Quantitative evaluation system based on contract file document dynamic risk map model

Abstract

The invention discloses a quantitative evaluation system based on a contract document dynamic risk map model, and relates to the field of contract document evaluation. The system comprises a data acquisition and preprocessing module, a dynamic risk map construction module and a quantitative evaluation module. And the data acquisition and preprocessing module is used for performing character recognition on the scanned contract PDF through an optical character recognition (OCR) engine. According to the method, the risk map can dynamically evolve along with changes of internal and external environments by defining the risk entities capable of being updated in real time and the incidence relation and combining a triple updating mechanism of time triggering, event triggering and rule triggering, and compared with a traditional static rule engine, the mechanism can respond to law and regulation updating, industry policy change and enterprise business adjustment in real time, and the efficiency is improved. And the problem that the risk assessment model lags behind the actual risk change is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of contract file document evaluation, and in particular to a quantitative evaluation system based on a dynamic risk map model of contract file documents. Background Art

[0002] In the field of enterprise contract management, traditional contract risk assessment methods mainly rely on static rule engines or manual review, which has significant defects. First, risk assessment relies on the experience of reviewers, which is highly subjective and inefficient, and cannot meet the needs of rapid review of massive contracts. Second, the risk model is rigid and cannot dynamically respond to updates to laws and regulations, changes in industry policies and corporate business adjustments. Third, only the contract text itself is reviewed in isolation, and there is a lack of correlation analysis between business operation risks and external environmental risks during the contract performance process, resulting in risk assessment results being out of touch with actual scenarios. Although there are attempts to introduce knowledge graphs for contract risk analysis in existing technologies, most of them remain at the static entity relationship modeling stage and have not formed a risk assessment system with dynamic evolution capabilities, making it difficult to meet the company's needs for real-time, accurate and comprehensive risk management.

[0003] Therefore, we have made improvements to this and proposed a quantitative assessment system based on the dynamic risk map model of contract file documents. Summary of the Invention

[0004] The present invention provides the following technical solutions:

[0005] The specific application is as follows:

[0006] A quantitative assessment system based on a dynamic risk map model for contract files, comprising: a data acquisition and preprocessing module, a dynamic risk map construction module, and a quantitative assessment module;

[0007] Data collection and preprocessing module: Use the optical character recognition (OCR) engine to perform character recognition on scanned contract PDFs; use the BERT pre-trained model to perform semantic analysis on unstructured text, segmenting the text by sentence, with each sentence length ≤ 512 tokens. Key entities such as the contract subject, subject matter, performance period, and breach of contract clauses are extracted, and the logical relationships between entities are extracted using the dependency parsing (LAC) model. Use ETL tools to structure and store entity and relationship data in a PostgreSQL database, establishing a relational data model consisting of four tables: contracts, entities, clauses, and relationships.

[0008] The contract subject storage format is: {subject ID, name, unified social credit code, legal representative};

[0009] The target storage format is: {target ID, name, specification, quantity, unit price};

[0010] The storage format of the fulfillment period is: {clause ID, start date, end date, fulfillment method};

[0011] The storage format of the breach of contract liability clause is: {clause ID, breach of contract situation, liquidated damages calculation method, dispute resolution method};

[0012] Dynamic risk graph construction module: uses Neo4j graph database to store risk entities and relationships, including entity definition unit, relationship unit and update unit;

[0013] Entity definition unit: defines contract text risk entities, business operation risk entities and external environment risk entities;

[0014] The storage format of the contract text risk entity is: {entity ID, clause ID, compliance score [0-100], rights and obligations balance score [0-100]};

[0015] The storage format of a business operation risk entity is: {entity ID, performance score [0-100], credit rating [AAA-E], performance rate in the last 12 months [0-100%], update timestamp}. The performance score, credit rating, and performance rate in the last 12 months support real-time dynamic updates.

