a multi-contract propagation path topology network model, including: Based on the characteristic data of propagation clauses, a multi-contract risk propagation network node is established. Each legal contract text is defined as a multi-contract risk propagation network node, and each propagation clause is defined as an edge of the propagation relationship between the multi-contract risk propagation network nodes. Specifically, each legal contract text corresponds to a node in a multi-contract risk propagation network. The data content of the network node includes the contract's unique identifier, information about the parties involved, the signing date, and the validity period, among other contract metadata. Each propagation clause corresponds to an edge in the propagation relationship between the multi-contract risk propagation network nodes. The data content of each edge includes the propagation clause type, trigger condition field, and scope field. To establish a network node, the feature data of all propagation clauses is traversed first. The unique identifier of the contract is extracted from the feature data, and the extracted unique identifier is used as the primary key of the multi-contract risk propagation network node to establish the node data structure, thereby determining the correspondence between the network node and the legal contract text. For the edges of the propagation relationship, the direction of propagation and the target node are determined according to the propagation clause type, trigger condition field, and scope field in the feature data of the propagation clause. The trigger condition field indicates the starting node of the edge, and the scope field indicates the ending node to which the edge points.

[0023] Based on the trigger condition field and scope field of the propagation clause, the direction of risk propagation between nodes in the multi-contract risk propagation network is determined; Specifically, the trigger condition field indicates the starting contract node where the risk propagates, i.e., the starting point of the risk propagation relationship; the scope field indicates the contract node to which the risk may spread or be transmitted, i.e., the ending point of the risk propagation relationship. Taking a cross-default type propagation clause as an example, if the trigger condition field is described as "when Contract A is overdue for more than 30 days" and the scope field is described as "Contract B may terminate early or request the debt to mature early", then the risk propagation direction defined by the propagation clause is from the multi-contract risk propagation network node corresponding to Contract A to the multi-contract risk propagation network node corresponding to Contract B. Each propagation clause determines the risk propagation direction according to the field rules.

[0024] Graph theory algorithms are used to traverse all nodes in the multi-contract risk propagation network to determine the direct and indirect propagation relationships between the nodes. Specifically, a breadth-first search algorithm is used to traverse the network nodes. Starting from each multi-contract risk propagation network node, the algorithm expands the search for direct propagation relationships between network nodes to determine direct propagation relationships. Each directly propagated node is then used as a new starting point to expand the search for other connected nodes, thus determining indirect propagation relationships through intermediate nodes. To ensure the accuracy of the traversal results, an access marking mechanism is used to record visited multi-contract risk propagation network nodes, avoiding duplicate searches and ensuring the completeness and accuracy of the traversal of direct and indirect propagation relationships. After the traversal is complete, a complete set of direct and indirect propagation relationships between each multi-contract risk propagation network node is output.

[0025] Based on the established direct and indirect propagation relationships, a multi-contract risk propagation path is constructed, and a multi-contract propagation path topology network model is generated. Specifically, each direct propagation relationship between network nodes is recorded as a basic risk propagation path and uniquely identified. Indirect propagation relationships are constructed using path expansion, connecting nodes sequentially based on the sequence of network nodes involved in the indirect propagation relationship to form a multi-level composite propagation path. Each path is also individually numbered and recorded. For example, if contract node A directly propagates risk to contract node B, and contract node B then directly propagates risk to contract node C, then there exists an indirect propagation relationship from contract node A through contract node B to contract node C. The resulting path construction forms a multi-level composite propagation path of contract node A - contract node B - contract node C. After all propagation paths are constructed, the path information is stored in a multi-contract risk propagation path data table. Each path record includes a path number, the sequence information of the multi-contract risk propagation network nodes traversed by the path, and the characteristic data of the corresponding propagation clauses. The multi-contract risk propagation path topology network model is a network graph model that includes a set of nodes in the multi-contract risk propagation network and a set of edges representing the propagation relationships between the nodes. It reflects the connection methods and directions between each node. The model extracts all path nodes and their connections from the multi-contract risk propagation path data table and uses data visualization tools or network modeling software, such as Gephi, to graphically represent the multi-contract risk propagation path topology network model. Each node in the multi-contract risk propagation network is labeled as an independent node element in the network graph model, with its name and unique contract identifier. Simultaneously, the risk propagation relationship defined by each propagation clause is represented by connecting lines with directional arrows, and the type, triggering condition field, and scope field of the propagation clause are marked on the connecting lines. The resulting multi-contract risk propagation path topology network model has a complete multi-node connection structure, and directional arrows mark the risk propagation direction between all nodes, expressing the risk propagation relationships of all contract nodes and the triggering mechanism of each risk propagation relationship.

