Artificial intelligence-based enterprise policy customization report generation method

Abstract

This invention relates to the field of artificial intelligence technology and discloses a method for generating customized enterprise policy reports based on artificial intelligence. The method includes S1: collecting policy data and enterprise data, fusing and enhancing them to generate a policy-enterprise fusion data pool; S2: based on the policy-enterprise fusion data pool, extracting policy knowledge elements and constructing enterprise profiles to generate an evolvable policy-enterprise related knowledge network and its query interface; by constructing an evolvable policy-enterprise related knowledge network and extracting related subnetworks from the network for strategic information filtering, the dynamics and accuracy of the policy-enterprise data association are ensured. At the same time, an evaluation vector is generated based on a multi-dimensional quantitative evaluation model to reduce policy matching errors and ensure the accuracy of policy matching analysis.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to a method for generating customized enterprise policy reports based on artificial intelligence. Background Technology

[0002] When matching with published corporate policy documents, companies need to meet a series of diverse and detailed matching conditions. These conditions typically involve multiple aspects such as the company's years of establishment, operating revenue, R&D investment, and employee composition, aiming to ensure that policy resources can effectively support eligible companies.

[0003] Currently, due to the dispersed and dynamically updated nature of corporate policy information, existing methods cannot construct a dynamically evolving knowledge network to achieve accurate correlation and quantitative assessment between policies and enterprises when generating customized corporate policy reports. If policy conditions change or corporate status shifts, it may cause large policy matching errors. Furthermore, the reports lack decision-making simulation and interactive exploration capabilities, which cannot guarantee the accuracy and foresight of the generated reports.

[0004] Therefore, we propose an AI-based method for generating customized enterprise policy reports to address the aforementioned issues. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an artificial intelligence-based method for generating customized enterprise policy reports. This method solves the problems mentioned in the background section, such as large policy matching errors and a lack of decision-making simulation and interactive exploration capabilities in the reports, which fail to guarantee the accuracy and foresight of the generated reports.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for generating customized enterprise policy reports based on artificial intelligence, comprising the following steps: S1. Collect policy data and enterprise data, integrate and enhance them to generate a policy-enterprise integrated data pool; S2. Based on the policy-enterprise integrated data pool, extract policy knowledge elements and construct enterprise profiles to generate an evolvable policy-enterprise related knowledge network and its query interface; S3. Receive a report generation request containing the target enterprise identifier and customized preference parameters, and extract related subnetworks from the policy-enterprise association knowledge network based on the report generation request and perform strategic information filtering; S4. The filtered sub-network information is structured, sliced, and contextualized to generate a modular report content unit set; S5. Quantitatively evaluate the modular report content unit set based on the preset evaluation model, and simulate the development path under different policy adoption strategies based on the preset inference algorithm to generate decision inference data package; S6. Assemble the modular report content unit set with the decision inference data package, and generate a preliminary customized report based on the report template selected from the preset template library; S7. Construct an interactive report exploration engine, which encapsulates the preliminary customized report, the modular report content unit set, the decision inference data package, and the query interface of the policy-enterprise related knowledge network to generate an interactive enterprise policy customized report; S8. Collect and analyze the interactive behavior data generated by users during the use of the interactive enterprise policy customization report, generate a user feedback profile, and optimize the parameters, recommendation logic and inference model in the report generation system based on this user feedback profile.

[0007] Preferably, the generation of the policy-enterprise integrated data pool in S1 includes the following steps: S11. Use a configurable network collector to collect policy data from the policy release platforms of multi-level government departments to obtain the original policy dataset, including the original policy text, attachments, and metadata. S12. Obtain enterprise form data through the enterprise self-reporting channel and obtain the enterprise's public information through the artificial intelligence information capture channel to form the enterprise's original data; S13. Perform format unification, field extraction, structure parsing and semantic segmentation on the original policy dataset to generate a structured policy knowledge base; S14. Perform entity alignment, information completion and feature extraction processing on the original enterprise data to generate enterprise panoramic profile data; S15. The structured policy knowledge base is associated and stored with the enterprise panoramic profile data to construct the policy-enterprise integrated data pool.

