Method and device for policy-enterprise intelligent matching

Abstract

The invention relates to the technical field of artificial intelligence and policy services, and provides a policy-enterprise intelligent matching method and device, and the method comprises the steps: carrying out the multi-stage granulation disassembly and structured conversion of a policy original text, and generating machine-readable structured policy data; integrating the multi-source enterprise data to construct a multi-dimensional enterprise data portrait; performing automatic comparison and bidirectional matching calculation on the structured policy data and the enterprise data portrait, and outputting a matching result; and calling a large model based on a matching result to carry out deep analysis, and generating a customized policy cultivation planning report. According to the method, the non-structured policy text is converted into the computable structured condition, accurate matching of the policy and the enterprise from the industry level to the condition level is achieved, the problems that a traditional matching mode is low in efficiency and insufficient in precision are solved, meanwhile, strategic interpretation and planning are provided for the matching result through a large model, and the matching efficiency is improved. And the intelligent level and the practical value of the policy service are obviously improved.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and policy service technology, specifically to a method and apparatus for intelligent matching of policy and enterprise. Background Technology

[0002] In the current policy service field, achieving effective matching between policies and enterprises remains a long-standing technical challenge. Traditional solutions primarily rely on two methods: keyword-based retrieval systems and human experts for interpretation and matching. While keyword retrieval systems can quickly filter documents containing specific terms from a massive amount of policies, they inherently cannot understand the deeper meaning of policies and complex application conditions. For example, when a policy requires that "the applicant enterprise's R&D investment in the previous year accounted for no less than 5%", keyword matching may only recognize the term "R&D investment" but cannot understand the crucial quantitative condition of "no less than 5%", let alone automatically compare it with the enterprise's actual financial data. This results in extremely low matching accuracy, generating a large amount of irrelevant information, requiring enterprise users to invest significant time in secondary screening. While human matching is relatively accurate, it heavily relies on the personal experience of experts, is inefficient, costly, and difficult to scale, failing to cope with the ever-increasing number of policies and the real-time needs of enterprises. The fundamental problem lies in the fact that policy texts are inherently unstructured natural language, while computer systems excel at processing structured, quantitative data. This semantic gap prevents computer systems from reading, understanding, and executing policy requirements in the same way humans do. This results in existing automated matching technologies remaining at a superficial level of "industry matching" or "keyword matching," failing to achieve precise "condition-level" matching based on specific application criteria. Therefore, there is an urgent need for an innovative technological solution that can transform unstructured policy texts into machine-understandable and computable data, and on this basis, achieve automated, high-precision matching and intelligent analysis. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a method and apparatus for intelligent matching of policies and enterprises. By transforming unstructured policy texts into computable structured conditions, it achieves precise matching of policies and enterprises from the "industry level" to the "condition level," solving the problems of low efficiency and insufficient accuracy of traditional matching methods.

[0004] To achieve the above objectives, the embodiments of this invention provide the following technical solutions:

[0005] This application provides a method for intelligent policy-enterprise matching, comprising: obtaining the original policy text and performing multi-level granular decomposition and structural transformation on the original policy text to generate machine-readable and computable structured policy data; collecting and integrating enterprise data from multiple sources to construct a multi-dimensional enterprise data profile containing quantitative indicators; automatically comparing and matching the structured policy data with the multi-dimensional enterprise data profile to output a matching result including matching degree and policy suitability; and based on the matching result, calling a large model for in-depth analysis and reasoning to generate a customized policy cultivation planning report.

[0006] Furthermore, the process of obtaining the original policy text and performing multi-level granular decomposition and structural transformation on the original policy text to generate machine-readable and computable structured policy data specifically includes: collecting and recording the basic attribute information of the policy, which at least includes the policy name, issuing department, application time window, subsidy amount range, and applicable regional restrictions; decomposing the overall policy into several independent application project categories, and creating a category description for each project category that includes project background, purpose, and core application guidance; for each application project category, extracting specific application conditions, support levels, and key elements of the application cycle from the corresponding text content to form a granular attribute set; and converting the granular attribute set described in the text into structured policy data with parameter codes, parameter names, logical comparison methods, and specific comparison values.

[0007] Furthermore, before converting the granular attribute set described in the text into structured policy data with parameter encoding, parameter name, logical comparison method, and specific comparison value, the method further includes: performing data cleaning and standardization on the extracted granular attribute set. The standardization process includes format unification using regular expressions and semantic disambiguation and normalization based on knowledge graphs to generate cleaned policy elements. A unique feature label is assigned to the cleaned policy elements, and the feature label includes a label encoding and a label name.

