CN-121996840-A - System, matching method and medium for intelligently matching government benefit-enterprise policy of enterprise

CN121996840ACN 121996840 ACN121996840 ACN 121996840ACN-121996840-A

Abstract

The invention discloses a system, a matching method and a medium for intelligently matching government benefit-enterprise policies, which can comprehensively and efficiently acquire policy information, screen the policies and calculate the matching degree based on the similarity between policy semantic vectors formed by the policy information and enterprise portrait vectors formed by the enterprise information, and finally realize accurate matching of the policies to enterprises.

Inventors

ZHOU YIJIE
GU ZHUOPING
ZHANG YI
Liu Chuankuan

Assignees

中通服软件科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251226

Claims (10)

1. A system for intelligently matching government benefit-enterprise policies of enterprises is characterized by comprising a data crawling module, a policy screening and filtering module, a large language model analysis module, an enterprise portrait construction module, a similarity retrieval module and an AI discrimination analysis module, wherein the data crawling module is used for collecting policy information, the policy screening and filtering module screens and filters the policy information based on keywords and filtering rules to obtain candidate policy files with priority labels, the large language model analysis module introduces a conditional action structure extraction mechanism to disassemble the candidate policy files to obtain structured policy contents, the enterprise portrait construction module forms enterprise portrait vectors according to enterprise user information, the similarity retrieval module is used for converting the structured policy contents into policy semantic vectors suitable for semantic retrieval and calculating similarity between the policy semantic vectors and the enterprise portrait vectors to obtain a candidate policy set, and the AI discrimination analysis module is used for analyzing and comparing each piece of the structured policy contents in the candidate policy set with the enterprise portrait input large language model to judge the matching degree between the current policy and the enterprise portrait and obtaining the structured policy contents based on the matching degree.
2. The system for matching government benefit-enterprise policies according to claim 1, wherein the similarity retrieval module comprises a vectorization construction module, a vector retrieval module and a candidate policy generation module, the vectorization construction module converts structured policy content into policy semantic vectors suitable for semantic detection and stores the policy semantic vectors into a vector database, the vector detection module is respectively connected with the enterprise portrait construction module and the vectorization construction module, the vector detection module performs similarity retrieval on the policy semantic vectors and the enterprise portrait vectors, and the candidate policy generation module gathers policies with similarity higher than a similarity threshold to form a candidate policy set.
3. The system for intelligently matching government benefit policies of an enterprise according to claim 1, wherein the AI discriminant analysis module comprises a matching module, a cross-validation module and a logic consistency detection module, wherein the matching module is used for comparing structured policy content of each policy in the candidate policy set with attribute values in the enterprise representation input large language models and outputting matching degree between the current policy and the enterprise representation based on comparison results, the cross-validation module is used for simultaneously mobilizing at least two different large language models to participate in comparison flow of the structured policy content and attribute values in the enterprise representation, and the logic consistency detection module is used for checking condition information related to the policies output by the large language models to obtain the policies finally pushed to the enterprise.
4. The system for intelligently matching government benefit policies according to claim 3, wherein the AI discriminant analysis module includes a risk analysis module for identifying whether there is a compliance risk of the condition information and enterprise history data involved in the push policy.
5. The system for intelligently matching government benefit policies of an enterprise according to claim 1, further comprising a visual output display module, wherein the visual output display module is configured to perform structural arrangement on the matching policies pushed to the enterprise by the AI discriminant analysis module, and visually output a policy recommendation list and detailed reports to the enterprise.
6. The system for intelligently matching government benefit policies according to claim 1, wherein the policy filtering module comprises a keyword module and a rule module, wherein the keyword module is used for identifying and extracting keywords in the policy information, and the rule module is used for filtering the policy information based on syntax rules and judging whether the filtered policy information is related to a benefit theme.
7. The system for intelligently matching government benefit policies according to claim 1, wherein the large language model parsing module comprises a condition module for extracting and describing condition information in candidate policy documents, and an action module for extracting benefit support mode information related to the current policy.
8. The system for intelligently matching government benefit policies according to claim 7, wherein the large language model parsing module includes a lifecycle management module for injecting time-based relevant lifecycle fields for each candidate policy file.
9. A matching method for intelligently matching government benefit policies of enterprises based on the system implementation of any one of claims 1-8, comprising the following steps: step 1, crawling the disclosed policy information through a data crawling module to form an original policy data set; step 2, screening and filtering the policy information based on the keywords and the filtering rules through a policy screening and filtering module to obtain candidate policy files with priority labels; Step3, inputting the candidate policy file into a large language model, and carrying out semantic analysis and automatic summarization on the text of the candidate policy file by the large language model to generate structured policy contents at least comprising a policy name, an applicable condition, a supporting mode, an application flow and a valid period; Step 4, based on condition information, benefit and enterprise support mode information and related life cycle fields in the structured policy contents, converting the structured policy contents into policy semantic vectors suitable for semantic retrieval through a similarity retrieval module; Step 5, inputting the enterprise data into an enterprise portrait construction module, and carrying out semantic analysis and vectorization processing on the enterprise data by the enterprise portrait construction module to obtain an enterprise portrait vector; Step 6, calculating the similarity between the policy semantic vector and the enterprise portrait vector, sorting the policies from high value to low value according to the similarity, and extracting the first n policies to form a candidate policy set; step 7, inputting the candidate policy set and the enterprise portraits into a large language model, extracting applicable conditions in the candidate policy set and enterprise characteristics in the enterprise portraits by the large language model, and comparing and matching the using conditions with the enterprise characteristics; and 8, sorting the policies based on the matching degree from high to low, and extracting the first m policies to push to enterprises.
10. A medium having stored therein a computer program for an enterprise to intelligently match government benefit policies, the computer program when executed implementing the method of claim 9.