[0008] Preferably, the policy knowledge element extraction and enterprise profile construction in S2 include the following steps: S21. Extract knowledge elements including policy subjects, applicable objects, application conditions, and support methods from the structured policy knowledge base, and assign semantic identifiers to each knowledge element; S22. Based on the enterprise panoramic profile data, construct a multi-layered enterprise profile that includes static attributes, operational strength, and innovation potential features; S23. Using the policy knowledge elements and the enterprise profile features as nodes, and semantic association and condition satisfaction relationships as edges, initialize the policy-enterprise association knowledge graph; S24. Configure timeliness attributes and weight decay functions for nodes and edges in the knowledge graph, and set triggers; S25. When the policy-enterprise integrated data pool detects data changes, it triggers a local incremental update and weight recalculation of the policy-enterprise related knowledge graph to form an evolvable policy-enterprise related knowledge network.

[0009] Preferably, the step S3, which involves extracting related sub-networks and performing strategy-based information filtering, includes the following steps: S31. Parse the report generation request to obtain the target enterprise identifier and customized preference parameters; S32. Starting from the corresponding node of the target enterprise in the policy-enterprise association knowledge network, perform multi-hop association subgraph extraction to form an initial association subnetwork; S33. Filter and prune the initial associated subnetwork according to the strategic filtering conditions in the customized preference parameters; S34. Supplement the selected subnetworks with relevant policy interpretation links and enterprise case summary information to generate rich information subnetworks for report customization.

[0010] Preferably, the structured slicing and contextualized encapsulation in S4 includes the following steps: S41. Perform multi-dimensional slicing on the information-rich sub-network, including policy type, matching degree, and urgency degree. S42. Encapsulate a standard contextual paragraph framework for each sliced ​​information module, including core points, matching point analysis, and gap analysis. S43. Adjust the module content according to the level of detail parameter in the customized preference parameters; S44. Sort the processed modules according to logical priority to generate the modular report content unit set.

[0011] Preferably, the generation of the decision deduction data package in S5 includes the following steps: S51. Call the multi-dimensional quantitative evaluation model that includes matching degree evaluation, benefit potential evaluation, and application complexity evaluation to generate a multi-dimensional evaluation vector for each policy matching item in the modular report content unit set; S52. Based on historical declaration data and machine learning models, construct a decision-making inference algorithm, taking the current status of the enterprise and policy matching items as inputs, and simulate the impact path of different declaration combination strategies on the enterprise's indicators in the future time period. S53. Run the decision-making simulation algorithm for key policy matching items to generate simulation results including application sequence recommendations, resource input and expected return curves; S54. Package the multidimensional evaluation vector and the inference results to form a decision inference data package associated with the corresponding modular report content unit.

[0012] Preferably, generating a preliminary customized report in step S6 includes the following steps: S61. Provides a variety of preset report templates, designed according to the audience role and purpose; S62. Select the appropriate report template based on the customized preference parameters; S63. Based on the framework of the selected report template, automatically fill each unit in the modular report content unit set into the corresponding position of the template; S64. Embed the evaluation conclusions and simulation charts in the decision simulation data package into the main body and appendix of the report in a visual form; S65. Perform coherence verification on the assembled content and generate a preliminary customized report document.

[0013] Preferably, generating an interactive, customized enterprise policy report in step S7 includes the following steps: S71. Embed client controls for the interactive report exploration engine in the initial customized report document; S72. The client control has a dynamic filtering function, allowing report readers to adjust the filtering conditions on the front end, and the engine back end to re-filter information and partially refresh the report content based on the rich information sub-network. S73. The client control provides drill-down exploration functionality, providing an entry point for viewing detailed analysis processes and data sources for the conclusions and data in the report; S74. The client control also provides a hypothesis simulation function, allowing users to manually adjust enterprise assumptions and select different policy application combination strategies, and calculate and display the changes in the adjusted inference path in real time based on the decision inference algorithm. S75. Encapsulate the preliminary customized report document, the client control, and the query interface of the policy-enterprise related knowledge network to generate an interactive enterprise policy customized report.