[0016] The storage format of the external environment risk entity is: {entity ID, legal change risk score [0-100], industry regulatory risk score [0-100], last update time};

[0017] The association unit establishes logical dependencies within the text, text-business relationships, and external environmental impact relationships. The establishment of logical dependencies within the text uses LAC model dependency syntax analysis to extract conditional, causal, and parallel relationships between clauses, with a confidence threshold of ≥0.75. The establishment of text-business relationships uses a preset rule engine to associate contract clauses with business indicators. The number of association rules is ≥500, and each rule contains a mapping relationship between contract clause characteristics and business indicators. When the contract clause characteristics and business indicator data meet the rule trigger conditions, dynamic matching is achieved through the rule engine. External environmental impact relationships are established through an API interface. The latest legal provisions are obtained from the national laws and regulations database at 00:00 every day. The TF-IDF+Cosine algorithm is used to calculate the matching degree between legal and contract terms. When the matching degree between legal and contract terms is less than 0.8, the corresponding contract term's "Compliance" risk entity is automatically marked as pending assessment, triggering a manual review process. The system pushes the pending assessment task to the legal department. The reviewer must complete the clause reassessment and update the risk score within 2 working days.

[0018] Update unit: set three update mechanisms: time trigger, event trigger and rule trigger;

[0019] Quantitative assessment module: Multi-source data is input into the fusion model, and the GCN model algorithm is used to explore the implicit associations between risk entities to construct a three-dimensional assessment system that includes the probability of risk occurrence, the degree of impact, and the complexity of the transmission path. The dimension weights are determined through the AHP hierarchical analysis method, and the random forest model is trained based on historical cases to generate a quantitative score for the overall contract risk. Among them, low risk is 1-3 points, including 3 points, medium risk is 4-6 points, including 6 points, high risk is 7 points, and 8 points or above is extremely high risk. The risk heat map and transmission path map are displayed through a visual interface, and risk warnings such as email notifications and system pop-up windows are supported.

[0020] Among them, the specific steps of the data acquisition and preprocessing module are as follows:

[0021] SA1, start;

[0022] SA2: Use the OCR engine to perform character recognition on the scanned contract document to generate a text file. Then, use the ETL tool to collect operational data from the enterprise business system and establish an association with the contract subject ID.

[0023] SA3: Use the BERT pre-trained model to perform named entity recognition on the text and extract the contract principal entity, subject entity, performance period entity, and breach of contract liability clause entity. Specifically, the text is segmented into sentences, with each sentence length ≤ 512 tokens. The model extracts a 768-dimensional embedding vector for each token. The embedding vectors of all tokens in the sentence are aggregated into a sentence vector using the maximum pooling method. The sentence vector is then mapped to an entity label using a fully connected layer.

[0024] SA4, extract the logical relationship between entities through the dependency parsing LAC model, with a confidence threshold ≥ 0.75;

[0025] SA5. Use ETL tools to store the extracted entity and relationship data in a structured manner in a PostgreSQL database, establishing a relational data model consisting of four tables: contracts, entities, clauses, and relationships.

[0026] SA6, end.

[0027] The specific steps for building a dynamic risk map module are as follows:

[0028] SB1, start;

[0029] SB2. Import structured data into the Neo4j graph database and create risk entity nodes, including: contract text risk entity, business operation risk entity, and external environment risk entity;

[0030] SB3. Establish inter-entity relationships through a pre-set rule engine, including: logical dependencies within the text, text-business relationships, and external environmental impact relationships. The rule engine is used to establish inter-entity relationships. Specific rules are stored in the relationship_rules table. Each rule includes: rule ID, trigger condition, associated action, and effective time.

[0031] SB4. Set dynamic update triggers, including time triggers, event triggers, and rule triggers;

[0032] Time trigger: Automatically update the 'last updated time' attribute of the external environment risk entity at 2:00 am every day, and retrieve the latest legal provisions through the API interface to calculate the legal change risk score and industry regulatory risk score;

[0033] Event triggering: Subscribe to update messages from external event sources through the message queue. After receiving the event message, the message consumer parses the message content and updates the corresponding entity attributes. It calls the APOC plug-in of the graph database to trigger the recalculation of related entities. When a lawsuit involving a contract entity is detected, the "credit rating" and "litigation record" attributes of the entity are updated within 30 minutes.

[0034] Rule triggering: The default threshold is 70%, and the preset threshold can be customized. When the financial indicators of an enterprise exceed the preset threshold, the recalculation of all contract risks of the enterprise as the subject is triggered;

[0035] SB5, end.