[0026] S3, based on a multi-contract propagation path topology network model, identifies the cascading triggering mechanism of propagation clauses among multiple contracts through temporal causal analysis, and outputs the cascading risk propagation chain, including: Based on the multi-contract propagation path topology network model, a propagation clause triggering time sequence is defined. The event occurrence time corresponding to the triggering condition field of each propagation clause is defined as the starting time point of clause triggering, and the starting time of the effect or liability agreed in the scope field of the propagation clause is defined as the ending time point of clause triggering. Specifically, based on the multi-contract propagation path topology network model, a propagation clause triggering time sequence is defined to determine the cascading triggering mechanism of propagation clauses among nodes in the multi-contract risk propagation network. The propagation clause triggering time sequence includes a start time point and an end time point. The start time point of the propagation clause is defined as the time when the event corresponding to the triggering condition field in the characteristic data of the propagation clause first occurs or reaches the triggering standard. The end time point of the propagation clause is defined as the time when the liability stipulated in the scope of action field of the propagation clause takes effect, the implementation of the impact, or the actual start of risk diffusion. For example, if the trigger condition field in the propagation clause feature data is "when Contract A experiences a payment delay exceeding 30 days", then the event occurrence time corresponding to the trigger condition field, i.e., the time when the payment delay reaches 30 days, will be set as the starting point of the propagation clause trigger time series. The starting point is determined by extracting and recording it through real-time contract performance monitoring data, corporate payment record data, or contract performance log records. For the scope field in the propagation clause feature data, "Contract B may claim early termination within 10 working days after Contract A's delayed payment", then the first working day on which Contract B's right to claim early termination actually takes effect and begins to have legal impact, i.e., the earliest definite date on which Contract B can begin to exercise its right, will be determined as the termination point of the propagation clause trigger time series. Similarly, the termination point is obtained by combining the provisions in the relevant contractual clauses and determining and recording the date based on the contract performance monitoring system or corporate legal management records.

[0027] Based on the start and end times of the triggering of the propagation clause, determine the order of risk cascading propagation among nodes in the multi-contract risk propagation network. Specifically, based on the start and end times of all propagation clause triggering time sequences, the order of risk cascading propagation among nodes in the multi-contract risk propagation network is determined: The triggering time sequences of all propagation clauses are sorted according to their start times to establish a temporal ordering relationship; the end time of each propagation clause is used to determine whether its actual triggering will induce triggering events for other propagation clauses in adjacent contract nodes; a risk propagation cascading relationship is established when there is reasonable temporal continuity between the end time of a propagation clause corresponding to one contract node and the start time of a propagation clause corresponding to another contract node. For example, if the end time of a propagation clause for Contract A is January 15, 2024, and the start time of a propagation clause for Contract B is January 20, 2024, then there is a sequential risk cascading propagation relationship between Contract A and Contract B; by traversing all propagation paths in the multi-contract propagation path topology network model, the cascading relationship between the start and end times of triggering propagation clauses corresponding to each network node is determined, thus forming a data set of the order of risk cascading propagation.