[0014] Preferably, generating a user feedback profile in step S8 includes the following steps: S81. Collect user interaction behavior data generated during the use of the interactive enterprise policy customization report, analyze user focus and decision-making preferences based on the interaction behavior data, and generate user feedback profiles. S82. Based on the user feedback profile, optimize the default values ​​of the customized preference parameters, the recommendation logic of the report template, and the parameters of the decision inference algorithm.

[0015] Preferably, the method further includes: S91. The generated interactive customized enterprise policy report is pushed to the target user terminal through the delivery channel, and a binding relationship between the report version and the user identity is established; S92. Monitoring and reporting related policy dynamics, enterprise data changes, and user feedback data constitute the report effectiveness tracking dataset; S93. Based on the report effect tracking dataset, analyze the correlation between policy recommendation adoption, user focus evolution trends, changes in enterprise status, and policy matching degree; S94. Based on the analysis results, the evolution rules of the policy-enterprise related knowledge network, the recommendation logic of customized preference parameters, and the decision inference algorithm are dynamically adjusted, and the adjusted strategy is applied to the subsequent report generation.

[0016] Compared with existing technologies, this invention provides a method for generating customized enterprise policy reports based on artificial intelligence, which has the following beneficial effects: 1. In this invention, an evolvable policy-enterprise related knowledge network is constructed, and related subnetworks are extracted from this network for strategic information filtering, ensuring the dynamism and accuracy of the policy-enterprise data association. At the same time, an evaluation vector is generated based on a multi-dimensional quantitative evaluation model to reduce policy matching errors and ensure the accuracy of policy matching analysis.

[0017] 2. In this invention, by simulating development paths under different policy adoption strategies based on decision inference algorithms and generating decision inference data packages, the report is provided with inference results including application sequence recommendations, resource input and expected return curves, so that the generated report has decision support depth and ensures the report's forward-looking nature and strategic reference value.

[0018] 3. In this invention, an interactive report exploration engine is constructed to provide dynamic filtering, drill-down exploration, and hypothesis simulation functions. The initial customized report, modular report content unit set, decision inference data package, and knowledge network query interface are encapsulated to generate an interactive enterprise policy customized report. At the same time, user feedback profiles are generated based on collected user interaction behavior data to optimize system parameters and models, ensuring the intelligence, adaptability, and continuous optimization capability of the reporting system. Attached Figure Description

[0019] Figure 1 This is a flowchart of the AI-based enterprise policy customization report generation method of the present invention; Figure 2 This is a flowchart for generating an interactive, customized enterprise policy report in S7 of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] For specific implementation examples, please refer to: Figure 1-2 The method for generating customized enterprise policy reports based on artificial intelligence includes the following steps: S1. Collect policy data and enterprise data, integrate and enhance them to generate a policy-enterprise integrated data pool; S2. Based on the policy-enterprise integrated data pool, policy knowledge elements are extracted and enterprise profiles are constructed to generate an evolvable policy-enterprise related knowledge network and its query interface. S3. Receive a report generation request containing the target enterprise identifier and customized preference parameters, and extract related subnetworks from the policy-enterprise relationship knowledge network based on the report generation request and perform strategic information filtering; S4. The filtered sub-network information is structured, sliced, and contextualized to generate a modular report content unit set; S5. Based on the preset evaluation model, the modular report content unit set is quantitatively evaluated, and based on the preset inference algorithm, the development path under different policy adoption strategies is simulated to generate decision inference data package. S6. Assemble the modular report content unit set with the decision inference data package, and generate a preliminary customized report based on the report template selected from the preset template library; S7. Construct an interactive report exploration engine, which encapsulates the preliminary customized report, the modular report content unit set, the decision inference data package, and the query interface of the policy-enterprise related knowledge network to generate an interactive enterprise policy customized report; S8. Collect and analyze user interaction data generated during the use of interactive enterprise policy customization reports, generate user feedback profiles, and optimize the parameters, recommendation logic, and inference models in the report generation system based on these user feedback profiles.

