Energy policy text structuring and correlation analysis system and method

Abstract

The invention provides an energy policy text structuring and correlation analysis method, and belongs to the technical field of policy analysis. The method comprises the following steps: collecting and preprocessing an energy policy text; performing semantic analysis and structured processing on the policy text based on a large language model, determining release time and semantic tags, and disassembling the policy text into structured semantic units; constructing a time sequence association relationship, a hyponymy logic association relationship and a semantic association relationship among policies according to the release time and the semantic unit; constructing an energy policy knowledge graph based on the association relationship and visually displaying the energy policy knowledge graph; and policy interpretation, trend tracking and cross-time and cross-region policy association comparative analysis are carried out by using the constructed knowledge graph. According to the method, intelligent processing and deep mining of the policy text are realized through structured processing and multi-dimensional correlation analysis of the energy policy text, and a systematic policy analysis tool is provided for policy formulators and researchers.

Description

Technical Field

[0001] This document relates to the field of energy policy analysis technology, and in particular to a system and method for structuring and correlating energy policy texts. Background Technology

[0002] With the deepening of energy transition and sustainable development, a large number of energy policy documents have been issued. These policy documents are of great significance in guiding the development of the energy industry and promoting energy structure adjustment. However, due to the large number of energy policy documents, their complex content, and the fact that most of them are unstructured, how to efficiently and systematically manage, analyze, and apply these policies has become an urgent problem to be solved.

[0003] Currently, the structured processing and knowledge graph construction of policy texts have been applied in multiple fields. For example, sorting and transforming data according to the "entity-relationship-entity" format helps to clarify the business context of carbon policies and provides support for the formulation and evaluation of dual-carbon policies. By collecting, filtering, and labeling electricity price policy data from multiple data sources and extracting basic attribute information to generate structured policy data, a knowledge graph of electricity price policies is ultimately formed. Entity extraction models are used to learn the semantic information in policy texts and accurately obtain the tag information, thereby supporting the accuracy and efficiency of intelligent policy retrieval.

[0004] However, existing technologies still have the following shortcomings when processing energy policy texts: First, existing policy text processing methods are mostly geared towards specific fields or single types of policy texts, lacking a systematic consideration of energy policy as a special field. Energy policy involves multiple sub-fields (such as electricity, oil, natural gas, renewable energy, etc.) and multiple levels (such as industry standards, local implementation rules, etc.), and existing methods are insufficient to fully cover these complex policy systems.

[0005] Secondly, existing technologies still have limitations in the semantic analysis and structured processing of policy texts. Traditional keyword matching and rule extraction methods struggle to accurately understand the technical terms, implicit logic, and complex expressions in policy language, resulting in inaccurate extracted policy elements and poor structuring effects.

[0006] Third, existing methods for constructing policy knowledge graphs primarily focus on static relationships between policy entities, lacking the ability to dynamically track policy evolution. Energy policies exhibit clear time-series characteristics and hierarchical progressions, making it difficult for existing methods to effectively capture the time-series relationships and hierarchical logical connections between policies.

[0007] Fourth, existing technologies have limited functionality in policy analysis and application. Most systems only provide basic search and query functions and lack in-depth policy interpretation, trend tracking, and cross-dimensional comparative analysis capabilities, making it difficult to meet the advanced needs of energy policy research and decision support.

[0008] Finally, most existing policy text processing systems use traditional natural language processing techniques and fail to fully utilize the advantages of large language models in semantic understanding and correlation analysis, which limits the accuracy and efficiency of the system when processing complex policy texts. Summary of the Invention

[0009] In view of the above-mentioned solutions, this application aims to propose a system and method for structuring and analyzing energy policy texts to solve at least one of the above problems.

[0010] In a first aspect, one or more embodiments of this application provide an energy policy text structuring and correlation analysis system, including: The data collection and preprocessing module is used to collect various energy policy texts from both domestic and international sources, and to clean and segment the collected texts. The policy text semantic analysis and structuring module is used to perform semantic analysis on preprocessed policy texts based on a large language model, determine the release time and semantic tags, and decompose the policy text into structured semantic units according to the semantic tags. The policy association construction module is used to determine the time-series association relationship between policy texts based on the release time and the semantic units, using a large language model; and to determine the hierarchical logical association relationship and semantic association relationship between policy texts based on the large language model and the semantic units. The knowledge graph construction and visualization module is used to construct an energy policy knowledge graph based on the time series association, the hierarchical logical association, and the semantic association. The policy analysis and application module is used to perform policy interpretation, trend tracking, and cross-time and cross-regional policy correlation comparison analysis based on the energy policy knowledge graph.

[0011] Furthermore, the semantic tags include at least one or more of the following: policy objectives, policy means, policy targets, and issuing authorities.

[0012] Furthermore, the policy association construction module includes: The semantic association construction unit is used to determine the semantic similarity between different policy texts based on the large language model and the semantic unit; and to determine policy texts with similar policy objectives, similar policy means or similar policy objects based on the similarity and to establish semantic association relationships.

[0013] Furthermore, for policy texts with semantic relationships, the policy analysis and application module is used to determine the corresponding evaluation dimensions based on semantic units; and to comment on the corresponding policy texts according to the evaluation dimensions.