[0028] The causal relationship of the sequence of all risk cascading propagation is identified to determine the cascading triggering order between propagation clauses and form a cascading risk propagation chain; Specifically, the causal relationship identification process uses a time-series causal analysis method based on Granger causality test: A time-series dataset is constructed, each containing the triggering event time series of multiple contract risk propagation network nodes, i.e., the starting time point corresponding to the propagation clause of each node; Granger causality test analysis is performed on the starting time point sequences of propagation clauses between two adjacent network nodes to determine whether a statistically significant causal relationship exists between the time series of nodes; if the Granger causality test confirms that the time series of the triggering event of the preceding network node has a predictive effect on the time series of the triggering event of the following network node and reaches a statistically significant level, such as a confidence level of 95% or higher, then a causal relationship exists; if the above conditions are not met, then a causal relationship does not exist, and a cascading propagation link is not formed. The statistical significance level can be set using historical risk event data. For example, by analyzing historical data on legal contract risk events from the past year, the false positive and false negative rates corresponding to different statistical significance thresholds can be determined. A relationship curve between the statistical significance threshold and the false positive and false negative rates can be constructed, and the optimal confidence threshold can be determined based on the principle of minimizing the false positive and false negative rates. For instance, if statistical analysis shows that the identification results of the cascading propagation chain are optimal at a 95% confidence level, then the confidence level for the Granger causality test can be set to 95%. After causal relationship identification is completed, the causal relationship sequence will constitute cascading risk propagation link data. Each cascading risk propagation link contains a complete causal sequence record from the initial triggering clause to the final affected clause. For example, the link information is expressed as: after contract node A triggers a propagating clause on a specific date, it triggers contract node B to trigger a propagating clause on a specific date, which in turn triggers the propagating clause of contract node C. All cascading risk propagation link information is stored in the cascading risk propagation link data table, including the link number, the name of each network node, the type of each propagating clause, the trigger start and end times, and the statistical confidence level of the causal relationship test.

[0029] S4, based on a multi-contract propagation path topology network model, determines the risk propagation sensitivity level of each propagation path through risk sensitivity analysis, and outputs propagation path sensitivity classification data, including: Based on the multi-contract propagation path topology network model, the propagation clause type, trigger condition field, and scope field of each propagation relationship are read; Specifically, the data corresponding to each propagation relationship originates from the characteristic data of the propagation clause and is recorded in the multi-contract propagation path data table, including the propagation clause type, the triggering condition field of the propagation clause, and the scope of the propagation clause field. The propagation clause type includes, but is not limited to, various types with clear legal characteristics such as cross-default clause type, cross-acceleration clause type, liability diffusion clause type, and joint and several liability clause type. The triggering condition field includes the event or state conditions that can trigger the activation of the propagation clause, such as the number of days of contract payment delay, the standard for occurrence of default events stipulated in the contract, the degree of deterioration of specific economic indicators, or the conditions for third-party debt default. The scope of the propagation clause records the scope of target contracts involved in the risk diffusion or liability transfer stipulated in the propagation clause, such as the affected contract identifier, the subject of liability, the range of affected amount, the affected business projects, and the effective time of the corresponding liability.

[0030] According to the preset risk weight mapping rules, the combination of the propagation clause type and the triggering condition field is mapped to the propagation relationship risk weight value; Specifically, the risk weight mapping rules are structured rule bases formed in advance based on a large amount of historical contract risk event data and expert knowledge. These rules quantify the importance and potential impact of different propagation clause types and triggering conditions on the overall risk propagation path. The method for determining the risk weight mapping rules is as follows: historical data of propagation risk events that have occurred are obtained from the enterprise's historical contract risk event database; a multi-dimensional combination feature space of propagation clause type and triggering condition fields is constructed, and quantified based on historical statistical indicators such as the frequency of risk occurrence, economic losses caused by the risk, and the scope of legal liability diffusion corresponding to each feature combination in the feature space; based on the quantification results, each feature combination is mapped to a risk weight value; the range of the risk weight value can be set using a normalization method, for example, limiting the risk weight value to between 0 and 1, where the closer the weight value is to 1, the higher the degree of impact of the corresponding propagation relationship on the overall risk; after the mapping rules are determined, they are stored in the risk weight mapping rule database in the form of structured data tables. The database records include the unique identifier of the feature combination, the propagation clause type, the triggering condition field, the mapping weight value, and the historical statistical data indicators on which the weight value is determined.