[0022] The generation of the policy-enterprise converged data pool in S1 includes the following steps: S11. Use a configurable network collector to collect policy data from the policy release platforms of multi-level government departments to obtain the original policy dataset, including the original policy text, attachments, and metadata. S12. Obtain enterprise form data through the enterprise self-reporting channel and obtain the enterprise's public information through the artificial intelligence information capture channel to form the enterprise's original data; S13. Perform format unification, field extraction, structure parsing, and semantic segmentation on the original policy dataset to generate a structured policy knowledge base; S14. Perform entity alignment, information completion and feature extraction processing on the enterprise's original data to generate enterprise panoramic profile data; S15. Link and store the structured policy knowledge base with the enterprise panoramic profile data to build a policy-enterprise integrated data pool.

[0023] The extraction of policy knowledge elements and the construction of enterprise profiles in S2 include the following steps: S21. Extract knowledge elements from the structured policy knowledge base, including policy subjects, applicable objects, application conditions, and support methods, and assign semantic identifiers to each knowledge element; Specifically, knowledge element extraction is implemented using deep learning-based Natural Language Processing (NLP) technology: First, the pre-trained language model BERT based on the Transformer architecture was fine-tuned to complete the Named Entity Recognition (NER) task, extracting entities of the aforementioned knowledge element types from policy texts. Second, the Relation Extraction (RE) model was used to identify and structure the semantic relationships between entities such as "policy subject-applicable object" and "application conditions-support methods." The training corpus for the NER and RE models came from a dedicated dataset constructed by manually annotating historical policy documents. The annotation specifications clearly defined the boundaries and relationship definitions of each knowledge element type. The coding specifications for "semantic identifiers" were generated based on a top-down "enterprise policy domain ontology," which defines the conceptual hierarchy, attributes, and relationships of policy elements. Each extracted knowledge element was assigned a standard coded URI that conforms to the ontology specifications, thereby achieving unified identification and association at the semantic level. S22. Based on the enterprise panoramic profile data, construct a multi-layer enterprise profile that includes static attributes, operational strength, and innovation potential characteristics; S23. Using policy knowledge elements and enterprise profile features as nodes and semantic association and condition satisfaction relationships as edges, initialize the policy-enterprise association knowledge graph; S24. Configure timeliness attributes and weight decay functions for nodes and edges in the knowledge graph, and set triggers; Dynamic attributes are attached to each connection edge in the knowledge graph, and a weight update rule is defined. A weight model based on time decay is adopted, and its calculation formula is as follows: ; in For the current time The effective weight of the associated edge, This is the initial association strength weight for this edge. Create a timestamp for this edge and the timestamp of the last time it was refreshed by a related event. This is the half-life parameter corresponding to this edge type; S25. When the policy-enterprise integrated data pool detects data changes, it triggers a local incremental update and weight recalculation of the policy-enterprise related knowledge graph, forming an evolvable policy-enterprise related knowledge network.

[0024] Extracting related subnetworks and performing policy-based information filtering in S3 includes the following steps: S31. Parse the report generation request to obtain the target enterprise identifier and customized preference parameters; S32. Starting from the corresponding node of the target enterprise in the policy-enterprise related knowledge network, perform multi-hop related subgraph extraction to form an initial related subnetwork; In the specific implementation of multi-hop related subgraph extraction, the corresponding node of the target enterprise in the knowledge network is used as the query starting point; the number of related hops and the set of related edge types are defined, and a rule-based graph traversal algorithm is adopted. Starting from the query starting point, a breadth-first search is performed along the edges that conform to the set of related edge types until the preset number of hops is reached; during the traversal process, all visited nodes and edges are recorded to form the initial related subgraph. S33. Filter and prune the initial associated subnetwork based on the strategic screening conditions in the customized preference parameters; In the specific implementation of filtering and pruning, the initial associated subgraph is filtered according to the strategic screening conditions defined in the customized preference parameters. First, based on the report's focus, weight thresholds are set for different types of associated edges; when the focus is on "qualification compliance," the weight threshold for "condition matching" type edges is increased; when the focus is on "profit forecast," the weight threshold for "financial support" type edges is increased. Then, all edges in the initial associated subgraph are traversed, and edges with weights lower than their corresponding type thresholds are removed. Next, isolated nodes resulting from edge removal are pruned and deleted. Finally, a filtered and streamlined associated subnetwork highly relevant to the customized needs is obtained. S34. Supplement the selected subnetworks with relevant policy interpretation links and enterprise case summary information to generate rich information subnetworks for report customization.