[0008] Furthermore, the process of collecting and integrating enterprise data from multiple sources to construct a multi-dimensional enterprise data profile containing quantitative indicators includes: acquiring enterprise business registration information, financial data, intellectual property information, R&D investment data, and project initiation information through a combination of automatic import via interface and manual input; cleaning, aligning, and standardizing the data from different sources to generate a standardized dataset; and constructing a multi-dimensional enterprise data profile containing quantitative evaluation indicators based on the standardized dataset, wherein the quantitative evaluation indicators are used at least for direct comparison with structured policy data.

[0009] Furthermore, the step of automatically comparing and matching the structured policy data with the enterprise data profiles to output matching results including matching degree and policy suitability includes: comparing one or more target structured policy data with multiple multi-dimensional enterprise data profiles, calculating the matching degree between each enterprise and the target structured policy data, and generating a first matching result set with enterprises as entities; comparing the multi-dimensional enterprise data profile of a target enterprise with multiple structured policy data in the policy library, calculating the matching degree between the target enterprise and each structured policy data, and generating a second matching result set with the structured policy data as entities; integrating and sorting the entities in the first matching result set or the second matching result set according to the corresponding matching degree and predefined rules, and outputting the corresponding matching result list.

[0010] Furthermore, the step of calling a large model for in-depth analysis and reasoning based on the matching results to generate a customized policy cultivation planning report includes: selecting any of the first matching result set or the second matching result set, and combining the corresponding original policy text, the multi-dimensional enterprise data profile, and matching details to form a prompt context; submitting the prompt context to the large model through an API interface, and requesting the large model to perform policy interpretation, enterprise status analysis, and strategic planning reasoning, and generating analysis results; receiving the analysis results returned by the large model, and formatting the analysis results to generate a customized policy cultivation planning report.

[0011] Accordingly, this application also provides an apparatus for intelligent policy-enterprise matching, comprising: a policy processing module for acquiring the original policy text and performing multi-level granular decomposition and structural transformation on the original policy text to generate machine-readable and computable structured policy data; an enterprise data management module for collecting and integrating enterprise data from multiple sources to construct a multi-dimensional enterprise data profile containing quantitative indicators; an intelligent matching module for automatically comparing and matching the structured policy data with the multi-dimensional enterprise data profile, and outputting a matching result including matching degree and policy suitability; and a large model analysis module for calling a large model for in-depth analysis and reasoning based on the matching result to generate a customized policy cultivation planning report.

[0012] Furthermore, the policy processing module also includes: a collection unit, used to collect and record basic attribute information of the policy, the basic attribute information including at least the policy name, issuing department, application time window, subsidy amount range, and applicable regional restrictions; a decomposition unit, used to decompose the overall policy into several independent application project categories, and create a category description for each project category that includes project background, purpose, and core application guidance; an extraction unit, used to extract specific application conditions, support levels, and key elements of the application cycle from the corresponding text content for each application project category, forming a granular attribute set; and a conversion unit, used to convert the granular attribute set described in the text into structured policy data with parameter codes, parameter names, logical comparison methods, and specific comparison values.

[0013] Furthermore, the enterprise data management module also includes: an acquisition unit, used to acquire enterprise business registration information, financial data, intellectual property information, R&D investment data, and project initiation information through a combination of automatic import via interface and manual input; a preprocessing unit, used to clean, align, and standardize the format of data from different sources to generate a standardized dataset; and a construction unit, used to construct a multi-dimensional enterprise data profile containing quantitative evaluation indicators based on the standardized dataset, wherein the quantitative evaluation indicators are used at least for direct comparison with structured policy data.

[0014] Furthermore, the intelligent matching module further includes: a first matching unit, used to compare one or more target structured policy data with multiple multi-dimensional enterprise data profiles, calculate the matching degree between each enterprise and the target structured policy data, and generate a first matching result set with enterprises as entities; a second matching unit, used to compare a target enterprise's multi-dimensional enterprise data profile with multiple structured policy data in the policy database, calculate the matching degree between the target enterprise and each structured policy data, and generate a second matching result set with the structured policy data as entities; and a sorting unit, used to integrate and sort the entities in the first matching result set or the second matching result set according to the corresponding matching degree and predefined rules, and output the corresponding matching result list.

[0015] Furthermore, the large-scale model analysis module also includes: a fusion unit, used to select any of the first matching result set or the second matching result set, and combine it with the corresponding original policy text, the multi-dimensional enterprise data profile, and matching details to form a prompt context; a parsing unit, used to submit the prompt context to the large-scale model through an API interface, and request the large-scale model to perform policy interpretation, enterprise status analysis, and strategic planning reasoning, and generate analysis results; and a generation unit, used to receive the analysis results returned by the large-scale model, and format the analysis results to generate a customized policy cultivation planning report.