[0031] Traverse each multi-contract risk propagation path, sum up the risk weight values ​​of all propagation relationships, and obtain the path risk propagation value; Specifically, all risk propagation path information is extracted from the multi-contract propagation path topology network model. Each risk propagation path is marked with a path number and includes the sequence of all multi-contract risk propagation network nodes involved and the propagation relationship edge data between nodes. All propagation relationship edge data in each risk propagation path is read, and the propagation clause type and triggering condition field information are extracted. Based on risk weight mapping rules, the propagation clause type and triggering condition field in the propagation relationship edge data are combined and mapped to a propagation relationship risk weight value. Each propagation relationship edge corresponds to a separate risk weight value. For example, the multi-contract risk propagation path A-B-C includes two propagation relationships: A to B and B to C, resulting in risk weight values ​​of 0.9 and 0.7 respectively. The path risk propagation value is calculated by accumulating the risk weight values ​​of all propagation relationship edges in each propagation path. After traversal, the path risk propagation value is recorded in the data structure of each multi-contract risk propagation path. The path risk propagation value characterizes the overall risk impact of the path; a higher path risk propagation value indicates a greater impact of the path risk.

[0032] By comparing the path risk propagation value with the preset risk sensitivity threshold within a range, propagation path sensitivity classification data is generated. Specifically, the method for determining the risk sensitivity threshold is as follows: Based on the risk event sample data accumulated by the historical contract risk monitoring system, the risk sensitivity threshold is determined through cluster analysis or expert evaluation methods to classify different sensitivity levels. For example, risk propagation value data of all contract risk events that have occurred in the past three years are selected, and the optimal classification interval of the path risk propagation value is determined through K-means cluster analysis algorithm, dividing it into high-risk, medium-risk, and low-risk levels. For example, through cluster analysis, it is found that when the path risk propagation value is greater than or equal to 2.0, the risk propagation path is marked as high-risk; when the path risk propagation value is between 1.0 and 2.0, the risk propagation path is marked as medium-risk; and when the path risk propagation value is less than or equal to 1.0, the risk propagation path is marked as low-risk. The path risk propagation value is compared with a preset risk sensitivity threshold within an interval. Based on the comparison results, the sensitivity level of each risk propagation path is determined and marked. For example, if the path risk propagation value of a risk propagation path is 2.2, after comparing it with the preset risk sensitivity threshold within an interval, it is confirmed that the path risk propagation value of 2.2 is greater than or equal to the high-risk level threshold of 2.0, so it is marked as a high-risk level. After completing the sensitivity level labeling, the sensitivity level information of all risk propagation paths is stored in a structured manner in the propagation path sensitivity classification data table. The data structure of the propagation path sensitivity classification data table includes the path number, the path risk propagation value, and the corresponding sensitivity level. For example, the data table records the path number "Path 1001", the corresponding path risk propagation value "2.2", and the sensitivity level "high risk level".

[0033] S5 integrates cascading risk propagation links and sensitivity classification data of propagation paths to establish a comprehensive assessment model for contract network risk propagation, outputting comprehensive risk assessment results, including: Based on the cascading risk propagation chain, obtain the network nodes of multi-contract risk propagation and the triggering order of propagation clauses in each cascading risk propagation chain; Specifically, by calling the cascading risk propagation link data table that stores the cascading risk propagation link data, each cascading risk propagation link is traversed to extract the sequence of multi-contract risk propagation network nodes in the link and the triggering order of the propagation clauses triggered in the corresponding nodes. The triggering order of the multi-contract risk propagation network nodes is marked according to the results of time-series causal analysis. For example, for the cascading risk propagation link numbered 10001, the propagation clause of contract node A is triggered first, followed by the propagation clauses in contract nodes B and C in sequence. The propagation clause information triggered by each network node includes the propagation clause type, triggering condition field, and scope field.

[0034] Based on the sensitivity classification data of the propagation path, obtain the path risk propagation value and sensitivity level corresponding to each multi-contract risk propagation path; Specifically, path risk propagation value and sensitivity level information are extracted from the propagation path sensitivity grading data table; the path risk propagation value is obtained by accumulating the propagation relationship risk weight values ​​of each path in the multi-contract risk propagation path topology network model, and the corresponding sensitivity level is determined by the preset risk sensitivity threshold interval division method.