[0025] Structured slicing and contextualized encapsulation in S4 include the following steps: S41. Multi-dimensional slicing of the information-rich sub-network, including policy type, matching degree, and urgency; S42. Encapsulate a standard contextual paragraph framework for each sliced ​​information module, including core points, matching point analysis, and gap analysis. S43. Adjust the module content according to the level of detail parameter in the customized preference parameters; In the specific implementation of scaling, the level of detail parameter in the customized preference parameters is used. Linear interpolation is used to scale the amount of information within each modular report content unit; a core information set is predefined for each content unit. and extended information set Set a content display scaling factor. Its relationship with the level of detail is as follows: For extended information within a unit, it is sorted according to its preset importance weight, and only the top-ranked information is displayed. The information of the item, among which , To expand the total amount of information, this method enables a smooth scaling of report content from briefings containing only key points to detailed reports containing all the essential information. S44. Sort the processed modules according to logical priority to generate a modular report content unit set.

[0026] Generating decision simulation data packages in S5 includes the following steps: S51. Call the multi-dimensional quantitative assessment model that includes matching degree assessment, benefit potential assessment, and application complexity assessment to generate a multi-dimensional assessment vector for each policy matching item in the modular report content unit set; The multi-dimensional quantitative assessment model is a comprehensive assessment function that calculates a multi-dimensional assessment vector for each policy matching item. The comprehensive assessment vector is obtained by weighted summation of the scores from each dimension. Its core calculation formula is as follows: ; Among them, matching degree The calculation formula is: ; in This is the comprehensive evaluation vector for policy matching items. To score the matching degree, The text semantic matching score is calculated based on a deep semantic similarity model. This represents the rate at which enterprises actually meet the explicit reporting requirements of the policy, with a value ranging from 0 to 1. The weighting coefficients and , To score the potential for returns, policies are analyzed. The level of support and tax incentives, based on historical data, are used to quantify and estimate the amount of support provided and the tax benefits offered. To obtain a score for the complexity of the application, according to the policy The calculation is based on a combination of factors, including the number of application materials, the number of process steps, and the historical average approval time. S52. Based on historical declaration data and machine learning models, construct a decision-making inference algorithm, taking the current status of the enterprise and policy matching items as inputs, and simulate the impact path of different declaration combination strategies on the enterprise's indicators in the future time period. A decision-making inference algorithm is constructed based on historical declaration data and machine learning models. Specifically, a historical decision state transition dataset is first built. A state transition prediction model is trained using this dataset. The core of this model's derivation is the state transition equation: ; in For the time predicted by the model The enterprise state vector, Indicates a point in time The enterprise state vector, For businesses in time State feature vector, For businesses in time The policy application process adopted In time The policies declared and evaluated eigenvectors, For parameters State transition prediction model; S53. Run the decision-making simulation algorithm for key policy matching items to generate simulation results including application sequence recommendations, resource input and expected return curves; S54. Package the multidimensional evaluation vectors and inference results to form a decision inference data package associated with the corresponding modular report content unit.

[0027] Generating a preliminary customized report in S6 includes the following steps: S61. Provides a variety of preset report templates, designed according to the audience role and purpose; S62. Select the appropriate report template based on the customized preference parameters; S63. Based on the framework of the selected report template, automatically fill each unit in the modular report content unit set into the corresponding position in the template; S64. Embed the evaluation conclusions and simulation charts from the decision simulation data package into the main body and appendix of the report in a visual format; S65. Perform coherence verification on the assembled content and generate a preliminary customized report document.