[0014] Furthermore, the policy association construction module includes: The time-series correlation construction unit is used to sort the policy texts according to the release time; determine the evolution pattern of the policy texts in the time dimension according to the semantic unit; and construct the time-series correlation between the policy texts according to the sorting and the evolution pattern.

[0015] Furthermore, the time series correlation construction unit is also used to determine the event processing time nodes in the semantic unit; and to determine the time series correlation between the policy texts based on the event processing time nodes.

[0016] Furthermore, the policy association construction module includes: A hierarchical logical relationship construction unit is used to determine the relationship between each policy text in terms of policy level and content coverage based on a large language model and the semantic unit, and to construct a hierarchical logical relationship. Secondly, one or more embodiments of this application provide a method for structuring and analyzing energy policy texts, based on the energy policy text structuring and analysis system described in any one of the first aspects, comprising: collecting and processing energy policy texts using a data collection and preprocessing module; The policy text is semantically analyzed and structured using the policy text semantic analysis and structuring module to obtain the release time and semantic units; The policy association construction module constructs semantic associations, time-series associations, and hierarchical logical associations between policies based on the release time and the semantic unit. The knowledge graph construction and visualization module constructs and visualizes an energy policy knowledge graph based on the semantic relationships, time-series relationships, and hierarchical logical relationships. Through the policy analysis and application module, based on the energy policy knowledge graph, energy policies are interpreted, trends are tracked, and correlations are compared and analyzed.

[0017] Furthermore, constructing semantic relationships between policies includes: Based on the large language model, the semantic similarity between different policy texts is determined according to the semantic units; based on the similarity, policy texts with similar policy objectives, similar policy means or similar policy objects are identified, and semantic associations are established.

[0018] Thirdly, embodiments of this application provide a storage medium for storing computer-executable instructions, characterized in that, when executed, the computer-executable instructions implement the steps of the energy policy text structuring and correlation analysis method described in any one of the first aspects.

[0019] Compared with the prior art, this application can achieve at least the following technical effects: 1. Through the data collection and preprocessing module, the system achieves automated collection and unified cleaning and word segmentation of various domestic and international energy policy texts, solving the problems of diverse sources, inconsistent formats, and difficulty in large-scale collection of energy policy texts, and significantly improving the efficiency and standardization of policy text processing.

[0020] 2. Through the policy text semantic analysis and structuring module, deep semantic analysis of policy texts is performed based on a large language model. The policy release time is accurately extracted and semantic tags such as policy objectives, policy means, policy targets, and issuing authorities are automatically generated. This overcomes the shortcomings of traditional keyword matching methods in terms of the accuracy of identifying policy terminology and implicit logic, and significantly improves the accuracy of policy text structuring.

[0021] 3. By constructing a policy association module, the system builds time-series, hierarchical, and semantic relationships between policies based on a large language model. This breaks through the limitation of existing technologies that can only achieve static entity associations and realizes dynamic association analysis of energy policies in three dimensions: time, hierarchy, and semantics for the first time.

[0022] 4. By using the knowledge graph construction and visualization module, the above three relationships are integrated into a unified energy policy knowledge graph and visualized, which solves the problems of isolated policy text information, difficulty in intuitive understanding and overall grasp, and significantly improves the readability and visualization interaction capabilities of policy information.

[0023] 5. Through the policy analysis and application module, based on the aforementioned knowledge graph, policy interpretation, trend tracking, and cross-time and cross-regional policy correlation comparative analysis are realized. This overcomes the shortcomings of existing technologies that are limited to single functions and only provide basic search queries, significantly improving the systematicness and depth of energy policy analysis, and providing policymakers and researchers with an efficient, accurate, and comprehensive decision support tool. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in one or more embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A schematic diagram of the structure of an energy policy text structuring and correlation analysis system provided for one or more embodiments of this application; Figure 2 A flowchart illustrating a method for structuring and correlating energy policy texts, provided for one or more embodiments of this application. Detailed Implementation

[0026] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this application, the technical solutions in one or more embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on one or more embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the protection scope of this document.

[0027] Currently, data processing in the energy sector faces the following specific challenges: (1) Large data volume The primary characteristic of energy big data is that it replaces sampling analysis with full-sample analysis. Therefore, the data penetrates into every actual business segment of the energy group, and the data volume usually reaches the petabyte level. At the same time, the characteristics of big data determine that the energy database needs to use data "raw materials" with extremely low value density to produce high-value analytical conclusions. Therefore, when faced with a large amount of random data information, how to transform disordered data resources into valuable data assets, and then analyze them from the perspective of the full sample, accurately distinguish and locate different groups, and provide effective decision support for enterprises to expand business areas, innovate profit models, and promote enterprise transformation and upgrading has become the current challenge of the project.

[0028] (2) Data diversity The data types include unstructured, semi-structured, and structured data such as video, audio, text, and numbers. This project requires the use of big data technologies to discover potential rules from massive amounts of complex data, explore the inherent development trends of things, and make early warnings and predictions based on this. To this end, based on the analysis of various data types, this project supports the rapid access and high-quality processing of various types of data, including but not limited to mainstream relational databases, data interfaces, data import support for common file types, understanding and identification of satellite GIS information, analysis and recognition of multimedia materials, and natural language semantic processing.