[0036] The specific steps of the quantitative evaluation module are as follows:

[0037] SC1, start;

[0038] SC2: Extract risk entities and associated relationship data from the graph database and construct a feature matrix;

[0039] SC3. Input the feature matrix into the trained GCN model algorithm to generate the risk node embedding vector;

[0040] SC4. Based on the embedded vectors, a three-dimensional assessment system is constructed, including the probability of risk occurrence, the degree of impact, and the complexity of the transmission path. The weights of the dimensions are determined through the Analytic Hierarchy Process (AHP). The random forest model is trained based on historical cases to generate a quantitative score for the overall contract risk:

[0041] SC5. Calculate the final risk score using the following formula: Risk score = Probability of risk × 0.5396 × Impact coefficient + Base impact value × 0.2970 + Transmission path complexity index × 0.1634, where the Base impact value is determined by the corresponding value of the level output by the gradient boosting tree model. Level I is 1.0, Level II is 1.5, Level III is 2.0, and Level IV is 2.5. The Base impact value is determined by the corresponding value of the level output by the gradient boosting tree model.

[0042] SC6. Generate corresponding risk levels based on risk scores, which are divided into low, medium, high and very high;

[0043] SC7. Store the assessment results in the risk_assessment_results table and output them through the API.

[0044] SC8, end.

[0045] In SC3, the training parameters of the GCN model are:

[0046] Training dataset: Contains 5,000 historical contract risk cases, divided into training set, validation set, and test set in a ratio of 7:2:1;

[0047] Model structure: input layer dimension is 128, hidden layer dimension is 64, and output layer dimension is 32;

[0048] Training parameters: learning rate is 0.001, batch size is 64, optimizer is Adam, loss function is MSE loss;

[0049] Number of training rounds: 200;

[0050] Evaluation metrics: The test set MSE is 0.021, and the risk level prediction accuracy is 87.3%.

[0051] In SC4, the three-dimensional evaluation system is specifically constructed as follows:

[0052] Probability of risk occurrence: Calculated through a random forest model, the input features include clause compliance score, entity credit rating, and historical compliance rate, and the output probability value range is [0-1];

[0053] Impact level: Calculated through a gradient boosting tree model, the input features include the contract amount, the subject matter type, and the liquidated damages ratio of the breach of contract liability clause, and the output is the impact level, which is divided into levels I-IV;

[0054] Conduction path complexity: Calculated through graph algorithms, by calculating the shortest path length between risk entities and the PageRank value of the risk entity, and mapping the results to the range of [0-100] after linear normalization, the conduction path complexity index is comprehensively calculated.

[0055] In SC4, the weights of the three-dimensional evaluation system are determined by the AHP method. The specific steps are as follows:

[0056] SD1, start;

[0057] SD2. Construct a 3×3 judgment matrix A;

[0058] SD3. Calculate the maximum eigenvalue of the judgment matrix λmax = 3.009;

[0059] SD4, calculated consistency index CI = 0.0045;

[0060] SD5, query random consistency index RI = 0.58;

[0061] SD6, calculated consistency ratio CR = 0.0077 < 0.1, passed the consistency test;

[0062] SD7. Calculate the weight vector W = [0.5396, 0.2970, 0.1634];

[0063] SD8, end.

[0064] The quantitative evaluation module also includes:

[0065] The visualization interface is configured to use ECharts to implement a risk heat map, with color mapping rules: 1-3 points are green, 4-6 points are yellow, 7 points are orange, and 8-10 points are red;

[0066] The warning thresholds are configured as follows: 4-6 points for yellow warning, 7 points for orange warning, and 8 points and above for red warning, which correspond to pop-up notifications, email notifications to department heads, and email notifications to company executives respectively.

[0067] Compared with the prior art, the present invention has the following beneficial effects:

[0068] 1. By defining risk entities and relationships that can be updated in real time, and combining a triple update mechanism of time triggering, event triggering, and rule triggering, the risk map can dynamically evolve with changes in the internal and external environment. Compared with traditional static rule engines, this mechanism can respond to updates to laws and regulations, changes in industry policies, and adjustments to corporate business in real time, reducing the problem of risk assessment models lagging behind actual risk changes.

[0069] 2. Break through the limitations of single contract text review, deeply associate contract text risks with business operation data and external environment data, and use the external environment API interface to match the compliance of legal terms and contract terms in real time. The assessment results are closer to the risk status of the company's actual operations.

[0070] 3. The GCN graph neural network algorithm is used to mine the implicit associations between risk entities, and the AHP hierarchical analysis method is combined to determine the weights of the three-dimensional assessment system. This method reduces reliance on manual experience, reduces the subjective bias of traditional manual review, and improves the objectivity and accuracy of the assessment results.