[0016] The beneficial effects of this invention are as follows: by breaking down policies into multi-level granular components, the structure and machine readability of policies are achieved, solving the problem of difficulty in understanding policies; by integrating enterprise data from multiple dimensions to construct profiles, precise matching at the parameter level is achieved; bidirectional matching is supported, meeting the needs of enterprises and policies for two-way search; customized cultivation plans are generated based on large models, providing long-term strategic guidance for enterprises; and the integration of the entire process improves matching efficiency and value, demonstrating significant practicality and innovation. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a method for intelligent matching of policy and enterprise provided in an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of the structure of a system for intelligent matching of policy and enterprise provided in an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram of a device structure for intelligent matching of policy and enterprise provided in an embodiment of the present invention. Detailed Implementation

[0020] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0021] In this invention, the terms "system" and "network" are used interchangeably. "Multiple" refers to two or more; therefore, in this invention, "multiple" can also be understood as "at least two." "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / ", unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, it should be understood that in the description of this invention, terms such as "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance or order.

[0022] Example 1

[0023] In the existing policy-enterprise matching process, the unstructured policy texts make it difficult for enterprises to quickly locate suitable policies; the matching only stays at the macro level and lacks precision; it only supports one-way matching, and enterprises cannot reverse-engineer policies; the matching results lack in-depth analysis and cannot support enterprises' long-term planning; and the data processing process is fragmented and inefficient.

[0024] like Figure 1 As shown in the figure, this embodiment provides a method for intelligent matching of policy and enterprise, including: obtaining the original policy text and performing multi-level granular decomposition and structural transformation on the original policy text to generate machine-readable and computable structured policy data; collecting and integrating enterprise data from multiple sources to construct a multi-dimensional enterprise data profile containing quantitative indicators; automatically comparing and matching the structured policy data with the multi-dimensional enterprise data profile to output a matching result including matching degree and policy suitability; and based on the matching result, calling a large model for in-depth analysis and reasoning to generate a customized policy cultivation planning report.

[0025] In one possible embodiment, the process begins with a policy structuring step. After obtaining the original policy text, it undergoes multi-level granular decomposition and structural transformation to obtain structured policy data containing at least a project configuration layer, a project category layer, and a parameter condition layer. The parameter condition layer includes quantifiable comparison conditions, transforming the policy from unstructured text into a machine-readable and computable form. Next, an enterprise profile construction step is performed, aggregating multi-dimensional enterprise operational data to form an enterprise data profile. Then, a two-way matching step is executed, matching the parameter conditions of the structured policy data with the corresponding data in the enterprise data profile to generate results supporting both forward and reverse matching. Finally, an intelligent planning generation step is performed, inputting the matching results and related data into a large model to generate a planning report containing policy adaptation analysis and cultivation suggestions. Simultaneously, a data cleaning step is performed during the policy structuring and enterprise profile construction processes, using regular expressions and knowledge graph technology to unify the format and disambiguate semantics.

[0026] By breaking down policies into multi-level granular components, unstructured policy texts are transformed into machine-readable structured data containing project category and parameter condition layers, significantly lowering the threshold for policy understanding and allowing enterprises to quickly locate core conditions without reading every word. Multi-dimensional integration of enterprise operating data constructs a unified profile covering key dimensions such as business registration, finance, and intellectual property, ensuring a comprehensive characterization of enterprise features and providing a data foundation for precise parameter-level matching. It supports both forward matching ("policy → enterprise") and reverse matching ("enterprise → policy"), meeting the policy delivery needs of government departments while addressing the pain point of enterprises proactively seeking policies. The introduction of large-scale models for in-depth analysis of matching results generates planning reports containing adaptation analysis and cultivation suggestions, rather than simple adaptation conclusions, providing enterprises with strategic guidance from short-term application to long-term planning. The integration of policy decomposition, data cleaning, matching calculation, and large-scale model analysis forms an integrated processing system, avoiding efficiency losses caused by fragmented processes and improving overall matching efficiency.

[0027] In this embodiment of the invention, the step of obtaining the original policy text and performing multi-level granular decomposition and structural transformation on the original policy text to generate machine-readable and computable structured policy data specifically includes: collecting and recording the basic attribute information of the policy, which at least includes the policy name, issuing department, application time window, subsidy amount range, and applicable regional restrictions; decomposing the overall policy into several independent application project categories, and creating a category description for each project category that includes project background, purpose, and core application guidance; for each application project category, extracting specific application conditions, support levels, and key elements of the application cycle from the corresponding text content to form a granular attribute set; and converting the granular attribute set described in the text into structured policy data with parameter codes, parameter names, logical comparison methods, and specific comparison values.

[0028] The current policy breakdown only scratches the surface, failing to delve into details such as application requirements and support levels, resulting in low structuring and insufficient matching accuracy.