[0029] (3) Multiple sources Data source systems are diverse, ranging in size. These systems include massive big data systems like electricity consumption information collection systems and wide-area monitoring systems, as well as multiple small to medium-sized systems with relatively small data volumes. These systems differ in their construction standards, data specifications, and even the naming conventions for the same physical equipment vary across different systems. Therefore, to analyze the big data systems comprised of this data, data integration and fusion become the first key technologies that energy big data databases must address.

[0030] (4) Data real-time performance Energy data platforms need to be highly timely to support intelligent functions such as policy change tracking, market fluctuation monitoring, and trend early warning. Achieving real-time synchronization and high-frequency data processing is a major challenge for the platform's practical application.

[0031] (5) Uneven value density Because of the differences in density and granularity between different data, it is impossible to directly reflect the correlation between data. Therefore, when faced with data of different value densities, how to conduct cross-domain analysis and realize the correlation between seemingly unrelated data sets is also an important challenge.

[0032] (6) Low transaction frequency From the perspective of overall development, my country's big data trading is still in its initial stage and faces some difficulties, such as the need to improve the data trading environment; the lack of a unified standard for data pricing in data trading, making it difficult to accurately measure the inherent value of data; and in terms of data quality, some transaction data has problems such as non-standard format and incomplete content, which affect data trading.

[0033] To address the above problems, this application provides the following embodiments: Example 1 An energy policy text structuring and correlation analysis system includes a data collection and preprocessing module 101, a policy text semantic analysis and structuring module 102, a policy correlation construction module 103, a knowledge graph construction and visualization module 104, and a policy analysis and application module 105.

[0034] The data collection and preprocessing module 101 is used to collect various domestic and international energy policy texts and to clean and segment the collected texts. Specifically, the data collection and preprocessing module automatically collects energy policy text data from official websites, energy authority websites, industry association websites, and other channels using web crawler technology. The collected data includes, but is not limited to, energy laws and regulations, administrative regulations, departmental rules, normative documents, and local-level energy policy documents. For the collected raw policy texts, the data collection and preprocessing module first performs text cleaning to remove special characters, redundant spaces, incorrect punctuation, and other interfering information; then, it performs word segmentation on the cleaned text, using Chinese word segmentation algorithms to divide the policy text into word sequences, laying the foundation for subsequent semantic analysis.

[0035] The policy text semantic analysis and structuring module 102 is used to perform semantic analysis on preprocessed policy texts based on a large language model, determining the release time and semantic tags; and decomposing the policy text into structured semantic units based on the semantic tags. Specifically, the module first calls a pre-trained large language model to perform deep semantic understanding on the segmented policy text. During the semantic analysis process, the module automatically identifies and extracts the release time information of the policy text, including time elements such as year, month, and day. Simultaneously, based on the semantic understanding capabilities of the large language model, the module generates semantic tags for the policy text. These semantic tags include at least one or more of the following: policy objectives, policy means, policy targets, and issuing authorities. After determining the semantic tags, the module further decomposes the policy text into structured semantic units. These semantic units are the smallest units with complete semantic meaning in the policy text, typically corresponding to clauses, paragraphs, or sentences. Each semantic unit is associated with a specific semantic tag, forming a structured representation of policy content.

[0036] The policy association construction module 103 is used to determine the temporal sequence relationship between policy texts based on the large language model, according to the release time and semantic units; and to determine the hierarchical logical relationship and semantic relationship between policy texts based on the large language model and semantic units. Specifically, the policy association construction module includes a time sequence relationship construction unit, a hierarchical logical relationship construction unit, and a semantic relationship construction unit.

[0037] The time-series correlation construction unit sorts policy texts chronologically based on their release dates, establishing a sequential relationship between them over time. Then, this unit analyzes the evolution of policy texts over time using semantic units, identifying continuity, points of change, and turning points in policy content. Based on the chronological order and evolutionary patterns, it constructs time-series correlations between policy texts, forming a timeline of policy evolution. Furthermore, the time-series correlation construction unit extracts event processing time nodes from semantic units, such as policy implementation dates, transition period deadlines, and completion dates of phased goals, and further refines the time-series correlations between policy texts based on these time nodes.

[0038] The hierarchical logical relationship construction unit is based on a large language model, analyzing the relationships between policy texts in terms of policy level and content coverage according to semantic units. The policy level relationship is mainly reflected in the hierarchical relationship between policy texts of different levels of effectiveness, such as laws, administrative regulations, departmental rules, and normative documents; the content coverage relationship is mainly reflected in the hierarchical relationship between general policies and specific policies, and between framework policies and implementing rules. By analyzing these relationships, the hierarchical logical relationships between policy texts are constructed, forming the hierarchical structure of the policy system.

[0039] The semantic association construction unit is based on a large language model, calculating the semantic similarity between different policy texts based on semantic units. Specifically, the semantic units of the policy texts are converted into vector representations, and then the cosine similarity or other similarity indices are calculated between the vectors. Based on the calculated similarity, policy texts with similar policy objectives, similar policy measures, or similar policy objects are identified, and semantic associations are established. For example, two policy texts on photovoltaic industry subsidies issued at different times and by different departments, although with different backgrounds, will be identified as having a semantic association because of their similar policy objects and policy measures.

