Policy dynamic monitoring method and system based on multi-agent autonomous collaboration

Abstract

The invention relates to the technical field of law and regulation monitoring, and provides a policy dynamic monitoring method and system based on multi-agent autonomous collaboration, and the method comprises the steps: selecting the most similar old policy file from a policy knowledge base for a newly published policy file, and calculating the semantic change degree of the newly published policy file and the most similar old policy file; for a newly published policy file, calculating the maximum semantic similarity between the newly published policy file and the description text of each service in the service background knowledge base to obtain a service association degree, and performing weighted summation on the semantic change degree and the service association degree to obtain a comprehensive influence score of the newly published policy file on each service; and evaluating a risk level for each service through a preset rule or a large language model based on the newly published policy file, the comprehensive influence score and the description text of the service. And regulation risk assessment is upgraded from experience driving to data driving, so that the compliance working efficiency is remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of regulatory monitoring technology, and in particular relates to a method and system for dynamic policy monitoring based on multi-agent autonomous collaboration. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In China's current consumer finance industry, compliance is the lifeline of business development. Regulatory policies are characterized by: broad sources (People's Bank of China, State Financial Regulatory Commission, local financial regulatory bureaus, etc.), high frequency (frequent release of regulations, notices, measures, draft opinions, etc.), and significant impact (directly affecting core aspects such as product design, marketing, risk control, and post-loan management). Currently, companies mainly rely on the following methods to cope with regulatory changes, but all of them have significant pain points: Manual monitoring and interpretation: Relying on legal or compliance teams to manually refresh regulatory websites and read industry news is inefficient, slow to respond, prone to omissions or misinterpretations, and costly in terms of manpower. From the release of regulations to the completion of impact analysis within the company, it often takes several days or even weeks.

[0004] Traditional public opinion monitoring systems rely on keyword matching (such as "consumer finance" and "interest rate cap") for information capture and early warning. Their main drawbacks are low signal-to-noise ratio and a lack of in-depth understanding. These systems cannot distinguish between general news mentions and the release of key regulatory clauses, nor can they understand the internal logic of regulations, identify subtle changes between clauses, or provide any insights into the impact on business.

[0005] Passive question-and-answer systems (such as basic RAG): While these systems can answer user questions about specific regulations, they are essentially passive knowledge bases. They cannot proactively inform users that "new regulations have been issued," nor can they analyze the potential impact of these new regulations on the company's existing business. Compliance personnel must first know of the existence of the new regulations before they can ask the system questions, which fails to address the most critical issue in compliance work: "first-time discovery and response."

[0006] In summary, the existing solutions cannot meet the core demands of consumer finance institutions for "proactive discovery, timely analysis, and accurate assessment" of regulatory changes. Summary of the Invention

[0007] To address the technical problems mentioned above, this invention provides a policy dynamic monitoring method and system based on multi-agent autonomous collaboration. It comprehensively considers the importance of policy changes themselves and their relevance to business operations, assesses the impact of policy changes on each business, and evaluates the business risk level based on this assessment. This upgrades regulatory risk assessment from experience-driven to data-driven, significantly improving the efficiency of compliance work.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of this invention provides a policy dynamic monitoring method based on multi-agent autonomous collaboration, comprising: Obtain newly released policy documents; For newly released policy documents, select the most similar old policy document from the policy knowledge base and calculate the semantic change degree between the newly released policy document and the most similar old policy document; For newly released policy documents, the maximum semantic similarity between the policy document and the descriptive text of each business in the business background knowledge base is calculated to obtain the business relevance. The semantic change degree and the business relevance are weighted and summed to obtain the comprehensive impact score of the newly released policy document on each business. Based on the newly released policy document, the comprehensive impact score and the descriptive text of the business, a risk level is assessed for each business through preset rules or a large language model.

[0009] Furthermore, the semantic variability is: ;in, This represents the high-dimensional vector of the newly released policy document after transformation by the embedding model.

[0010] This represents the high-dimensional vector of the most similar old policy document after being transformed by the embedding model.

[0011] Furthermore, it also includes generating summaries of newly released policy documents using a large language model.

[0012] Furthermore, it also includes: classifying newly released policy documents according to semantic change degree to obtain change categories; and generating change summaries based on change categories, newly released policy documents, and the most similar old policy documents through a large language model.

[0013] A second aspect of the present invention provides a policy dynamic monitoring system based on multi-agent autonomous collaboration, comprising: The monitoring module is configured to: retrieve newly released policy documents; The parsing module is configured to: for a newly released policy document, select the most similar old policy document from the policy knowledge base and calculate the semantic change between the newly released policy document and the most similar old policy document; The assessment module is configured to: calculate the maximum semantic similarity between a newly released policy document and the descriptive text of each business in the business background knowledge base to obtain the business relevance; weight the semantic change degree and the business relevance degree to obtain the comprehensive impact score of the newly released policy document on each business; and, based on the newly released policy document, the comprehensive impact score, and the descriptive text of the business, assess a risk level for each business through preset rules or a large language model.