Description

System, matching method and medium for intelligently matching government benefit-enterprise policy of enterprise Technical Field The invention belongs to the technical field of computer data artificial intelligence processing, and particularly relates to a system, a matching method and a medium for intelligently matching government benefit-enterprise policies. Background Various levels of government typically issue benefit-to-enterprise policies through official websites, government public platforms, weChat public numbers, and the like. The enterprise user can review the policy text by means of keyword retrieval and classified browsing. A government service integrated platform is built in a part of areas, and centralized policy issuing and searching services are provided. Enterprises can browse policy announcements, download policy files and conduct reporting operations to a certain extent on the same platform. In addition to government platforms and information aggregation systems, many enterprises also acquire policy information through channels such as accounting offices, legal service institutions, campus service centers, etc., and professional staff provides policy interpretation and declaration coaching. Although the existing government service platform, policy information aggregation website and manual consultation service alleviate the difficulty of enterprises to acquire policy information to a certain extent, the following technical problems and disadvantages still exist: 1. The information is scattered, the whole coverage is difficult, the policy release channels of different levels and departments are various, the information is updated frequently, enterprises need to search among a plurality of websites or systems repeatedly, and policies related to the enterprises are easy to miss. 2. The policy text is unstructured, the understanding threshold is high, most of policy files are presented in the forms of bulletins, official documents and the like, the clauses are numerous, the expression is complex, a unified structured information extraction mode is lacking, and enterprises are difficult to quickly understand and apply. 3. The existing platform mainly uses keyword search or classification index, and is difficult to carry out accurate policy matching and recommendation by combining information such as industry, scale, qualification, development stage and the like of enterprises. 4. The manual interpretation cost is high, the efficiency is low, enterprises, especially small and medium-sized enterprises, often rely on third party consultation institutions to conduct policy interpretation and screening, extra cost is increased, and the efficiency and popularity of policy benefit are reduced. Therefore, based on the problems in the prior art in the process of matching enterprises with policies, the invention discloses a system, a matching method and a medium for intelligently matching government benefits and enterprises policies for enterprises. Disclosure of Invention The invention discloses a system, a matching method and a medium for intelligently matching government benefit-enterprise policies, which can comprehensively and efficiently acquire policy information, screen the policies and calculate the matching degree based on the similarity between policy semantic vectors formed by the policy information and enterprise portrait vectors formed by the enterprise information, and finally realize accurate matching of the policies to enterprises. The invention is realized by the following technical scheme: The system comprises a data crawling module, a policy screening and filtering module, a large language model analysis module, an enterprise portrait construction module, a similarity retrieval module and an AI discrimination analysis module, wherein the data crawling module is used for collecting policy information, the policy screening and filtering module screens and filters the policy information based on keywords and filtering rules to obtain candidate policy files with priority labels, the large language model analysis module introduces a conditional action structure extraction mechanism to disassemble the candidate policy files to obtain structured policy contents, the enterprise portrait construction module forms enterprise portrait vectors according to enterprise user information, the similarity retrieval module is used for converting the structured policy contents into policy semantic vectors suitable for semantic retrieval, calculating similarity between the policy semantic vectors and the enterprise portrait vectors to obtain candidate policy sets, and the AI discrimination analysis module carries out analysis comparison on each piece of the structured policy contents in the candidate policy sets and the enterprise portrait input large language model to judge the matching degree between the current policy and the enterprise portrait and pushes the corresponding policy to the enterprise b