[0040] The knowledge graph construction and visualization module 104 is used to construct an energy policy knowledge graph based on time-series relationships, hierarchical logical relationships, and semantic relationships. Specifically, this module uses policy texts as nodes in the knowledge graph and various relationships as edges, constructing a multi-dimensional, multi-layered energy policy knowledge graph. In the knowledge graph, node attributes include policy title, publication time, issuing authority, semantic tags, etc.; edge attributes include relationship type, relationship strength, etc. The completed knowledge graph is displayed using visualization technology, supporting interactive operations such as node filtering, relationship filtering, and path tracing, enabling users to intuitively understand the overall structure and internal relationships of the energy policy system.

[0041] The knowledge graph construction and visualization module 104 organizes documents or data during the visualization process. Specifically, based on a preset chapter framework, it calls a large language model to integrate and create daily report content. The model is based on the highest quality materials selected by RAG, and combines them with chapter themes to logically connect and polish the language; it adjusts the expression style according to the characteristics of different chapters (such as concise and succinct in the "Market Dynamics" section, and in-depth analysis in the "Key Focus" section); it automatically supplements policy background information and related policy prompts to enhance the completeness and readability of the content; During the generation process, the model must strictly adhere to the principle of "authenticity of materials," ensuring that all statements are based on the original policy text, avoiding subjective assumptions, and guaranteeing the authority of the daily report content. The generation time for a single policy daily report has been reduced from 3 hours manually to 10 minutes.

[0042] After the content is generated, the system enters the automated formatting process: The daily report is formatted uniformly according to a preset template (including font, font size, paragraph spacing, headers and footers, chapter title styles, etc.). It automatically inserts charts (such as visualization elements like listed company information tables). It supports exporting in multiple formats (PDF, Word, HTML, etc.) and integrates with internal distribution systems (such as email push notifications and platform publishing) to enable rapid deployment and application of the daily report. This step eliminates the tedious process of manual formatting, reducing the 15 minutes required for manual formatting to 1 minute, ensuring the timeliness of the daily policy report.

[0043] The knowledge graph construction and visualization module 104 enables the efficient transformation of scattered information into structured daily reports. Compared with the traditional manual mode, it not only reduces the production cycle from an average of 8 hours to within 1-2 hours, but also improves the coverage, accuracy and interpretation depth of policy information through technical means, providing strong support for decision-makers to quickly grasp policy dynamics and formulate response strategies.

[0044] The Policy Analysis and Application Module 105 is used for policy interpretation, trend tracking, and cross-time and cross-regional policy correlation comparative analysis based on an energy policy knowledge graph. Specifically, this module provides the following functions: the policy interpretation function automatically generates policy key point interpretations and policy background analyses by analyzing node attributes and edge relationships in the knowledge graph; the trend tracking function identifies the development trends and key changes in energy policies by analyzing time series correlations; and the policy correlation comparative analysis function allows users to select policies from different periods, regions, or departments for parallel comparison, highlighting policy differences and commonalities. For policy texts with semantic relationships, the Policy Analysis and Application Module determines corresponding evaluation dimensions based on semantic units, such as policy coordination, policy continuity, and policy innovation; then, based on these evaluation dimensions, it comments on the corresponding policy texts, providing decision-making references for policymakers and researchers.

[0045] The policy analysis and application module performs in-depth analysis of the full text of policies collected daily, automatically extracting high-value "highlights." Traditionally, manual processing of a single policy text and extraction of core segments takes an average of 3 hours, while an AI agent leveraging NLP technology can complete the same task within 30 minutes. In terms of completeness and accuracy, it is essentially on par with manual results. This significant improvement in efficiency provides core support for the efficient production of daily policy reports.

[0046] Example 2 A method for structuring and correlating energy policy texts based on the system described in Embodiment 1 includes the following steps: Step 1: Collect and process energy policy texts using the data collection and preprocessing module.

[0047] Specifically, energy policy text data is automatically collected using web crawling technology. The collected data includes various forms of text such as laws and regulations, administrative rules, normative documents, and policy interpretations. The raw text data is cleaned, including removing HTML tags, special characters, redundant spaces, and other interfering information, and standardizing the text format and encoding. Then, a professional Chinese word segmentation tool is used to segment the cleaned text, identifying technical terms and policy-specific vocabulary, and generating word sequences. Simultaneously, part-of-speech tagging and named entity recognition are performed on the segmentation results to provide a foundation for subsequent semantic analysis.

[0048] In the actual data collection process, emphasis was placed on researching and learning from advanced international energy databases. Given the mature systems and widespread influence of internationally renowned energy databases such as IEA, BP, and BNEF in terms of indicator design, data standardization, service interfaces, and intelligent analysis, the project team systematically analyzed their data types, organizational logic, update mechanisms, and application models from a practical user perspective, summarizing their core advantages. Since most external databases are commercial platforms, making it difficult to gain in-depth access to their internal structure and development details, the team focused on combining daily usage experience with publicly available literature to extract valuable elements from a user's perspective. Ultimately, the goal was to ensure that the collected data possessed international benchmarking capabilities in areas such as indicator system construction, automated data collection, knowledge integration, and intelligent services.

[0049] Step 2: Perform semantic analysis and structuring on the policy text using the policy text semantic analysis and structuring module to obtain the release time and semantic units.