[0028] Generating interactive, customized enterprise policy reports in S7 includes the following steps: S71. Embed client-side controls for the interactive report exploration engine in the initial customized report document; S72. The client-side control has a dynamic filtering function, which allows report readers to adjust the filtering conditions on the front end. The engine back end re-filters information based on the rich information sub-network and partially refreshes the report content. S73. The client-side control provides drill-down exploration functionality, offering an entry point for viewing detailed analysis processes and data sources for conclusions and data in the report; S74. The client-side control also provides a hypothesis simulation function, allowing users to manually adjust the enterprise's assumptions and select different policy application combination strategies. Based on the decision-making inference algorithm, it calculates and displays the changes in the adjusted inference path in real time. In practical implementation, within the hypothesis simulation interface of the interactive report exploration engine, when users manually adjust the enterprise's assumptions and select different policy application combination strategies through interactive controls, the engine triggers real-time calculations. The system takes the current target enterprise's baseline state and the user's adjusted state and new strategy as input, calls the pre-trained decision-making inference algorithm model, and performs multi-step iterative inference. The inference process is as follows: ; in Let be the system state vector at step k+1. Let k be the system state vector at step k. For the new strategy in the first The movement of the step, The corresponding policies; the simulation results are rendered and updated in real time in the form of dynamic charts in the user interface, and compared with the baseline simulation path to intuitively reflect the potential changes brought about by user adjustments; S75. Encapsulate the preliminary customized report document, client controls, and query interface of the policy-enterprise related knowledge network to generate an interactive enterprise policy customized report.

[0029] Generating user feedback profiles in S8 includes the following steps: S81. Collect user interaction behavior data generated during the use of interactive enterprise policy customization reports, analyze user concerns and decision-making preferences based on the interaction behavior data, and generate user feedback profiles. S82. Based on user feedback profiles, optimize the default values ​​of customized preference parameters, the recommendation logic of report templates, and the parameters of decision inference algorithms; The objective function for optimizing the parameters of the inference algorithm model is: ; in The optimization objective is to find the loss function that makes the loss function... Minimal model parameters , For loss function, This is a set of state transition samples constructed from user feedback data. A training sample represents the initial state, action, policy, and observed subsequent actual state, respectively. The square of the L2 norm, For regularization terms, is the regularization coefficient.

[0030] The method also includes: S91. Push the generated interactive customized enterprise policy report to the target user through the delivery channel, and establish a binding relationship between the report version and the user identity; S92. Monitoring and reporting related policy dynamics, enterprise data changes, and user feedback data constitute the report effectiveness tracking dataset; S93. Based on the report effect tracking dataset, analyze the correlation between policy recommendation adoption, user focus evolution trends, changes in enterprise status, and policy matching degree; S94. Based on the analysis results, the evolution rules of the policy-enterprise related knowledge network, the recommendation logic of customized preference parameters, and the decision inference algorithm are dynamically adjusted, and the adjusted strategy is applied to the subsequent report generation.

[0031] The steps of this method are as follows: First, raw policy datasets are obtained by targeting policy release platforms of multi-level government departments through a configurable network collector. Then, raw enterprise data is obtained through enterprise self-reporting channels and artificial intelligence information capture channels. The two are then merged and enhanced to generate a policy-enterprise integrated data pool. Then, based on this data pool, policy knowledge elements are extracted and enterprise profiles are constructed. An evolvable policy-enterprise related knowledge network and its query interface are built with policy knowledge elements and profile features as nodes. Next, a report generation request containing the target enterprise identifier and customized preference parameters is received, and related subnetworks are extracted from the knowledge network and strategic information filtering is performed to generate information-rich subnetworks. Subsequently, the information-rich subnetwork is subjected to multi-dimensional structured slicing and contextualized encapsulation to generate a set of modular report content units; then, a multi-dimensional quantitative evaluation model is invoked to quantitatively evaluate the set of modular report content units, and a decision inference data package is generated by simulating development paths under different policy adoption strategies based on a decision inference algorithm. Then, the modular report content unit set is assembled with the decision inference data package, and a preliminary customized report is generated based on the report template selected from the preset template library; then, an interactive report exploration engine is built to encapsulate the preliminary customized report, the modular report content unit set, the decision inference data package, and the query interface of the policy-enterprise related knowledge network to generate an interactive enterprise policy customized report; Finally, user feedback profiles are generated by collecting user interaction data during the use of interactive reports, and the parameters, recommendation logic, and inference models in the report generation system are optimized based on these user feedback profiles.