[0029] In one possible implementation, the project configuration sub-step is first executed. This involves collecting policy metadata information through methods such as parsing the original policy text and connecting to government data interfaces. This metadata includes the policy's source department (e.g., "Provincial Department of Science and Technology"), the validity period (e.g., "January 2025 - December 2025"), and subsidy information (e.g., "maximum 5 million yuan"), and then categorizing and recording this information in preset fields. Next, the category decomposition sub-step is executed. Based on the "I, II, III" or other chapter divisions in the original policy text, or the description of core support directions, the policy is decomposed into multiple independent project categories, such as "High-tech Enterprise Certification Subsidy" and "Enterprise R&D Platform Construction Subsidy." Record the background of each project category, such as "to promote industrial upgrading" and the purpose, such as "to enhance the enterprise's R&D capabilities," and clarify the differences in application for different categories. Finally, execute granular sub-steps to extract granular attributes from the original policy text for each project category, such as "must be established for at least 3 years," support level, such as "subsidy at 20% of R&D investment," and application cycle, such as "application in March every year." Then, use natural language processing technology to transform these textual attributes into structured conditions containing parameter codes, comparison methods, and comparison values. For example, "must be established for at least 3 years" is transformed into "parameter code PAR002, comparison method ≥, comparison value 3 years."

[0030] The project configuration sub-step collects metadata such as policy source and validity period to clarify basic policy information and prevent enterprises from making ineffective preparations due to missing application deadlines or misunderstanding the policy issuing entity. The category decomposition sub-step breaks down the policy into independent project categories and records the background purpose, clarifying the differences between different support directions within the policy, such as the distinction between "technology R&D subsidy" and "talent introduction subsidy" under the same policy, to prevent enterprises from confusing the application categories. The granularization sub-step further decomposes the application requirements, support intensity, and other attributes, and transforms them into structured conditions with parameter codes and comparison methods. It transforms vague descriptions such as "high R&D investment" into precise conditions such as "R&D investment ratio ≥ 15% (parameter code PAR001, comparison method ≥)," providing an operational basis for automated matching calculations and completely solving the problem of insufficient matching accuracy caused by vague conditions.

[0031] In existing technologies, policy granular attributes are not cleaned and standardized before being converted into structured data, resulting in semantic ambiguity, inconsistent formats, and other problems. This leads to poor quality of structured data and affects matching accuracy.

[0032] In this embodiment of the invention, before converting the granular attribute set described in the text into structured policy data with parameter encoding, parameter name, logical comparison method, and specific comparison value, the method further includes: performing data cleaning and standardization on the extracted granular attribute set. The standardization process includes format unification using regular expressions and semantic disambiguation and normalization based on knowledge graphs to generate cleaned policy elements. A unique feature label is assigned to the cleaned policy elements, and the feature label includes a label encoding and a label name.

[0033] In one possible implementation, the granular attributes extracted from the project categories are first cleaned. Regular expressions are used to standardize the format of data such as dates, amounts, and percentages in the attributes, for example, unifying "subsidy amount 5 million" and "subsidy 5 million yuan" into "subsidy amount: 5,000,000 yuan". Then, a preset policy knowledge graph is invoked to perform semantic disambiguation on ambiguous terms, such as distinguishing "technology-based enterprise" and "high-tech enterprise" according to the knowledge graph definition to ensure that the semantics of each attribute are clear. Next, the cleaned granular attributes are labeled with features. A unique label code is assigned to each attribute according to preset coding rules, and a corresponding label name is also labeled. For example, the attribute "R&D investment ratio ≥ 15%" is assigned the label code TAG001 and the label name "R&D investment ratio condition", and the attribute "employee size ≥ 100 people" is assigned the label code TAG002 and the label name "employee size condition", finally forming a set of granular attributes with standard labels.

[0034] Data cleaning is performed on granular attributes. Regular expressions are used to unify the format of dates, amounts, etc., for example, unifying "March 2025" and "2025.03" into "2025-03". Semantic ambiguity is eliminated through knowledge graphs, for example, unifying "R&D expenses" and "R&D investment" into the same concept, ensuring that the data format and semantics are consistent. Feature annotation is performed, assigning a unique label code and name to each attribute. For example, "R&D investment ratio condition" corresponds to the label code TAG001. This facilitates quick location of attributes during subsequent data management and clarifies the attribute correspondence during matching calculations, avoiding comparison errors caused by confusing attribute labels. This improves the quality of structured data from the source and ensures matching accuracy.

[0035] Existing enterprise data collection methods are simplistic, data integration is low, and quantitative evaluation indicators are lacking, resulting in incomplete enterprise profiles and insufficient matching accuracy.

[0036] In this embodiment of the invention, the step of collecting and integrating enterprise data from multiple sources to construct a multi-dimensional enterprise data profile containing quantitative indicators includes: acquiring enterprise business registration information, financial data, intellectual property information, R&D investment data, and project initiation information through a combination of automatic import via interface and manual input; cleaning, aligning, and standardizing the data from different sources to generate a standardized dataset; and constructing a multi-dimensional enterprise data profile containing quantitative evaluation indicators based on the standardized dataset, wherein the quantitative evaluation indicators are used at least for direct comparison with structured policy data.