[0014] Furthermore, the semantic variability is: ;in, This represents the high-dimensional vector of the newly released policy document after transformation by the embedding model. This represents the high-dimensional vector of the most similar old policy document after being transformed by the embedding model.

[0015] Furthermore, the parsing module is also configured to generate a summary of the newly released policy document using a large language model.

[0016] Furthermore, the parsing module is also configured to: classify newly released policy documents according to semantic change degree to obtain change categories; and generate change summaries based on change categories, newly released policy documents, and the most similar old policy documents through a large language model.

[0017] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the policy dynamic monitoring method based on multi-agent autonomous collaboration as described above.

[0018] A fourth aspect of the present invention provides a computer device including a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein the processor executes the program to implement the steps of the policy dynamic monitoring method based on multi-agent autonomous collaboration as described above.

[0019] Compared with the prior art, the beneficial effects of the present invention are: This invention comprehensively considers the importance of policy changes themselves and their relevance to business operations, assesses the impact of policy changes on each business activity, and conducts a business risk level assessment based on this assessment. This upgrades regulatory risk assessment from experience-driven to data-driven, significantly improving the efficiency of compliance work.

[0020] This invention reduces labor costs, freeing compliance personnel from tedious and repetitive tasks of collecting, reading, and performing preliminary analysis of regulations. It significantly improves compliance efficiency, allowing them to focus on more valuable tasks such as developing compliance strategies, providing business guidance, and facilitating cross-departmental communication, thereby maximizing organizational effectiveness. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1 This is a flowchart of a policy dynamic monitoring method based on multi-agent autonomous collaboration according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of data and intelligent flow pipeline in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the structure of a computer device according to Embodiment 4 of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0024] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] Example 1 This embodiment provides a policy dynamic monitoring method based on multi-agent autonomous collaboration.

[0026] This embodiment provides a policy dynamic monitoring method based on multi-agent autonomous collaboration, which can simulate expert workflow and automatically complete the entire process from monitoring to impact analysis.

[0027] This embodiment provides a policy dynamic monitoring method based on multi-agent autonomous collaboration, such as... Figure 2 As shown, it relies on four core agents, three supporting knowledge bases, and a distribution module, which work together to form a closed-loop data and intelligence flow pipeline. Unlike a single LLM call, this embodiment breaks down complex tasks into multiple agents with clearly defined roles, and they communicate and collaborate through standardized information flows (such as "new regulation clues" and "change summaries"), exhibiting high modularity, scalability, and robustness.

[0028] This embodiment provides a policy dynamic monitoring method based on multi-agent autonomous collaboration, such as... Figure 1 As shown, it includes the following steps: Step 1: Monitoring.

[0029] The monitoring agent (Scout Agent) scans multiple pre-defined regulatory agency websites periodically (e.g., every 5 minutes) or in real time; by comparing changes in page structure or content hash values, it efficiently identifies newly released policy documents (regulations, notices, drafts for comments, etc.) and passes the links or content of the new documents as clues to the parsing agent.

[0030] Step 2, Parsing.

[0031] After receiving clues, the Analyst Agent automatically performs the following tasks: (1) Content acquisition and cleaning: Download the full text, remove HTML tags, advertisements and other irrelevant information, and retain the main text.

[0032] (2) Core summary generation: Using the summarization capabilities of LLM, generate a “one-sentence summary” and a “list of key points” to quickly clarify the main points of the regulations.

[0033] (3) Difference Analysis: Retrieve existing effective regulations with similar themes to the new regulations from the policy knowledge base (e.g., the regulatory knowledge base). Through semantic comparison, accurately identify the "new clauses", "revised clauses" and "repealed clauses" of the new regulations relative to the old regulations. That is, for newly released policy documents, calculate the semantic change degree with each old policy document in the policy knowledge base, and classify the newly released policy documents according to the semantic change degree to obtain the change category. Based on the change category, the newly released policy document and the most similar old policy document, use LLM to output a structured "change summary".

[0034] Step 3, Assessment.

[0035] The impact assessment agent is the core of the decision-making process. It receives a "change summary" and performs the following actions: (1) Business association: Read the business background knowledge base, which stores information such as the company's product manuals, business process diagrams, and target customer profiles.

[0036] (2) Impact reasoning: Semantically match each "change clause" with the business background knowledge base and select products with a comprehensive impact score higher than the threshold. For example, if a new clause about "prohibiting lending to college students" is found, it can be automatically associated with a product called "Youth Campus Loan" within the company and judged to have a high compliance risk.