[0050] Specifically, a pre-trained large language model is invoked to perform deep semantic understanding on the segmented policy text. The model first identifies and extracts time information from the policy text, determining time attributes such as the policy's release date, effective date, and expiration date. Then, based on its semantic understanding capabilities, the model generates semantic tags for the policy text, including key information such as policy objectives, policy measures, policy targets, and issuing authorities. Based on the generated semantic tags, the policy text is decomposed into structured semantic units. During this decomposition process, the model identifies the logical structure of the policy text, such as chapters, clauses, and paragraphs, and associates each structural unit with its corresponding semantic tag, forming structured units with semantic attributes. These semantic units are the basic units for subsequent association analysis.

[0051] Step 3: Using the policy association construction module, construct semantic associations, time-series associations, and hierarchical logical associations between policies based on their release time and semantic units.

[0052] Specifically, the process begins by constructing semantic relationships. Based on a large language model, the semantic units of policy texts are converted into vector representations, and the vector similarity between semantic units of different policy texts is calculated. Based on a similarity threshold, policy texts with similar policy objectives, similar policy measures, or similar policy objects are identified, and semantic relationships are established. Next, a temporal sequence relationship is constructed. Policy texts are ordered chronologically according to their release date. The ordered policy texts are analyzed to identify the evolution of policy content over time, such as adjustments to policy objectives and changes in policy measures. Based on the chronological order and evolution patterns, a temporal sequence relationship is constructed between policy texts. Furthermore, event processing time nodes, such as policy implementation time and phased target time, are extracted from the semantic units to further refine the temporal sequence relationship. Finally, a hierarchical logical relationship is constructed. Based on a large language model, the semantic units of policy texts are analyzed to determine the relationship between policy texts in terms of policy level and content coverage. The hierarchical relationship between policies of different levels of effectiveness, such as laws, administrative regulations, and departmental rules, as well as the inclusion relationship between general policies and specific policies, and between framework policies and implementation details, are identified, and a hierarchical logical relationship is constructed.

[0053] Step 4: Using the knowledge graph construction and visualization module, construct an energy policy knowledge graph based on semantic relationships, time series relationships, and hierarchical logical relationships, and then visualize it.

[0054] Specifically, a multi-dimensional, multi-layered energy policy knowledge graph is constructed by using policy texts as nodes and various relationships as edges. Node attributes in the knowledge graph include policy title, publication date, issuing authority, and semantic tags; edge attributes include relationship type and relationship strength. The constructed knowledge graph is stored using graph database technology, supporting efficient graph query and analysis operations. Visualization techniques are used to display the knowledge graph, including force-directed graphs, tree diagrams, and timeline graphs. Interactive functions such as node filtering, relationship filtering, and path tracing are implemented, allowing users to explore the energy policy system from different dimensions. Zooming, panning, and focusing operations on the knowledge graph are provided to enhance the user experience.

[0055] In real-world scenarios, the following patterns can also be used to construct semantic connections: Information filtering: BERT / TextCNN is used for text classification (identifying policy topics and categories), combined with a rule engine to accurately filter high-value policies, eliminate invalid information, and solve the problem of "information overload"; Fragment Mining: Focusing on key policy dimensions (key measures, corporate impact, energy dynamics, etc.), based on natural language processing technology, it performs fine-grained analysis of policy texts, automatically identifies and extracts key content such as core viewpoints, industry trends, and corporate-related information, helping users quickly obtain decision-level information; Translation Alignment: Based on the Transformer model, Chinese-English translation is achieved, coupled with "domain terminology alignment" to break down language barriers and make domestic and foreign policies comparable and referential. RAG Enhancement: Through text vectorization and knowledge retrieval, and then using a re-ranking model to ensure that the output content is highly matched with user needs, intelligent and three-dimensional policy interpretation is achieved; Daily Report Generation: Based on a template engine, it pre-sets standardized daily report formats (covering modules such as policy summary, core measures, impact analysis, and action recommendations), achieving decoupled configuration of content and format. Through automated typesetting tools, it generates standardized policy daily reports that meet office needs with a single click, improving information application efficiency.

[0056] Step 5: Through the policy analysis and application module, based on the energy policy knowledge graph, interpret energy policies, track trends, and conduct comparative analysis.

[0057] Specifically, based on node attributes and edge relationships in the knowledge graph, the system automatically generates interpretations of policy key points and analyses of policy background. The interpretations include crucial information such as the policy's core objectives, main measures, applicable targets, and expected effects. By analyzing time-series relationships, the system identifies development trends and key changes in energy policy. Trend analysis includes shifts in policy focus, evolution of policy tools, and changes in policy intensity. Users can select policies from different periods, regions, or departments for side-by-side comparison, highlighting policy differences and commonalities. Comparison dimensions include the consistency of policy objectives, the complementarity of policy tools, and the coordination of policy implementation. For policy texts with semantic relationships, corresponding evaluation dimensions are determined based on semantic units, such as policy coordination, policy continuity, and policy innovation. Then, based on these evaluation dimensions, the system provides commentary on the corresponding policy texts, offering decision-making references for policymakers and researchers.

[0058] Example 3 Based on Example 1, semantic tags include at least one or more of the following: policy objectives, policy means, policy targets, and issuing authorities.