[0037] In one possible implementation, enterprise data is first acquired through a data collection process. On one hand, this involves connecting to official data interfaces such as those of the industrial and commercial departments, tax systems, and intellectual property offices to automatically import data such as enterprise registration information, annual financial statements, and authorized patent information. On the other hand, a manual input interface is provided for enterprise staff to supplement data that cannot be obtained through interfaces, such as R&D project approvals, special industry qualification certificates, and awards. Next, the collected multi-source data is preprocessed. Duplicate and erroneous data are cleaned up, such as deleting duplicate registration records of the same patent and correcting calculation errors in financial data. Then, the data is aligned to unify the accounting period of financial data and the statistical caliber of patent data, generating a standardized dataset with consistent format and semantics. Finally, an enterprise data profile is constructed based on the standardized dataset. Quantifiable indicators such as "R&D investment ratio," "number of invention patents," "employee size," and "years of establishment" are extracted from the dataset as evaluation dimensions. These indicators are then integrated by category to form a multi-dimensional enterprise data profile that can be directly compared with policy parameters.

[0038] The data collection method combines automatic import via API with manual entry. API import ensures the real-time performance and accuracy of high-frequency data such as business registration and financial data, while manual entry supplements low-frequency but critical data such as R&D project initiation and special qualifications, avoiding data gaps caused by a single collection method. Multi-source data is cleaned and aligned, for example, by standardizing account names across different financial systems and aligning data from different statistical periods, resolving data format issues and improving data quality. Enterprise profiles with quantitative evaluation indicators are constructed, transforming abstract characteristics such as "R&D capability" and "innovation level" into quantifiable indicators such as "R&D investment ratio" and "number of invention patents," ensuring that enterprise profiles can be directly compared with the quantitative parameters of policies, fundamentally solving the problems of incomplete profiles and insufficient matching accuracy.

[0039] Existing matching methods can only achieve one-way, macro-level matching, and cannot simultaneously support two-way matching centered on policies and enterprises, and the integration of matching results is low.

[0040] In this embodiment of the invention, the step of automatically comparing and matching the structured policy data with the enterprise data profile, and outputting a matching result including matching degree and policy suitability, includes: comparing one or more target structured policy data with multiple multi-dimensional enterprise data profiles, calculating the matching degree between each enterprise and the target structured policy data, and generating a first matching result set with enterprises as entities; comparing the multi-dimensional enterprise data profile of a target enterprise with multiple structured policy data in the policy library, calculating the matching degree between the target enterprise and each structured policy data, and generating a second matching result set with the structured policy data as entities; integrating and sorting the entities in the first matching result set or the second matching result set according to the corresponding matching degree and predefined rules, and outputting the corresponding matching result list.

[0041] In one possible embodiment, the matching calculation rules are first determined using a weighted scoring algorithm. Each parameter in the structured policy data is assigned a weight; core conditions such as "R&D investment ratio" are weighted at 0.3, and secondary conditions such as "employee size" are weighted at 0.1. Scores are then calculated based on the matching between the enterprise data and the conditions. A perfect match receives full marks, a partial match receives proportional marks, and a mismatch receives 0 marks. Finally, the scores for each condition are summed to obtain a comprehensive matching score, and matching levels such as "five stars" and "four stars" are assigned based on the score. Next, a two-way matching operation is performed. For forward matching, the target structured policy data is input, and the system iterates through the enterprise profile database, matching the policy parameters with each enterprise profile. Data comparison is performed, a comprehensive matching score is calculated, and a first matching result set is generated, with each enterprise as the entity and including the score and matching level. If reverse matching is performed, the target enterprise data profile is input, the system traverses the policy database, compares the enterprise data with the parameters of each policy, calculates the comprehensive matching score, and generates a second matching result set with the policy as the entity. Finally, the matching results are integrated and sorted, with the first and second matching result sets sorted from high to low according to the comprehensive matching score. At the same time, core matching conditions such as "R&D investment ratio 18% ≥ 15% (matched)" and non-matching conditions such as "employee size 80 < 100 (non-matched)" are marked in the results, forming a clearly structured matching result list and outputting it.

[0042] The system employs a weighted scoring algorithm to calculate the comprehensive matching score. This algorithm considers not only the number of successfully matched parameters but also assigns different weights based on the importance of each parameter. For example, "R&D investment ratio" has a higher weight than "employee size," ensuring the matching results better align with core policy needs and avoiding misjudgments caused by simple counting. It supports a two-way matching mode: a "policy → enterprise" mode to meet the needs of government departments for batch screening of eligible enterprises, and an "enterprise → policy" mode to address the pain point of enterprises actively searching for applicable policies, covering different usage scenarios. Matching results are sorted by comprehensive score and integrated into a list, clearly displaying key information such as matching degree and core matching conditions. Enterprises or government departments can quickly locate the optimal matching target without manual screening, improving the efficiency of result utilization.