[0071] 4. Through ECharts, risk heat maps and transmission path maps can be visualized, supporting users to interactively query risk details. Combined with threshold configuration, pop-up notifications and email warning mechanisms can be triggered to help corporate management quickly identify high-risk areas, formulate targeted risk control strategies, and improve contract management efficiency and risk prevention and control response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 This is a schematic diagram of the overall system architecture of this application;

[0073] Figure 2 This is a flow chart of the data acquisition and preprocessing module of this application;

[0074] Figure 3 This is a schematic diagram of the dynamic risk map construction module process of this application;

[0075] Figure 4 This is a schematic diagram of the update unit process of this application;

[0076] Figure 5 This is a flow chart of the quantitative evaluation module of this application;

[0077] Figure 6 This is a schematic table of risk entities and related relationships for this application;

[0078] Figure 7 This is a schematic diagram of the risk assessment visualization and early warning mechanism for this application. DETAILED DESCRIPTION

[0079] To make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions in 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 part of the embodiments of the present invention, not all of them.

[0080] Therefore, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the claimed invention, but merely represents some embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0081] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features and technical solutions therein may be combined with each other.

[0082] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0083] In the description of the present invention, it should be noted that the terms "upper" and "lower" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, or the orientations or positional relationships in which the inventive product is typically placed when in use, or the orientations or positional relationships commonly understood by those skilled in the art. Such terms are intended solely to facilitate the description of the present invention and simplify the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limitations on the present invention. Furthermore, the terms "first" and "second" and the like are used solely for distinction and should not be construed as indicating or implying relative importance.

[0084] The present invention provides the following technical solutions:

[0085] Please refer to Figure 1-Figure 7 , a quantitative assessment system based on the dynamic risk map model of contract files, including: data collection and preprocessing module, dynamic risk map construction module and quantitative assessment module;

[0086] Data collection and preprocessing module: Use the optical character recognition (OCR) engine to perform character recognition on scanned contract PDFs; use the BERT pre-trained model to perform semantic analysis on unstructured text, segmenting the text by sentence, with each sentence length ≤ 512 tokens. Key entities such as the contract subject, subject matter, performance period, and breach of contract clauses are extracted, and the logical relationships between entities are extracted using the dependency parsing (LAC) model. Use ETL tools to structure and store entity and relationship data in a PostgreSQL database, establishing a relational data model consisting of four tables: contracts, entities, clauses, and relationships.

[0087] The contract subject storage format is: {subject ID, name, unified social credit code, legal representative};

[0088] The target storage format is: {target ID, name, specification, quantity, unit price};

[0089] The storage format of the fulfillment period is: {clause ID, start date, end date, fulfillment method};

[0090] The storage format of the breach of contract liability clause is: {clause ID, breach of contract situation, liquidated damages calculation method, dispute resolution method};

[0091] Dynamic risk graph construction module: uses Neo4j graph database to store risk entities and relationships, including entity definition unit, relationship unit and update unit;

[0092] Entity definition unit: defines contract text risk entities, business operation risk entities and external environment risk entities;

[0093] The storage format of the contract text risk entity is: {entity ID, clause ID, compliance score [0-100], rights and obligations balance score [0-100]};

[0094] The storage format of a business operation risk entity is: {entity ID, performance score [0-100], credit rating [AAA-E], performance rate in the last 12 months [0-100%], update timestamp}. The performance score, credit rating, and performance rate in the last 12 months support real-time dynamic updates.

[0095] The storage format of the external environment risk entity is: {entity ID, legal change risk score [0-100], industry regulatory risk score [0-100], last update time};

[0096] The association unit establishes logical dependencies within the text, text-business relationships, and external environmental impact relationships. The establishment of logical dependencies within the text uses LAC model dependency syntax analysis to extract conditional, causal, and parallel relationships between clauses, with a confidence threshold of ≥0.75. The establishment of text-business relationships uses a preset rule engine to associate contract clauses with business indicators. The number of association rules is ≥500, and each rule contains a mapping relationship between contract clause characteristics and business indicators. When the contract clause characteristics and business indicator data meet the rule trigger conditions, dynamic matching is achieved through the rule engine. External environmental impact relationships are established through an API interface. The latest legal provisions are obtained from the national laws and regulations database at 00:00 every day. The TF-IDF+Cosine algorithm is used to calculate the matching degree between legal and contract terms. When the matching degree between legal and contract terms is less than 0.8, the corresponding contract term's "Compliance" risk entity is automatically marked as pending assessment, triggering a manual review process. The system pushes the pending assessment task to the legal department. The reviewer must complete the clause reassessment and update the risk score within 2 working days.