[0059] Specifically, policy objectives refer to the expected effects or states explicitly stated in the policy text, such as "promoting the development of renewable energy," "controlling carbon emissions," and "ensuring energy security." Policy means refer to the specific measures or tools taken to achieve policy objectives, such as "fiscal subsidies," "tax incentives," "administrative licensing," and "planning guidance." Policy targets refer to the subjects or objects of policy action, such as "the photovoltaic industry," "wind power companies," and "electricity consumers." Issuing authorities refer to the departments or institutions that formulate and issue policies.

[0060] When performing semantic analysis, the policy text semantic analysis and structuring module first identifies keywords and phrases related to these semantic tags in the policy text. For example, for policy objectives, the module identifies goal-oriented verbs such as "promote," "strengthen," "improve," and "achieve," as well as related objective descriptions; for policy means, the module identifies nouns indicating specific measures such as "subsidy," "tax reduction," "approval," and "planning"; for policy targets, the module identifies professional terms and entity names in the energy sector; and for issuing authorities, the module identifies the standard names of departments and institutions.

[0061] After identifying these keywords and phrases, the module leverages the semantic understanding capabilities of large language models to integrate this scattered information into structured semantic tags.

[0062] These semantic tags provide an important basis for subsequent policy text structuring and correlation analysis. The policy text semantic analysis and structuring module breaks down the policy text into structured semantic units based on these semantic tags. Each semantic unit is associated with a specific semantic tag, forming a structured representation with semantic attributes. For example, clauses in the policy text concerning fiscal subsidies are marked as semantic units related to "policy means - fiscal subsidies," and clauses concerning photovoltaic module production are marked as semantic units related to "policy targets - photovoltaic industry."

[0063] This structured processing based on semantic tags enables the system to understand the content and intent of policy texts at the semantic level, laying the foundation for subsequent policy correlation analysis and knowledge graph construction.

[0064] Example 4 Based on Example 1, the policy association construction module includes a semantic association construction unit, which is used to determine the semantic similarity between different policy texts based on the large language model and the semantic unit; and to determine policy texts with similar policy objectives, similar policy means or similar policy objects based on the similarity, and to establish semantic association relationships.

[0065] Specifically, the semantic association construction unit first converts the semantic units of the policy text into vector representations. This process utilizes the semantic encoding capabilities of large language models to map the text content to vectors in a high-dimensional semantic space. For each policy text, multiple vectors can be generated, each corresponding to a different semantic unit, such as a policy objective vector, a policy means vector, and a policy object vector.

[0066] Then, the semantic association building unit calculates the similarity of semantic unit vectors between different policy texts. Commonly used similarity calculation methods include cosine similarity, Euclidean distance, and dot product. For example, using cosine similarity to calculate the similarity between the policy objective vectors of two policy texts can assess the degree of similarity between the two policies in their goal setting.

[0067] Based on the calculated similarity, the semantic association construction unit identifies policy texts with semantic relationships. Specifically, a similarity threshold is set. When the similarity between two policy texts in a certain semantic dimension (such as policy objectives, policy means, or policy objects) exceeds the threshold, the two policy texts are considered to have a semantic relationship in that dimension.

[0068] The semantic association building unit also distinguishes the association strength across different semantic dimensions. For example, if two policy texts have high similarity across three dimensions—policy objectives, policy means, and policy objects—then the semantic association between them is strong; conversely, if they only have high similarity across one dimension, the association strength is relatively low.

[0069] The established semantic relationships are recorded as edges in the knowledge graph. The attributes of the edges include the type of relationship (such as goal relationship, means relationship, object relationship) and the strength of the relationship. These semantic relationships provide an important basis for subsequent policy analysis, enabling the system to discover policy texts that appear unrelated but are actually intrinsically connected.

[0070] For policy texts with semantic relationships, the policy analysis and application module is used to determine the corresponding evaluation dimensions based on semantic units; and to comment on the corresponding policy texts according to the evaluation dimensions.

[0071] Specifically, the policy analysis and application module first determines the dimensions suitable for evaluating relevant policy texts based on the content and attributes of the semantic units. These evaluation dimensions include, but are not limited to: policy coordination, assessing the consistency of goals and complementarity of means among different policies; policy continuity, assessing the continuity and stability of policies over time; policy innovation, assessing the degree of innovation in the problem-solving methods of policies; policy feasibility, assessing the resource conditions and environmental constraints for policy implementation; and policy targeting, assessing the targeting of policies for specific problems or targets.

[0072] Then, the policy analysis and application module systematically reviews policy texts with semantic relationships based on the determined evaluation dimensions. During the review process, the module analyzes the semantic units of the policy texts, extracts content relevant to the evaluation dimensions, and generates evaluation conclusions based on the understanding capabilities of a large language model. For example, for two policy texts with related policy objectives, the module reviews them from the perspective of policy coordination, analyzing whether the objectives of the two policies are consistent, complementary, or conflicting; for policy texts with a time-series relationship, the module reviews them from the perspective of policy continuity, analyzing whether the changes in policy objectives and means over time are reasonable and whether necessary stability has been maintained.

[0073] The review results are presented in a structured format, including evaluation dimensions, key evaluation points, and evaluation conclusions. These reviews provide valuable reference information for policymakers and researchers, helping them understand the connections and consistency between policies and identify potential problems and areas for improvement within the policy system.

[0074] Example 5 Based on Example 1, the policy association construction module includes a time-series association construction unit, which is used to sort policy texts according to their release time; determine the evolution pattern of policy texts in the time dimension according to semantic units; and construct time-series associations between policy texts based on the sorting and evolution patterns.