[0043] Existing technologies only provide simple matching results, lacking in-depth analysis based on large models, and thus cannot provide enterprises with customized, strategically valuable cultivation plans.

[0044] In this embodiment of the invention, the step of calling a large model to perform in-depth analysis and reasoning based on the matching results to generate a customized policy cultivation planning report includes: selecting any of the first matching result set or the second matching result set, and combining the corresponding original policy text, the multi-dimensional enterprise data profile, and matching details to form a prompt context; submitting the prompt context to the large model through an API interface, and requesting the large model to perform policy interpretation, enterprise status analysis, and strategic planning reasoning, and generating analysis results; receiving the analysis results returned by the large model, and formatting the analysis results to generate a customized policy cultivation planning report.

[0045] In one possible implementation, a large model prompt template is first constructed. This template includes core information such as enterprise industry attributes (e.g., "high-end equipment manufacturing"), matching policy names (e.g., "provincial science and technology enterprise subsidy policy"), highly matched project categories (e.g., "R&D platform construction subsidy"), unmet policy conditions (e.g., "R&D investment ratio 13% < 15%)", and industry development trends (e.g., "R&D investment requirements for high-end equipment manufacturing will be increased to 18% by 2025"). This ensures the large model obtains comprehensive analytical basis. Next, the analysis data is prepared by selecting any set of matching results and extracting information such as the comprehensive matching score, core matching conditions, and unmatched conditions. This information is then combined with key fragments of the corresponding original policy text and core indicators of the enterprise data profile to form a prompt context. Then, the system calls a large-scale model for analysis. Through the API interface, it submits the prompt context and preset template to the large-scale model, requesting it to perform policy clause interpretation, enterprise adaptation advantages and disadvantages analysis, and industry trend correlation analysis. The system then generates analysis results including application suggestions, cultivation goals, and implementation steps. Finally, the system processes the analysis results, receiving the text analysis content returned by the large-scale model, formatting the content, adding elements such as tables of contents and charts, and generating a downloadable PDF or Word format policy cultivation planning report. Simultaneously, it records user operation data on the report, such as "viewing the cultivation suggestion chapter" and "downloading the report." Based on this data, it adjusts the parameter weights of the matching model or optimizes the large-scale model's prompt word template to improve the accuracy and practicality of subsequent analyses.

[0046] The design includes prompt templates containing information such as the company's industry attributes and unmet policy conditions, providing ample analytical basis for the large model and avoiding biased analysis due to insufficient information. The large model not only interprets policy requirements and analyzes the company's suitability, but also generates cultivation suggestions based on industry trends, such as "It is recommended to increase the R&D investment ratio from 13% to 15% in the next year to meet policy requirements," providing companies with complete guidance from short-term application to long-term development. It supports formatted output of planning reports, which can be exported to common formats such as PDF and Word, facilitating company archiving and internal circulation. At the same time, it records user operation data to optimize the model, making subsequent analysis more in line with actual needs and continuously improving report quality.

[0047] Example 2

[0048] Please refer to Figure 2In this embodiment of the invention, an apparatus for intelligent policy-enterprise matching is also provided, comprising: a policy processing module 701, used to acquire the original policy text and perform multi-level granular decomposition and structural transformation on the original policy text to generate machine-readable and computable structured policy data; an enterprise data management module 702, used to collect and integrate enterprise data from multiple sources to construct a multi-dimensional enterprise data profile containing quantitative indicators; an intelligent matching module 703, used to automatically compare and match the structured policy data with the multi-dimensional enterprise data profile, and output a matching result including matching degree and policy suitability; and a large model analysis module 704, used to call a large model for in-depth analysis and reasoning based on the matching result to generate a customized policy cultivation planning report.

[0049] In this embodiment of the invention, the policy processing module 701 further includes: a collection unit 7011, used to collect and record basic attribute information of the policy, the basic attribute information including at least the policy name, issuing department, application time window, subsidy amount range, and applicable regional restrictions; a decomposition unit 7012, used to decompose the overall policy into several independent application project categories, and create a category description for each project category containing project background, purpose, and core application guidance; an extraction unit 7013, used to extract specific application conditions, support strength, and application cycle key elements from the corresponding text content for each application project category, forming a granular attribute set; and a conversion unit 7014, used to convert the granular attribute set described in the text into structured policy data with parameter codes, parameter names, logical comparison methods, and specific comparison values.

[0050] In this embodiment of the invention, the enterprise data management module 702 further includes: an acquisition unit 7021, used to acquire enterprise business registration information, financial data, intellectual property information, R&D investment data, and project establishment information through a combination of automatic import via interface and manual input; a preprocessing unit 7022, used to clean, align, and standardize the format of data from different sources to generate a standardized dataset; and a construction unit 7023, used to construct a multi-dimensional enterprise data profile containing quantitative evaluation indicators based on the standardized dataset, wherein the quantitative evaluation indicators are used at least for direct comparison with structured policy data.