[0097] Update unit: set three update mechanisms: time trigger, event trigger and rule trigger;

[0098] Quantitative assessment module: Multi-source data is input into the fusion model, and the GCN model algorithm is used to explore the implicit associations between risk entities to construct a three-dimensional assessment system that includes the probability of risk occurrence, the degree of impact, and the complexity of the transmission path. The dimension weights are determined through the AHP hierarchical analysis method, and the random forest model is trained based on historical cases to generate a quantitative score for the overall contract risk. Among them, low risk is 1-3 points, including 3 points, medium risk is 4-6 points, including 6 points, high risk is 7 points, and 8 points or above is extremely high risk. The risk heat map and transmission path map are displayed through a visual interface, and risk warnings such as email notifications and system pop-up windows are supported.

[0099] Among them, the specific steps of the data acquisition and preprocessing module are as follows:

[0100] SA1, start;

[0101] SA2: Use the OCR engine to perform character recognition on the scanned contract document to generate a text file. Then, use the ETL tool to collect operational data from the enterprise business system and establish an association with the contract subject ID.

[0102] SA3: Use the BERT pre-trained model to perform named entity recognition on the text and extract the contract principal entity, subject entity, performance period entity, and breach of contract liability clause entity. Specifically, the text is segmented into sentences, with each sentence length ≤ 512 tokens. The model extracts a 768-dimensional embedding vector for each token. The embedding vectors of all tokens in the sentence are aggregated into a sentence vector using the maximum pooling method. The sentence vector is then mapped to an entity label using a fully connected layer.

[0103] SA4, extract the logical relationship between entities through the dependency parsing LAC model, with a confidence threshold ≥ 0.75;

[0104] SA5. Use ETL tools to store the extracted entity and relationship data in a structured manner in a PostgreSQL database, establishing a relational data model consisting of four tables: contracts, entities, clauses, and relationships.

[0105] SA6, end.

[0106] The specific steps for building a dynamic risk map module are as follows:

[0107] SB1, start;

[0108] SB2. Import structured data into the Neo4j graph database and create risk entity nodes, including: contract text risk entity, business operation risk entity, and external environment risk entity;

[0109] SB3. Establish inter-entity relationships through a pre-set rule engine, including: logical dependencies within the text, text-business relationships, and external environmental impact relationships. The rule engine is used to establish inter-entity relationships. Specific rules are stored in the relationship_rules table. Each rule includes: rule ID, trigger condition, associated action, and effective time.

[0110] SB4. Set dynamic update triggers, including time triggers, event triggers, and rule triggers;

[0111] Time trigger: Automatically update the 'last updated time' attribute of the external environment risk entity at 2:00 am every day, and retrieve the latest legal provisions through the API interface to calculate the legal change risk score and industry regulatory risk score;

[0112] Event triggering: Subscribe to update messages from external event sources through the message queue. After receiving the event message, the message consumer parses the message content and updates the corresponding entity attributes. It calls the APOC plug-in of the graph database to trigger the recalculation of related entities. When a lawsuit involving a contract entity is detected, the "credit rating" and "litigation record" attributes of the entity are updated within 30 minutes.

[0113] Rule triggering: The default threshold is 70%, and the preset threshold can be customized. When the financial indicators of an enterprise exceed the preset threshold, the recalculation of all contract risks of the enterprise as the subject is triggered;

[0114] SB5, end.

[0115] The specific steps of the quantitative evaluation module are as follows:

[0116] SC1, start;

[0117] SC2: Extract risk entities and associated relationship data from the graph database and construct a feature matrix;

[0118] SC3. Input the feature matrix into the trained GCN model algorithm to generate the risk node embedding vector;

[0119] SC4. Based on the embedded vectors, a three-dimensional assessment system is constructed, including the probability of risk occurrence, the degree of impact, and the complexity of the transmission path. The weights of the dimensions are determined through the Analytic Hierarchy Process (AHP). The random forest model is trained based on historical cases to generate a quantitative score for the overall contract risk:

[0120] SC5. Calculate the final risk score using the following formula: Risk score = Probability of risk × 0.5396 × Impact coefficient + Base impact value × 0.2970 + Transmission path complexity index × 0.1634, where the Base impact value is determined by the corresponding value of the level output by the gradient boosting tree model. Level I is 1.0, Level II is 1.5, Level III is 2.0, and Level IV is 2.5. The Base impact value is determined by the corresponding value of the level output by the gradient boosting tree model.