[0037] In this embodiment, a fusion reasoning approach combining business background knowledge and legal knowledge is employed: by embedding unstructured internal business documents (product introductions, user manuals, etc.) and external regulatory texts into the same high-dimensional vector space, "language alignment" of the two types of knowledge is achieved. This enables the impact assessment agent to perform semantic matching and logical reasoning across domains, accurately "connecting" regulatory requirements with specific business scenarios.

[0038] In summary, regulation-oriented semantic diff analysis is key to distinguishing it from traditional text comparison, as it compares not only literal differences but also semantic connotations. This process can be abstracted into the following model: For a clause in a new regulation... Find the most relevant corresponding clause in the old regulations knowledge base. (May be empty).

[0039] (a) Semantic Change Score (SCS): This measures the semantic difference between the old and new clause text embeddings by calculating the cosine distance between them. ;in, The high-dimensional vector representing the text of the terms after being transformed by the embedding model; It is the cosine similarity function; the range of SCS is The closer the value is to 1, the greater the semantic change; if the old clause If it does not exist, SCS can be defined as the maximum value and marked as "new".

[0040] (b) Business Relevance Score (BRS): This measures relevance by calculating the maximum semantic similarity between a new clause and the text describing a specific business (product). ;in, This represents the company's i-th business (such as a credit product). This is the first description of the business. A text fragment (such as a paragraph in a product manual); the value range of BRS is... A higher value indicates a closer connection between the new terms and the business.

[0041] (c) Overall Impact Score (OIS): This score is a weighted sum of semantic change and business relevance, used to rank risks. The overall impact score is as follows: ;in, and It is a non-negative weight that can be set based on experience to balance the "importance of the change itself" and the "relevance of the change to the business"; the higher the OIS score of the clause-business pair, the more important the impact point needs to be warned.

[0042] This formalized model transforms impact assessment from a subjective judgment into a quantifiable and rankable computational process.

[0043] (3) Risk rating: Based on newly released policy documents, comprehensive impact scores, and business background knowledge, a risk level (e.g., high risk, warning, alert) is assigned to each product impact assessment through preset rules or LLM judgment.

[0044] Step 4: Notification.

[0045] The Notifier Agent is responsible for effectively communicating the analysis results, specifically including: (1) Report integration: integrate the regulatory summary, change summary, business impact analysis and risk level into a clear and easy-to-read "Regulatory Change Intelligence Briefing".

[0046] (2) Targeted Push: Access the user profile database and push personalized information based on the recipient's role (such as product manager, risk control director) and area of ​​interest. For example, regulatory changes regarding marketing and promotion are mainly pushed to the marketing department, while information regarding data reporting is pushed to the IT and data departments.

[0047] Step 5: Multi-channel distribution: Send the briefing to WeChat Work, DingTalk, email system or internal compliance dashboard via API.

[0048] This embodiment provides a policy dynamic monitoring method based on multi-agent autonomous collaboration, which realizes a qualitative change from "passive query" to "proactive early warning", solves the problem of lag in manual monitoring, and buys enterprises valuable response time.

[0049] This embodiment provides a policy dynamic monitoring method based on multi-agent autonomous collaboration. Based on LLM deep semantic understanding, it significantly improves the accuracy and depth of analysis, can penetrate the surface of the text, accurately identify the substantive changes in regulations, and combine them with the company's own situation to conduct impact assessment, outputting insights that are "relevant to me".

[0050] This embodiment provides a policy dynamic monitoring method based on multi-agent autonomous collaboration, which reduces labor costs, frees compliance personnel from tedious and repetitive work of collecting, reading and preliminary analysis of regulations, significantly improves compliance work efficiency, and allows them to focus on more valuable compliance strategy formulation, business guidance and cross-departmental communication, thereby maximizing organizational effectiveness.

[0051] This embodiment provides a policy dynamic monitoring method based on multi-agent autonomous collaboration, which effectively reduces the operational and compliance risks of enterprises, helps enterprises avoid potential violations in the early stages of product design and business promotion, avoids regulatory penalties, economic losses and reputational damage caused by compliance issues, and safeguards the sound operation of enterprises.

[0052] Example 2 This embodiment provides a policy dynamic monitoring system based on multi-agent autonomous collaboration, which specifically includes: The monitoring module is configured to: retrieve newly released policy documents; The parsing module is configured to: for a newly released policy document, select the most similar old policy document from the policy knowledge base and calculate the semantic change between the newly released policy document and the most similar old policy document; The assessment module is configured to: calculate the maximum semantic similarity between a newly released policy document and the descriptive text of each business in the business background knowledge base to obtain the business relevance; weight the semantic change degree and the business relevance degree to obtain the comprehensive impact score of the newly released policy document on each business; and, based on the newly released policy document, the comprehensive impact score, and the descriptive text of the business, assess a risk level for each business through preset rules or a large language model.