[0051] In this embodiment of the invention, the intelligent matching module 703 further includes: a first matching unit 7031, configured to compare one or more target structured policy data with multiple multi-dimensional enterprise data profiles, calculate the matching degree between each enterprise and the target structured policy data, and generate a first matching result set with enterprises as entities; a second matching unit 7032, configured to compare one target enterprise's multi-dimensional enterprise data profile with multiple structured policy data in a policy database, calculate the matching degree between the target enterprise and each structured policy data, and generate a second matching result set with the structured policy data as entities; and a sorting unit 7033, configured to integrate and sort the entities in the first matching result set or the second matching result set according to the corresponding matching degree and predefined rules, and output the corresponding matching result list.

[0052] In this embodiment of the invention, the large model analysis module 704 further includes: a fusion unit 7041, used to select any of the first matching result set or the second matching result set, and combine it with the corresponding original policy text, the multi-dimensional enterprise data profile, and matching details to form a prompt context; a parsing unit 7042, used to submit the prompt context to the large model through an API interface, and request the large model to perform policy interpretation, enterprise status analysis, and strategic planning reasoning, and generate analysis results; and a generation unit 7043, used to receive the analysis results returned by the large model, and format the analysis results to generate a customized policy cultivation planning report.

[0053] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0054] Example 3

[0055] Corresponding to the above method embodiments, this disclosure also provides a device for policy-enterprise intelligent matching. The device for policy-enterprise intelligent matching described below and the method for policy-enterprise intelligent matching described above can be referred to in correspondence with each other.

[0056] Figure 3 This is a block diagram illustrating a device 800 for policy-enterprise intelligent matching according to an exemplary embodiment. Figure 3 As shown, the device 800 for policy-enterprise intelligent matching may include a processor 801 and a memory 802. The device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0057] The processor 801 controls the overall operation of the device 800 for policy-enterprise intelligent matching to complete all or part of the steps in the aforementioned method for policy-enterprise intelligent matching. The memory 802 stores various types of data to support the operation of the device 800 for policy-enterprise intelligent matching. This data may include, for example, instructions for any application or method operating on the device 800 for policy-enterprise intelligent matching, as well as application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the device 800 for policy-enterprise intelligent matching and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0058] In an exemplary embodiment, the device 800 for policy-enterprise intelligent matching may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method for policy-enterprise intelligent matching.

[0059] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the method for policy-enterprise intelligent matching described above. For example, the computer-readable storage medium may be the memory 802 including the program instructions described above, which may be executed by the processor 801 of the device 800 for policy-enterprise intelligent matching to perform the method for policy-enterprise intelligent matching described above.

[0060] Example 4

[0061] Corresponding to the above method embodiments, this disclosure also provides a readable storage medium, which can be referred to in conjunction with the above-described method for policy-enterprise intelligent matching.

[0062] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for policy-enterprise intelligent matching described in the above method embodiments.

[0063] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0064] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention.

[0065] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not describe the various possible combinations separately.

[0066] Furthermore, various different implementations of the present invention can be combined arbitrarily, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed in the present invention.

Claims

1. A method for intelligent matching of policy and enterprise, characterized in that, include: The original policy text is obtained and then subjected to multi-level granular decomposition and structural transformation to generate machine-readable and computable structured policy data. Collect and integrate enterprise data from multiple sources to construct a multi-dimensional enterprise data profile that includes quantitative indicators; The structured policy data is automatically compared and matched with the multi-dimensional enterprise data profile, and the matching result, including the matching degree and the applicable policy, is output. Based on the matching results, a large model is invoked for in-depth analysis and reasoning to generate a customized policy development planning report.

2. The method for intelligent matching of policy and enterprise according to claim 1, characterized in that, The process of obtaining the original policy text and performing multi-level granular decomposition and structural transformation on the original policy text to generate machine-readable and computable structured policy data specifically includes: Collect and record basic attribute information of the policy, including at least the policy name, issuing department, application time window, subsidy amount range, and applicable regional restrictions; The overall policy is broken down into several independent application categories, and a category description is created for each category, including the project background, purpose, and core application guidelines. For each of the aforementioned project categories, key elements such as specific application conditions, support levels, and application cycles are extracted from the corresponding text content to form a granular attribute set. The granular attribute set described in the text is converted into structured policy data with parameter encoding, parameter name, logical comparison method, and specific comparison value.

3. The method for intelligent matching of policy and enterprise according to claim 2, characterized in that, Before converting the granular attribute set described in the text into structured policy data with parameter encoding, parameter names, logical comparison methods, and specific comparison values, the method further includes: The extracted granular attribute set is cleaned and standardized. The standardization process includes format unification using regular expressions and semantic disambiguation and normalization based on knowledge graphs, and then cleaned policy elements are generated. A unique feature label is assigned to the cleaned policy element, the feature label containing a label code and a label name.