[0121] SC6. Generate corresponding risk levels based on risk scores, which are divided into low, medium, high and very high;

[0122] SC7. Store the assessment results in the risk_assessment_results table and output them through the API.

[0123] SC8, end.

[0124] In SC3, the training parameters of the GCN model are:

[0125] Training dataset: Contains 5,000 historical contract risk cases, divided into training set, validation set, and test set in a ratio of 7:2:1;

[0126] Model structure: input layer dimension is 128, hidden layer dimension is 64, and output layer dimension is 32;

[0127] Training parameters: learning rate is 0.001, batch size is 64, optimizer is Adam, loss function is MSE loss;

[0128] Number of training rounds: 200;

[0129] Evaluation metrics: The test set MSE is 0.021, and the risk level prediction accuracy is 87.3%.

[0130] In SC4, the three-dimensional evaluation system is specifically constructed as follows:

[0131] Probability of risk occurrence: Calculated through a random forest model, the input features include clause compliance score, entity credit rating, and historical compliance rate, and the output probability value range is [0-1];

[0132] Impact level: Calculated through a gradient boosting tree model, the input features include the contract amount, the subject matter type, and the liquidated damages ratio of the breach of contract liability clause, and the output is the impact level, which is divided into levels I-IV;

[0133] Conduction path complexity: Calculated through graph algorithms, by calculating the shortest path length between risk entities and the PageRank value of the risk entity, and mapping the results to the range of [0-100] after linear normalization, the conduction path complexity index is comprehensively calculated.

[0134] In SC4, the weights of the three-dimensional evaluation system are determined by the AHP method. The specific steps are as follows:

[0135] SD1, start;

[0136] SD2. Construct a 3×3 judgment matrix A;

[0137] SD3. Calculate the maximum eigenvalue of the judgment matrix λmax = 3.009;

[0138] SD4, calculated consistency index CI = 0.0045;

[0139] SD5, query random consistency index RI = 0.58;

[0140] SD6, calculated consistency ratio CR = 0.0077 < 0.1, passed the consistency test;

[0141] SD7. Calculate the weight vector W = [0.5396, 0.2970, 0.1634];

[0142] SD8, end.

[0143] The quantitative evaluation module also includes:

[0144] The visualization interface is configured to use ECharts to implement a risk heat map, with color mapping rules: 1-3 points are green, 4-6 points are yellow, 7 points are orange, and 8-10 points are red;

[0145] The warning thresholds are configured as follows: 4-6 points for yellow warning, 7 points for orange warning, and 8 points and above for red warning, which correspond to pop-up notifications, email notifications to department heads, and email notifications to company executives respectively.

[0146] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0147] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features and technical solutions therein may be combined with each other.

[0148] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0149] The above embodiments are only used to illustrate the present invention and are not intended to limit the technical solutions described in the present invention. Although this specification has described the present invention in detail with reference to the above embodiments, the present invention is not limited to the above specific implementation methods. Therefore, any modification or equivalent replacement of the present invention; and all technical solutions and improvements thereof that do not depart from the spirit and scope of the invention are included in the scope of the claims of the present invention.

Claims

1. A quantitative assessment system based on a dynamic risk map model for contract files, characterized by: include: Data acquisition and preprocessing module, dynamic risk map construction module and quantitative assessment module; Data collection and preprocessing module: Use the optical character recognition (OCR) engine to perform character recognition on scanned contract PDFs; use the BERT pre-trained model to perform semantic analysis on unstructured text, segmenting the text by sentence, with each sentence length ≤ 512 tokens. Key entities such as the contract subject, subject matter, performance period, and breach of contract clauses are extracted, and the logical relationships between entities are extracted using the dependency parsing (LAC) model. Use ETL tools to structure and store entity and relationship data in a PostgreSQL database, establishing a relational data model consisting of four tables: contracts, entities, clauses, and relationships. The contract subject storage format is: {subject ID, name, unified social credit code, legal representative}; The target storage format is: {target ID, name, specification, quantity, unit price}; The storage format of the fulfillment period is: {clause ID, start date, end date, fulfillment method}; The storage format of the breach of contract liability clause is: {clause ID, breach of contract situation, liquidated damages calculation method, dispute resolution method}; Dynamic risk graph construction module: uses Neo4j graph database to store risk entities and relationships, including entity definition unit, relationship unit and update unit; Entity definition unit: defines contract text risk entities, business operation risk entities and external environment risk entities; The storage format of the contract text risk entity is: {entity ID, clause ID, compliance score [0-100], rights and obligations balance score [0-100]}; The storage format of a business operation risk entity is: {entity ID, performance score [0-100], credit rating [AAA-E], performance rate in the last 12 months [0-100%], update timestamp}. The performance score, credit rating, and performance rate in the last 12 months support real-time dynamic updates. The storage format of the external environment risk entity is: {entity ID, legal change risk score [0-100], industry regulatory risk score [0-100], last update time}; The association unit establishes logical dependencies within the text, text-business relationships, and external environmental impact relationships. The establishment of logical dependencies within the text uses LAC model dependency syntax analysis to extract conditional, causal, and parallel relationships between clauses, with a confidence threshold of ≥0.