[0035] Establish a comprehensive assessment model for the propagation of contract network risks. Based on the multiple contract risk propagation network nodes and the triggering order of propagating clauses in the cascading risk propagation chain, combine and judge the path risk propagation value and sensitivity level of each propagation path, and output the comprehensive risk assessment result of each cascading risk propagation chain. Specifically, the comprehensive assessment model for contract network risk propagation is defined with the inputs being the sequence of triggering clauses for propagating clauses in the cascading risk propagation chain, the path risk propagation value, and the sensitivity level corresponding to each path. A hybrid model combining expert rules and machine learning is used for combined assessment: an initial rule set is jointly developed by experts in the legal and risk management fields. Based on practical experience, this initial rule set determines the weight of different scenarios regarding the triggering sequence of each network node on the degree of risk propagation, and considers the impact of path risk propagation value and sensitivity level on the severity of risk diffusion. Machine learning algorithms are then used to optimize the initial expert rule set, specifically employing a random forest model: utilizing historical legal contract risk event data from enterprises. Real-world risk propagation events in the database serve as training samples. Input sample features include the trigger sequence of nodes in the multi-contract risk propagation network, path risk propagation values, and sensitivity levels. The output label represents the actual level of risk loss. The optimal parameters of the model are determined through repeated training and cross-validation. Key parameters of the random forest model, such as the number of trees, feature splitting criteria, and maximum tree depth, are determined through grid search and cross-validation. For example, the number of trees can be set between 100 and 500, and the maximum depth between 5 and 10. Finally, the parameter combination that minimizes the model's evaluation error is selected. The optimized random forest model will replace the original expert rule set as the basis for judging the comprehensive evaluation model of contract network risk propagation. The system sequentially obtains the trigger order of the propagation clauses of network nodes in each cascading risk propagation link, determines the corresponding propagation path number, and extracts the path risk propagation value and sensitivity level of each path. These are used as input data for the comprehensive assessment model of contract network risk propagation. The model combines the node trigger order, path risk propagation value, and sensitivity level to generate a comprehensive risk assessment result. This result is reflected as a comprehensive risk level, including but not limited to multiple risk assessment levels such as extremely high risk, high risk, medium risk, and low risk. The comprehensive risk assessment level is determined based on historical risk loss data statistics. For example, based on historical data statistics, when the path risk propagation value is high and the sensitivity level is high risk, the model typically outputs an extremely high risk or high risk level. If the path risk propagation value is... If the sensitivity level is medium risk, the model output is usually medium risk. If the path risk propagation value is low and the sensitivity level is low risk, the model output is usually low risk. For example, for a cascaded risk propagation link numbered 10001, the triggering order of the multi-contract risk propagation network nodes is contract node A - contract node B - contract node C, the path risk propagation value is 2.4, and the sensitivity level is high risk. When input into the comprehensive risk assessment model for contract network risk propagation, after judgment, the comprehensive risk assessment result of link 10001 is extremely high risk. The comprehensive risk assessment result is stored in the form of a structured data table. Each record contains the link number, node triggering sequence, path risk propagation value, sensitivity level, and comprehensive risk level output by the model.

[0036] S6, based on the comprehensive risk assessment results and risk warning threshold rules, trigger cross-contract cascading risk warnings and generate corresponding risk alerts, including: The comprehensive risk assessment results of each cascading risk propagation link are compared with the preset high-risk warning threshold and medium-risk warning threshold; Specifically, high-risk and medium-risk warning thresholds are preset: These are obtained through a combination of cluster analysis and expert scoring based on real risk loss event data recorded in the enterprise's historical legal contract risk event database; comprehensive risk assessment results corresponding to historical contract risk events are extracted, with each event's comprehensive risk assessment result data output by the contract network risk propagation comprehensive assessment model and recorded in a structured data table, containing two key indicators: the comprehensive risk assessment value and the actual amount of risk loss; all historical comprehensive risk assessment results are normalized, for example, by linearly mapping the original comprehensive risk assessment value... The target range is 0 to 1. A K-means clustering algorithm is used, initially setting the number of clusters to 3, dividing the normalized comprehensive risk assessment data into high-risk, medium-risk, and low-risk clusters. The clustering algorithm iteratively optimizes by calculating the sum of squared errors within each cluster, aiming to minimize the intra-cluster error, to determine the optimal cluster centers for the three clusters, serving as the initial basis for classifying high, medium, and low risk levels. After clustering, domain experts, combined with statistical results of actual risk losses, fine-tune the boundaries corresponding to the cluster centers to determine the final high-risk and medium-risk warning thresholds. For example, based on a combination of historical data clustering and expert scoring, the high-risk warning threshold is set at 0.75, and the medium-risk warning threshold is set at 0.45.