[0053] Furthermore, the semantic variability is: ;in, This represents the high-dimensional vector of the newly released policy document after transformation by the embedding model. This represents the high-dimensional vector of the most similar old policy document after being transformed by the embedding model.

[0054] Furthermore, the parsing module is also configured to generate a summary of the newly released policy document using a large language model.

[0055] Furthermore, the parsing module is also configured to: classify newly released policy documents according to semantic change degree to obtain change categories; and generate change summaries based on change categories, newly released policy documents, and the most similar old policy documents through a large language model.

[0056] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0057] Example 3 This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of a policy dynamic monitoring method based on multi-agent autonomous collaboration as described in Embodiment 1 above.

[0058] Example 4 This embodiment provides a computer device, such as... Figure 3 As shown, the system includes a display device, an input device, a computer-readable storage medium (volatile memory and non-volatile storage medium), a processor, a communication interface (i.e., a network interface), and a computer program stored on the computer-readable storage medium and executable on the processor. The processor, communication interface, and computer-readable storage medium can be connected via a bus or other means. The communication interface is used to receive and send data, and when the processor executes the program, it implements the steps of the policy dynamic monitoring method based on multi-agent autonomous collaboration described in Embodiment 1 above.

[0059] Any references to memory, storage, database, or other media used in this application and embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0060] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0061] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0062] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0063] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A policy dynamic monitoring method based on multi-agent autonomous collaboration, characterized in that, include: Obtain newly released policy documents; For newly released policy documents, select the most similar old policy document from the policy knowledge base and calculate the semantic change degree between the newly released policy document and the most similar old policy document; For newly released policy documents, the maximum semantic similarity between the policy document and the descriptive text of each business in the business background knowledge base is calculated to obtain the business relevance. The semantic change degree and the business relevance are weighted and summed to obtain the comprehensive impact score of the newly released policy document on each business. Based on the newly released policy document, the comprehensive impact score and the descriptive text of the business, a risk level is assessed for each business through preset rules or a large language model.

2. The policy dynamic monitoring method based on multi-agent autonomous collaboration as described in claim 1, characterized in that, The degree of semantic change is: ;in, This represents the high-dimensional vector of the newly released policy document after transformation by the embedding model. This represents the high-dimensional vector of the most similar old policy document after being transformed by the embedding model.

3. The policy dynamic monitoring method based on multi-agent autonomous collaboration as described in claim 1, characterized in that, Also includes: For newly released policy documents, a summary of the policy documents is generated using a large language model.

4. The policy dynamic monitoring method based on multi-agent autonomous collaboration as described in claim 1, characterized in that, Also includes: Newly released policy documents are classified according to their semantic change degree to obtain change categories; Based on the change category, newly released policy documents, and the most similar old policy documents, a change summary is generated using a large language model.

5. A policy dynamic monitoring system based on multi-agent autonomous collaboration, characterized in that, include: The monitoring module is configured to: retrieve newly released policy documents; The parsing module is configured to: for a newly released policy document, select the most similar old policy document from the policy knowledge base and calculate the semantic change between the newly released policy document and the most similar old policy document; The assessment module is configured to: calculate the maximum semantic similarity between a newly released policy document and the descriptive text of each business in the business background knowledge base to obtain the business relevance; weight the semantic change degree and the business relevance degree to obtain the comprehensive impact score of the newly released policy document on each business; and, based on the newly released policy document, the comprehensive impact score, and the descriptive text of the business, assess a risk level for each business through preset rules or a large language model.

6. A policy dynamic monitoring system based on multi-agent autonomous collaboration as described in claim 5, characterized in that, The degree of semantic change is: ;in, This represents the high-dimensional vector of the newly released policy document after transformation by the embedding model. This represents the high-dimensional vector of the most similar old policy document after being transformed by the embedding model.

7. A policy dynamic monitoring system based on multi-agent autonomous collaboration as described in claim 5, characterized in that, The parsing module is also configured to generate a summary of newly released policy documents using a large language model.

8. A policy dynamic monitoring system based on multi-agent autonomous collaboration as described in claim 5, characterized in that, The parsing module is further configured to: classify newly released policy documents according to semantic change degree to obtain change categories; and generate change summaries based on change categories, newly released policy documents, and the most similar old policy documents through a large language model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the policy dynamic monitoring method based on multi-agent autonomous collaboration as described in any one of claims 1-4.

10. A computer device comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the policy dynamic monitoring method based on multi-agent autonomous collaboration as described in any one of claims 1-4.

Images