75. The establishment of text-business relationships uses a preset rule engine to associate contract clauses with business indicators. The number of association rules is ≥500, and each rule contains a mapping relationship between contract clause characteristics and business indicators. When the contract clause characteristics and business indicator data meet the rule trigger conditions, dynamic matching is achieved through the rule engine. External environmental impact relationships are established through an API interface. The latest legal provisions are obtained from the national laws and regulations database at 00:00 every day. The TF-IDF+Cosine algorithm is used to calculate the matching degree between legal and contract terms. When the matching degree between legal and contract terms is less than 0.8, the corresponding contract term's "Compliance" risk entity is automatically marked as pending assessment, triggering a manual review process. The system pushes the pending assessment task to the legal department. The reviewer must complete the clause reassessment and update the risk score within 2 working days. Update unit: set three update mechanisms: time trigger, event trigger and rule trigger; Quantitative assessment module: Multi-source data is input into the fusion model, and the GCN model algorithm is used to explore the implicit associations between risk entities to construct a three-dimensional assessment system that includes the probability of risk occurrence, the degree of impact, and the complexity of the transmission path. The dimension weights are determined through the AHP hierarchical analysis method, and the random forest model is trained based on historical cases to generate a quantitative score for the overall contract risk. Among them, low risk is 1-3 points, including 3 points, medium risk is 4-6 points, including 6 points, high risk is 7 points, and 8 points or above is extremely high risk. The risk heat map and transmission path map are displayed through a visual interface, and risk warnings such as email notifications and system pop-up windows are supported.

2. A quantitative assessment system based on a dynamic risk map model for contract files according to claim 1, characterized in that: The specific steps of the data acquisition and preprocessing module are as follows: SA1, start; SA2: Use the OCR engine to perform character recognition on the scanned contract document to generate a text file. Then, use the ETL tool to collect operational data from the enterprise business system and establish an association with the contract subject ID. SA3: Use the BERT pre-trained model to perform named entity recognition on the text and extract the contract principal entity, subject entity, performance period entity, and breach of contract liability clause entity. Specifically, the text is segmented into sentences, with each sentence length ≤ 512 tokens. The model extracts a 768-dimensional embedding vector for each token. The embedding vectors of all tokens in the sentence are aggregated into a sentence vector using the maximum pooling method. The sentence vector is then mapped to an entity label using a fully connected layer. SA4, extract the logical relationship between entities through the dependency parsing LAC model, with a confidence threshold ≥ 0.75; SA5. Use ETL tools to store the extracted entity and relationship data in a structured manner in a PostgreSQL database, establishing a relational data model consisting of four tables: contracts, entities, clauses, and relationships. SA6, end.

3. The quantitative assessment system based on the contract file document dynamic risk map model according to claim 1 is characterized in that: The specific steps of the dynamic risk map construction module are as follows: SB1, start; SB2. Import structured data into the Neo4j graph database and create risk entity nodes, including: contract text risk entity, business operation risk entity, and external environment risk entity; SB3. Establish inter-entity relationships through a pre-set rule engine, including: logical dependencies within the text, text-business relationships, and external environmental impact relationships. The rule engine is used to establish inter-entity relationships. Specific rules are stored in the relationship_rules table. Each rule includes: rule ID, trigger condition, associated action, and effective time. SB4. Set dynamic update triggers, including time triggers, event triggers, and rule triggers; Time trigger: Automatically update the 'last updated time' attribute of the external environment risk entity at 2:00 am every day, and retrieve the latest legal provisions through the API interface to calculate the legal change risk score and industry regulatory risk score; Event triggering: Subscribe to update messages from external event sources through the message queue. After receiving the event message, the message consumer parses the message content and updates the corresponding entity attributes. It calls the APOC plug-in of the graph database to trigger the recalculation of related entities. When a lawsuit involving a contract party is detected, the "credit rating" and "litigation record" attributes of the party are updated within 30 minutes. Rule triggering: The default threshold is 70%, and the preset threshold can be customized. When the financial indicators of an enterprise exceed the preset threshold, the recalculation of all contract risks of the enterprise as the subject is triggered; SB5, end.