[0037] When the comprehensive risk assessment result is greater than or equal to the high-risk warning threshold, the corresponding cascading risk propagation link is marked as high-risk, and a high-risk cross-contract cascading risk warning is generated. When the comprehensive risk assessment result is greater than the medium risk warning threshold but less than the high risk warning threshold, the corresponding cascading risk propagation link is marked as medium risk level, and a cross-contract cascading risk warning of medium risk level is generated. When the comprehensive risk assessment result is less than or equal to the medium risk warning threshold, the corresponding cascading risk propagation link is marked as low risk level, and a low-risk cross-contract cascading risk warning is generated. Specifically, the comprehensive risk assessment result of each cascading risk propagation link is compared with the high-risk warning threshold and the medium-risk warning threshold, respectively. Based on the comparison results, the risk level corresponding to each cascading risk propagation link is determined. When the comprehensive risk assessment result is greater than or equal to the high-risk warning threshold, the corresponding cascading risk propagation link is marked as high-risk. When the comprehensive risk assessment result is between the medium-risk and high-risk warning thresholds, the corresponding cascading risk propagation link is marked as medium-risk. When the comprehensive risk assessment result is less than or equal to the medium-risk warning threshold, the corresponding cascading risk propagation link is marked as low-risk. For example, the comprehensive risk of link number 10001... The assessment result value is 0.82, which is greater than the preset high-risk warning threshold of 0.75. Therefore, link 10001 is marked as high-risk. The comprehensive risk assessment result value of link 10002 is 0.58, which is between the medium-risk warning threshold of 0.45 and the high-risk warning threshold of 0.75. Therefore, link 10002 is marked as medium-risk. The comprehensive risk assessment result value of link 10003 is 0.32, which is less than the medium-risk warning threshold of 0.45. Therefore, link 10003 is marked as low-risk. The risk level of each link is recorded in the cross-contract risk warning level data table, which includes the link number, the comprehensive risk assessment result value, and the marked risk level. Based on the marked risk level, corresponding levels of cross-contract cascading risk alerts are generated. The generation of cross-contract cascading risk alerts includes text alerts and visual alerts. The content of the text alerts is a structured text template, designed for high-risk, medium-risk, and low-risk levels respectively. For example, for the high-risk level link 10001, the final generated alert is: The risk propagation level of risk link 10001 has reached the high-risk level, which may lead to serious cross-contract risk spread. It is recommended to immediately activate cross-contract risk emergency measures and promptly notify the contract-related parties. Meanwhile, the visualization prompts include, but are not limited to, color coding. Specifically, each cascading risk propagation link is marked with a different color in the multi-contract propagation path topology network model, with high-risk levels highlighted in red, medium-risk levels in yellow, and low-risk levels in green.

[0038] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0039] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0040] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0041] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0042] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0043] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0044] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0045] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0046] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring and early warning of legal contract risks, characterized in that, Includes the following steps: S1: Acquire and identify dissemination clauses in multiple legal contracts, extract the type characteristics, triggering conditions and scope of effect of dissemination clauses, and output the characteristic data of dissemination clauses; S2, construct a multi-contract risk propagation network based on the characteristic data of propagation clauses, establish the propagation association path between contract nodes and clauses through graph theory algorithm, and output a multi-contract propagation path topology network model; S3, based on the multi-contract propagation path topology network model, identifies the cascading triggering mechanism of propagation clauses among multiple contracts through temporal causal analysis, and outputs the cascading risk propagation link; S4, based on the multi-contract propagation path topology network model, determines the risk propagation sensitivity level of each propagation path through risk sensitivity analysis and outputs propagation path sensitivity classification data; S5 integrates cascading risk propagation links and propagation path sensitivity classification data to establish a comprehensive assessment model for contract network risk propagation and output comprehensive risk assessment results. S6, based on the comprehensive risk assessment results, triggers cross-contract cascading risk warnings based on risk warning threshold rules, and generates corresponding risk alerts.