[0032] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0033] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for generating customized enterprise policy reports based on artificial intelligence, characterized by: Includes the following steps: S1. Collect policy data and enterprise data, integrate and enhance them to generate a policy-enterprise integrated data pool; S2. Based on the policy-enterprise integrated data pool, extract policy knowledge elements and construct enterprise profiles to generate an evolvable policy-enterprise related knowledge network and its query interface; S3. Receive a report generation request containing the target enterprise identifier and customized preference parameters, and extract related subnetworks from the policy-enterprise association knowledge network based on the report generation request and perform strategic information filtering; S4. The filtered sub-network information is structured, sliced, and contextualized to generate a modular report content unit set; S5. Quantitatively evaluate the modular report content unit set based on the preset evaluation model, and simulate the development path under different policy adoption strategies based on the preset inference algorithm to generate decision inference data package; S6. Assemble the modular report content unit set with the decision inference data package, and generate a preliminary customized report based on the report template selected from the preset template library; S7. Construct an interactive report exploration engine, which encapsulates the preliminary customized report, the modular report content unit set, the decision inference data package, and the query interface of the policy-enterprise related knowledge network to generate an interactive enterprise policy customized report; S8. Collect and analyze the interactive behavior data generated by users during the use of the interactive enterprise policy customization report, generate a user feedback profile, and optimize the parameters, recommendation logic and inference model in the report generation system based on this user feedback profile.

2. The method for generating customized enterprise policy reports based on artificial intelligence according to claim 1, characterized in that: The generation of the policy-enterprise integrated data pool in S1 includes the following steps: S11. Use a configurable network collector to collect policy data from the policy release platforms of multi-level government departments to obtain the original policy dataset, including the original policy text, attachments, and metadata. S12. Obtain enterprise form data through the enterprise self-reporting channel and obtain the enterprise's public information through the artificial intelligence information capture channel to form the enterprise's original data; S13. Perform format unification, field extraction, structure parsing and semantic segmentation on the original policy dataset to generate a structured policy knowledge base; S14. Perform entity alignment, information completion and feature extraction processing on the original enterprise data to generate enterprise panoramic profile data; S15. The structured policy knowledge base is associated and stored with the enterprise panoramic profile data to construct the policy-enterprise integrated data pool.

3. The method for generating customized enterprise policy reports based on artificial intelligence according to claim 2, characterized in that: The policy knowledge element extraction and enterprise profile construction in S2 include the following steps: S21. Extract knowledge elements including policy subjects, applicable objects, application conditions, and support methods from the structured policy knowledge base, and assign semantic identifiers to each knowledge element; S22. Based on the enterprise panoramic profile data, construct a multi-layered enterprise profile that includes static attributes, operational strength, and innovation potential features; S23. Using the policy knowledge elements and the enterprise profile features as nodes, and semantic association and condition satisfaction relationships as edges, initialize the policy-enterprise association knowledge graph; S24. Configure timeliness attributes and weight decay functions for nodes and edges in the knowledge graph, and set triggers; S25. When the policy-enterprise integrated data pool detects data changes, it triggers a local incremental update and weight recalculation of the policy-enterprise related knowledge graph to form an evolvable policy-enterprise related knowledge network.

4. The method for generating customized enterprise policy reports based on artificial intelligence according to claim 3, characterized in that: The process of extracting related sub-networks and filtering information using a strategy in step S3 includes the following steps: S31. Parse the report generation request to obtain the target enterprise identifier and customized preference parameters; S32. Starting from the corresponding node of the target enterprise in the policy-enterprise association knowledge network, perform multi-hop association subgraph extraction to form an initial association subnetwork; S33. Filter and prune the initial associated subnetwork according to the strategic filtering conditions in the customized preference parameters; S34. Supplement the selected subnetworks with relevant policy interpretation links and enterprise case summary information to generate rich information subnetworks for report customization.