[0075] Specifically, the time-series correlation construction unit first sorts all policy texts chronologically based on their release date. The sorting can be done with precision of year, month, or day, forming a time series of policy texts. For multiple policy texts released on the same day, a secondary sorting can be performed based on the issuing authority's hierarchy or the policy's importance.

[0076] Then, the time-series correlation construction unit, based on semantic units, analyzes the evolution of policy texts over time. This process mainly focuses on the changes in policy objectives, policy tools, and policy targets over time. For example, analyzing the changes in the objectives of energy structure adjustment policies at different periods may reveal an evolution from the early "increasing the proportion of renewable energy" to the mid-term "controlling fossil energy consumption" and then to the recent "achieving carbon peaking and carbon neutrality"; analyzing the changes in the means of photovoltaic industry support policies at different periods may reveal an evolution from the early "large-scale subsidies" to the mid-term "competitive bidding for grid connection" and then to the recent "grid parity".

[0077] Based on the analysis of evolutionary patterns, time-series correlation units are used to identify the continuity points, change points, and turning points of policy content. Continuation points refer to the parts of policy content that remain unchanged over time, indicating policy stability; change points refer to the parts of policy content that undergo gradual adjustments, indicating policy adaptability; and turning points refer to the parts of policy content that undergo fundamental changes, indicating major policy adjustments.

[0078] Based on the time sequence and evolution patterns, the time sequence relation construction unit constructs the time sequence relations between policy texts. These relations include: successive relations, indicating that the later policy is a continuation or adjustment of the earlier policy; substitution relations, indicating that the later policy replaces the earlier policy; and supplementary relations, indicating that the later policy supplements or elaborates on the earlier policy.

[0079] The constructed time-series relationships are recorded as edges in a knowledge graph, with attributes including relationship type (e.g., successor, substitute, supplement) and relationship strength. These time-series relationships provide important evidence for understanding the historical evolution and future trends of energy policy.

[0080] Example 6 Based on Example 5, the time series correlation construction unit is also used to determine the event processing time nodes in the semantic unit; and to determine the time series correlation between policy texts based on the event processing time nodes.

[0081] Specifically, the time-series correlation construction unit first extracts event processing time nodes from the semantic units of the policy text. These time nodes include, but are not limited to: policy implementation time, i.e., the time when the policy officially takes effect; transition period end time, i.e., the time when the policy transition arrangements end; phased goal completion time, i.e., the time when the phased goals set in the policy should be achieved; policy evaluation time, i.e., the time when the policy implementation effect is evaluated; and policy adjustment time, i.e., the time when the policy is planned to be adjusted.

[0082] Then, the time series correlation construction unit further refines the time series correlation between policy texts based on the extracted event processing time nodes. Specifically, this involves analyzing the relationships between the event processing time nodes of different policy texts to identify temporal connections and overlaps. For example, if the implementation time of policy A coincides with the transition period deadline of policy B, it indicates that policy A may be a follow-up policy to policy B; if the completion time of the phased objectives of policy C coincides with the policy evaluation time of policy D, it indicates that policy D may be evaluated based on the implementation effects of policy C.

[0083] Based on these temporal relationships, the time-series correlation construction unit can more accurately determine the temporal relationships between policy texts. For example, it can identify the update and iteration relationship of policies, that is, a later policy is an update or replacement of a previous policy; identify the supporting relationship of policies, that is, multiple policies are implemented in a coordinated manner in time to jointly serve a certain goal; and identify the evaluation and feedback relationship of policies, that is, a later policy is formulated based on the evaluation results of a previous policy.

[0084] This correlation analysis based on event processing time nodes enables the system to gain a deeper understanding of the temporal relationships between policy texts, providing a more accurate basis for policy trend analysis and policy effectiveness evaluation.

[0085] Example 7 Based on Example 1, the policy association construction module includes a hierarchical logical association construction unit, which is used to determine the relationship between each policy text in terms of policy level and content coverage based on the large language model and semantic units, and to construct hierarchical logical associations.

[0086] Specifically, the hierarchical logical relationship construction unit first analyzes the semantic units of policy texts based on a large language model to identify the type and nature of policy texts. Policy types include laws, administrative regulations, departmental rules, normative documents, etc.; policy natures include general policies, specific policies, framework policies, implementation rules, etc.

[0087] Then, the hierarchical logical relationship construction unit determines the relationship between policy texts in terms of legal effect hierarchy based on policy type. According to my country's legal system, different types of policy texts have different levels of legal effect, from highest to lowest: laws, administrative regulations, departmental rules, normative documents, etc. Based on this hierarchical relationship, the hierarchical relationship of legal effect between policy texts is constructed.

[0088] Meanwhile, the hierarchical logical relationship construction unit determines the relationship between policy texts in terms of content coverage based on the nature and content of the policy. This relationship is mainly reflected in the inclusion relationship between the general and specific policies, and between the framework and detailed rules.

[0089] Furthermore, the hierarchical logical relationship construction unit also analyzes the citation and basis relationships in policy texts. If policy B explicitly cites policy A as the basis for its formulation in its text, it indicates that policy A is in a superior position and policy B is in a subordinate position. This citation relationship is an important basis for determining the hierarchical logical relationship.