4. The quantitative assessment system based on the contract file document dynamic risk map model according to claim 1 is characterized in that: The specific steps of the quantitative evaluation module are as follows: SC1, start; SC2: Extract risk entities and associated relationship data from the graph database and construct a feature matrix; SC3. Input the feature matrix into the trained GCN model algorithm to generate the risk node embedding vector; SC4. Based on the embedded vectors, a three-dimensional assessment system is constructed, including the probability of risk occurrence, the degree of impact, and the complexity of the transmission path. The weights of the dimensions are determined through the Analytic Hierarchy Process (AHP). The random forest model is trained based on historical cases to generate a quantitative score for the overall contract risk: SC5. Calculate the final risk score using the following formula: Risk score = Probability of risk × 0.5396 × Impact coefficient + Base impact value × 0.2970 + Transmission path complexity index × 0.1634, where the Base impact value is determined by the corresponding value of the level output by the gradient boosting tree model. Level I is 1.0, Level II is 1.5, Level III is 2.0, and Level IV is 2.

5. The Base impact value is determined by the corresponding value of the level output by the gradient boosting tree model. SC6. Generate corresponding risk levels based on risk scores, which are divided into low, medium, high and very high; SC7. Store the assessment results in the risk_assessment_results table and output them through the API. SC8, end.

5. The quantitative assessment system based on the contract file document dynamic risk map model according to claim 4 is characterized in that: In the SC3, the training parameters of the GCN model are: Training dataset: Contains 5,000 historical contract risk cases, divided into training set, validation set, and test set in a ratio of 7:2:1; Model structure: input layer dimension is 128, hidden layer dimension is 64, and output layer dimension is 32; Training parameters: learning rate is 0.001, batch size is 64, optimizer is Adam, loss function is MSE loss; Number of training rounds: 200; Evaluation metrics: The test set MSE is 0.021, and the risk level prediction accuracy is 87.3%.

6. The quantitative assessment system based on the contract file document dynamic risk map model according to claim 4 is characterized in that: In the SC4, the three-dimensional evaluation system is specifically constructed as follows: Probability of risk occurrence: Calculated through a random forest model, the input features include clause compliance score, entity credit rating, and historical compliance rate, and the output probability value range is [0-1]; Impact level: Calculated through a gradient boosting tree model, the input features include the contract amount, the subject matter type, and the liquidated damages ratio of the breach of contract liability clause, and the output is the impact level, which is divided into levels I-IV; Conduction path complexity: Calculated through graph algorithms, by calculating the shortest path length between risk entities and the PageRank value of the risk entity, and mapping the results to the range of [0-100] after linear normalization, the conduction path complexity index is comprehensively calculated.

7. The quantitative assessment system based on the contract file document dynamic risk map model according to claim 4 is characterized in that: In the SC4, the weights of the three-dimensional evaluation system are determined by the AHP (Analytical Hierarchy Process), and the specific steps are as follows: SD1, start; SD2. Construct a 3×3 judgment matrix A; SD3. Calculate the maximum eigenvalue of the judgment matrix λmax = 3.009; SD4, calculated consistency index CI = 0.0045; SD5, query random consistency index RI = 0.58; SD6, calculated consistency ratio CR = 0.0077 < 0.1, passed the consistency test; SD7. Calculate the weight vector W = [0.5396, 0.2970, 0.1634]; SD8, end.

8. The quantitative assessment system based on the contract file document dynamic risk map model according to claim 1 is characterized in that: The quantitative evaluation module also includes: The visualization interface is configured to use ECharts to implement a risk heat map, with color mapping rules: 1-3 points are green, 4-6 points are yellow, 7 points are orange, and 8-10 points are red; The warning thresholds are configured as follows: 4-6 points for yellow warning, 7 points for orange warning, and 8 points and above for red warning, which correspond to pop-up notifications, email notifications to department heads, and email notifications to company executives respectively.

Images