5. The method for generating customized enterprise policy reports based on artificial intelligence according to claim 4, characterized in that: The structured slicing and contextualized encapsulation in S4 include the following steps: S41. Perform multi-dimensional slicing on the information-rich sub-network, including policy type, matching degree, and urgency degree. S42. Encapsulate a standard contextual paragraph framework for each sliced ​​information module, including core points, matching point analysis, and gap analysis. S43. Adjust the module content according to the level of detail parameter in the customized preference parameters; S44. Sort the processed modules according to logical priority to generate the modular report content unit set.

6. The method for generating customized enterprise policy reports based on artificial intelligence according to claim 5, characterized in that: The generation of the decision inference data package in S5 includes the following steps: S51. Call the multi-dimensional quantitative evaluation model that includes matching degree evaluation, benefit potential evaluation, and application complexity evaluation to generate a multi-dimensional evaluation vector for each policy matching item in the modular report content unit set; S52. Based on historical declaration data and machine learning models, construct a decision-making inference algorithm, taking the current status of the enterprise and policy matching items as inputs, and simulate the impact path of different declaration combination strategies on the enterprise's indicators in the future time period. S53. Run the decision-making simulation algorithm for key policy matching items to generate simulation results including application sequence recommendations, resource input and expected return curves; S54. Package the multidimensional evaluation vector and the inference results to form a decision inference data package associated with the corresponding modular report content unit.

7. The method for generating customized enterprise policy reports based on artificial intelligence according to claim 6, characterized in that: The initial customized report is generated in S6, including the following steps: S61. Provides a variety of preset report templates, designed according to the audience role and purpose; S62. Select the appropriate report template based on the customized preference parameters; S63. Based on the framework of the selected report template, automatically fill each unit in the modular report content unit set into the corresponding position of the template; S64. Embed the evaluation conclusions and simulation charts in the decision simulation data package into the main body and appendix of the report in a visual form; S65. Perform coherence verification on the assembled content and generate a preliminary customized report document.

8. The method for generating customized enterprise policy reports based on artificial intelligence according to claim 7, characterized in that: The S7 process generates an interactive, customized enterprise policy report, including the following steps: S71. Embed client controls for the interactive report exploration engine in the initial customized report document; S72. The client control has a dynamic filtering function, allowing report readers to adjust the filtering conditions on the front end, and the engine back end to re-filter information and partially refresh the report content based on the rich information sub-network. S73. The client control provides drill-down exploration functionality, providing an entry point for viewing detailed analysis processes and data sources for the conclusions and data in the report; S74. The client control also provides a hypothesis simulation function, allowing users to manually adjust enterprise assumptions and select different policy application combination strategies, and calculate and display the changes in the adjusted inference path in real time based on the decision inference algorithm. S75. Encapsulate the preliminary customized report document, the client control, and the query interface of the policy-enterprise related knowledge network to generate an interactive enterprise policy customized report.

9. The method for generating customized enterprise policy reports based on artificial intelligence according to claim 8, characterized in that: The process of generating a user feedback profile in S8 includes the following steps: S81. Collect user interaction behavior data generated during the use of the interactive enterprise policy customization report, analyze user focus and decision-making preferences based on the interaction behavior data, and generate user feedback profiles. S82. Based on the user feedback profile, optimize the default values ​​of the customized preference parameters, the recommendation logic of the report template, and the parameters of the decision inference algorithm.

10. The method for generating customized enterprise policy reports based on artificial intelligence according to claim 1, characterized in that: The method further includes: S91. The generated interactive customized enterprise policy report is pushed to the target user terminal through the delivery channel, and a binding relationship between the report version and the user identity is established; S92. Monitoring and reporting related policy dynamics, enterprise data changes, and user feedback data constitute the report effectiveness tracking dataset; S93. Based on the report effect tracking dataset, analyze the correlation between policy recommendation adoption, user focus evolution trends, changes in enterprise status, and policy matching degree; S94. Based on the analysis results, the evolution rules of the policy-enterprise related knowledge network, the recommendation logic of customized preference parameters, and the decision inference algorithm are dynamically adjusted, and the adjusted strategy is applied to the subsequent report generation.

Images