[0090] Based on the above analysis, the hierarchical logical relationship construction unit constructs the hierarchical logical relationships between policy texts. These relationships are recorded as edges in the knowledge graph, and the attributes of the edges include the relationship type (such as legal effect relationship, content coverage relationship, citation basis relationship) and the relationship strength.

[0091] The construction of this hierarchical logical relationship enables the system to present the hierarchical structure of the energy policy system, helping users understand the logical relationships and levels of effectiveness between policies, and providing guidance for policy interpretation and implementation.

[0092] It should be noted that Examples 1, 2, 3, 4, 5, 6, and 7 are all types of energy policy text structuring and correlation analysis systems.

[0093] This application provides a storage medium for storing computer-executable instructions, characterized in that, when executed, the computer-executable instructions implement the steps of the energy policy text structuring and correlation analysis method described in any one of the embodiments.

[0094] It should be noted that the embodiments concerning storage media in this application and the embodiments concerning the energy policy text structuring and correlation analysis method in this application are based on the same inventive concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the corresponding energy policy text structuring and correlation analysis method mentioned above, and the repeated parts will not be described again.

[0095] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0096] In the 1930s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many improvements to the methodology today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that an improvement to the methodology cannot be implemented using hardware physical modules. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must also be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0097] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0098] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0099] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, in implementing the embodiments of this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0100] Those skilled in the art will understand that one or more embodiments of this application can be provided as a method, system, or computer program product. Therefore, one or more embodiments of this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0101] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0102] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0103] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0104] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0105] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0106] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0107] It should also be noted that 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 limitation, 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.

[0108] One or more embodiments of this application can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. One or more embodiments of this application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can reside in local and remote computer storage media, including storage devices.

[0109] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0110] The above description is merely an embodiment of this document and is not intended to limit the scope of this document. Various modifications and variations can be made to this document by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this document should be included within the scope of the claims of this document.

Claims

1. A system for structuring and correlating energy policy texts, characterized in that, include: The data collection and preprocessing module is used to collect various energy policy texts from both domestic and international sources, and to clean and segment the collected texts. The policy text semantic analysis and structuring module is used to perform semantic analysis on preprocessed policy texts based on a large language model, determine the release time and semantic tags, and decompose the policy text into structured semantic units according to the semantic tags. The policy association construction module is used to determine the time-series association relationship between policy texts based on the release time and the semantic units, using a large language model; and to determine the hierarchical logical association relationship and semantic association relationship between policy texts based on the large language model and the semantic units. The knowledge graph construction and visualization module is used to construct an energy policy knowledge graph based on the time series association, the hierarchical logical association, and the semantic association. The policy analysis and application module is used to perform policy interpretation, trend tracking, and cross-time and cross-regional policy correlation comparison analysis based on the energy policy knowledge graph.

2. The system according to claim 1, characterized in that, The semantic tags include at least one or more of the following: policy objectives, policy means, policy targets, and issuing authorities.

3. The system according to claim 1, characterized in that, The policy association construction module includes: The semantic association construction unit is used to determine the semantic similarity between different policy texts based on the large language model and the semantic unit; and to determine policy texts with similar policy objectives, similar policy means or similar policy objects based on the similarity and to establish semantic association relationships.

4. The system according to claim 3, characterized in that, For policy texts with semantic relationships, the policy analysis and application module is used to determine the corresponding evaluation dimensions based on semantic units; and to comment on the corresponding policy texts according to the evaluation dimensions.

5. The system according to claim 1, characterized in that, The policy association construction module includes: The time-series correlation construction unit is used to sort the policy texts according to the release time; determine the evolution pattern of the policy texts in the time dimension according to the semantic unit; and construct the time-series correlation between the policy texts according to the sorting and the evolution pattern.

6. The system according to claim 5, characterized in that, The time series correlation construction unit is also used to determine the event processing time nodes in the semantic unit; and to determine the time series correlation between the policy texts based on the event processing time nodes.

7. The system according to claim 1, characterized in that, The policy association construction module includes: The hierarchical logical relationship construction unit is used to determine the relationship between the policy texts in terms of policy level and content coverage based on the large language model and the semantic unit, and to construct the hierarchical logical relationship.

8. A method for structuring and correlating energy policy texts based on the system described in any one of claims 1-7, characterized in that, Includes the following steps: The data collection and preprocessing module is used to collect and process energy policy texts. The policy text is semantically analyzed and structured using the policy text semantic analysis and structuring module to obtain the release time and semantic units; The policy association construction module constructs semantic associations, time-series associations, and hierarchical logical associations between policies based on the release time and the semantic unit. The knowledge graph construction and visualization module constructs and visualizes an energy policy knowledge graph based on the semantic relationships, time-series relationships, and hierarchical logical relationships. Through the policy analysis and application module, based on the energy policy knowledge graph, energy policies are interpreted, trends are tracked, and correlations are compared and analyzed.

9. The method according to claim 8, characterized in that, Constructing semantic relationships between policies includes: Based on the large language model, the semantic similarity between different policy texts is determined according to the semantic units; based on the similarity, policy texts with similar policy objectives, similar policy means or similar policy objects are identified, and semantic associations are established.

10. A storage medium for storing computer-executable instructions, characterized in that, When executed, the computer-executable instructions implement the steps of the method of claim 8 or 9.

Images