2. The legal contract risk monitoring and early warning method according to claim 1, characterized in that, S1, specifically: Batch read multiple legal contract texts; Each contract is segmented into paragraphs and its clause numbers are corrected to obtain the clause text segments; Candidate propagation clauses are detected in the clause text segments, and non-candidate clauses are filtered out based on a pre-set propagation clause dictionary and syntactic dependency rules. Classify candidate propagation clauses using multiple labels and output the propagation clause type; Based on regular rules and event extraction models, trigger condition fields and scope fields are extracted from candidate propagation clause texts; The propagation clause type, trigger condition field, and scope field are combined to generate the feature data of the propagation clause.

3. The legal contract risk monitoring and early warning method according to claim 2, characterized in that, S2, specifically: Based on the characteristic data of propagation clauses, a multi-contract risk propagation network node is established. Each legal contract text is defined as a multi-contract risk propagation network node, and each propagation clause is defined as an edge of the propagation relationship between the multi-contract risk propagation network nodes. Based on the trigger condition field and scope field of the propagation clause, the direction of risk propagation between nodes in the multi-contract risk propagation network is determined; Graph theory algorithms are used to traverse all nodes in the multi-contract risk propagation network to determine the direct and indirect propagation relationships between the nodes. Based on the established direct and indirect propagation relationships, a multi-contract risk propagation path is constructed, generating a multi-contract propagation path topology network model.

4. The legal contract risk monitoring and early warning method according to claim 3, characterized in that, S3, specifically: Based on the multi-contract propagation path topology network model, a propagation clause triggering time sequence is defined. The event occurrence time corresponding to the triggering condition field of each propagation clause is defined as the starting time point of clause triggering, and the starting time of the effect or liability agreed in the scope field of the propagation clause is defined as the ending time point of clause triggering. Based on the start and end times of the triggering of the propagation clause, determine the order of risk cascading propagation among nodes in the multi-contract risk propagation network. The causal relationship of all risk cascading propagation is identified to determine the cascading triggering order between propagation clauses, thus forming a cascading risk propagation chain.

5. The legal contract risk monitoring and early warning method according to claim 4, characterized in that, S4, specifically: Based on the multi-contract propagation path topology network model, the propagation clause type, trigger condition field, and scope field of each propagation relationship are read; According to the preset risk weight mapping rules, the combination of the propagation clause type and the triggering condition field is mapped to the propagation relationship risk weight value; Traverse each multi-contract risk propagation path, sum up the risk weight values ​​of all propagation relationships, and obtain the path risk propagation value; By comparing the path risk propagation value with the preset risk sensitivity threshold within a range, propagation path sensitivity classification data is generated.

6. The legal contract risk monitoring and early warning method according to claim 5, characterized in that, S5, specifically: Based on the cascading risk propagation chain, obtain the network nodes of multi-contract risk propagation and the triggering order of propagation clauses in each cascading risk propagation chain; Based on the sensitivity classification data of the propagation path, obtain the path risk propagation value and sensitivity level corresponding to each multi-contract risk propagation path; A comprehensive risk assessment model for contract network propagation is established. Based on the multiple contract risk propagation network nodes and the triggering order of propagating clauses in the cascading risk propagation chain, the path risk propagation value and sensitivity level of each propagation path are combined and judged, and the comprehensive risk assessment result of each cascading risk propagation chain is output.

7. The legal contract risk monitoring and early warning method according to claim 6, characterized in that, S6, specifically: The comprehensive risk assessment results of each cascading risk propagation link are compared with the preset high-risk warning threshold and medium-risk warning threshold; When the comprehensive risk assessment result is greater than or equal to the high-risk warning threshold, the corresponding cascading risk propagation link is marked as high-risk, and a high-risk cross-contract cascading risk warning is generated. When the comprehensive risk assessment result is greater than the medium risk warning threshold but less than the high risk warning threshold, the corresponding cascading risk propagation link is marked as medium risk level, and a cross-contract cascading risk warning of medium risk level is generated. When the comprehensive risk assessment result is less than or equal to the medium risk warning threshold, the corresponding cascading risk propagation link is marked as low risk level, and a low-risk cross-contract cascading risk warning is generated.

Images