4. The method for intelligent matching of policy and enterprise according to claim 1, characterized in that, The process of collecting and integrating enterprise data from multiple sources to construct a multi-dimensional enterprise data profile including quantitative indicators includes: By combining automatic import via API with manual entry, we can obtain enterprises' business registration information, financial data, intellectual property information, R&D investment data, and project initiation information. Data from different sources is cleaned, aligned, and standardized to generate a standardized dataset. Based on the standardized dataset, a multi-dimensional enterprise data profile containing quantitative evaluation indicators is constructed. These quantitative evaluation indicators are used at least for direct comparison with structured policy data.

5. The method for intelligent matching of policy and enterprise according to claim 1, characterized in that, The automated comparison and matching calculation of the structured policy data and the multi-dimensional enterprise data profile, outputting a matching result including matching degree and policy suitability, includes: Based on the structured policy data of one or more targets, compare it with multiple multi-dimensional enterprise data profiles, calculate the matching degree between each enterprise and the structured policy data of the targets, and generate a first matching result set with enterprises as entities; Based on the multi-dimensional enterprise data profile of a target enterprise, the data is compared with multiple structured policy data in the policy database to calculate the matching degree between the target enterprise and each structured policy data, and a second matching result set with the structured policy data as entities is generated. The entities in the first matching result set or the second matching result set are integrated and sorted according to the corresponding matching degree and predefined rules, and the corresponding matching result list is output.

6. A method for intelligent matching of policy and enterprise according to claim 5, characterized in that, Based on the matching results, a large model is invoked for in-depth analysis and reasoning to generate a customized policy development planning report, including: Select either the first matching result set or the second matching result set, and combine them with the corresponding original policy text, the multi-dimensional enterprise data profile, and the matching details to form a prompt context; The prompt context is submitted to the large model via the API interface, and the large model is requested to perform policy interpretation, enterprise status analysis and strategic planning reasoning, and generate analysis results. Receive the analysis results returned by the large model, and format the analysis results to generate a customized policy cultivation planning report.

7. A device for intelligent matching of policy and enterprise, characterized in that, include: The policy processing module is used to obtain the original policy text and perform multi-level granular decomposition and structural transformation on the original policy text to generate machine-readable and computable structured policy data; The enterprise data management module is used to collect and integrate enterprise data from multiple sources to build a multi-dimensional enterprise data profile that includes quantitative indicators. The intelligent matching module is used to automatically compare and match the structured policy data with the multi-dimensional enterprise data profile, and output the matching result including the matching degree and the applicable policy. The large model analysis module is used to perform in-depth analysis and reasoning based on the matching results, and generate customized policy cultivation planning reports.

8. The device for policy-enterprise intelligent matching according to claim 7, characterized in that, The policy processing module also includes: The data collection unit is used to collect and record the basic attribute information of the policy, which includes at least the policy name, issuing department, application time window, subsidy amount range, and applicable regional restrictions. The decomposition unit is used to break down the overall policy into several independent application project categories, and to create a category description for each project category that includes the project background, purpose and core application guidelines; The extraction unit is used to extract key elements such as specific application conditions, support levels, and application cycles from the corresponding text content for each application project category, forming a granular attribute set. The conversion unit is used to convert the granular attribute set described in the text into structured policy data with parameter encoding, parameter name, logical comparison method and specific comparison value.

9. The device for policy-enterprise intelligent matching according to claim 7, characterized in that, The enterprise data management module also includes: The acquisition unit is used to acquire enterprise business registration information, financial data, intellectual property information, R&D investment data and project establishment information through a combination of automatic import via interface and manual entry. The preprocessing unit is used to clean, align, and standardize data from different sources to generate a standardized dataset. The construction unit is used to construct a multi-dimensional enterprise data profile containing quantitative evaluation indicators based on the standardized dataset. The quantitative evaluation indicators are used at least for direct comparison with structured policy data.

10. The device for policy-enterprise intelligent matching according to claim 7, characterized in that, The intelligent matching module also includes: The first matching unit is used to compare the structured policy data of one or more targets with the multi-dimensional enterprise data profiles, calculate the matching degree between each enterprise and the structured policy data of the targets, and generate a first matching result set with enterprises as entities. The second matching unit is used to compare the multi-dimensional enterprise data profile of a target enterprise with multiple structured policy data in the policy database, calculate the matching degree between the target enterprise and each structured policy data, and generate a second matching result set with the structured policy data as entities; The sorting unit is used to integrate and sort the entities in the first matching result set or the second matching result set according to the corresponding matching degree and predefined rules, and output the corresponding matching result list.

Images