A method, system and device for checking the compliance of a system document architecture based on semantic mapping, and a storage medium

Abstract

The application discloses a kind of system file architecture compliance verification method, system, equipment and storage medium based on semantic mapping, related to computer information processing technical field, method includes: configuration verification rule and access enterprise data, system file is parsed and semantically classified, while organizational post information is structured modeling, constructs the full-factor semantic graph of connecting file, organization, platform and architecture, calls multiple models and carries out multidimensional semantic verification according to graph and rule, identifies non-compliance item;Automatic risk assessment, generate rectification suggestion and tracking report, realize closed-loop management;The application realizes the intelligentization of system file, full-dimensional architecture compliance verification, significantly improves verification accuracy and efficiency, guarantees the cooperation of management system and business and digitization, forms the rectification closed loop that can be traced.

Description

Technical Field

[0001] This invention relates to the field of computer information processing technology, and in particular to a method, system, device, and storage medium for verifying compliance with institutional document architecture based on semantic mapping. Background Technology

[0002] In modern enterprise management, institutional documents are the core carriers for standardizing operations, controlling risks, and ensuring the implementation of strategies. As enterprises expand and digital transformation deepens, the system of institutional documents becomes increasingly complex. Whether its content is complete, consistent, and effective, and whether it can coordinate with the organizational structure, business processes, and digital platforms, has become crucial to the effectiveness of the management system.

[0003] Currently, the industry primarily employs two approaches to address the compliance and consistency issues of policy documents: First, it relies entirely on manual review, with specialists comparing each item of the document content line by line. This method is inefficient, highly subjective, and struggles to cover all relevant dimensions. Second, it utilizes automated tools based on keyword matching or fixed rules, such as scanning specific fields or using simple logical judgments for conflict detection. These approaches also have significant drawbacks: document classification methods are crude, often relying solely on filename keywords, lacking a deep understanding of the content's semantics, leading to inaccurate classification and retrieval difficulties; verification dimensions are extremely limited, typically only checking version expiration, format compliance, or explicit textual duplication, failing to identify implicit semantic contradictions across documents, discrepancies between policy requirements and job responsibilities, and mismatches between policy processes and digital platform functions; furthermore, existing methods generally lack the ability to structurally model and semantically map the relationships between various elements of an enterprise (policies, organization, processes, systems), resulting in isolated and one-sided verification results. They fail to assess policy compliance from the perspective of the overall enterprise architecture, exhibiting low levels of intelligence and high reliance on manual intervention. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention provides a method, system, device and storage medium for verifying compliance of institutional document architecture based on semantic mapping.

[0005] Therefore, the technical problem solved by this invention is: how to achieve panoramic and intelligent architecture compliance verification of institutional documents based on deep semantic understanding.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for verifying compliance with institutional document architecture based on semantic mapping, comprising: Obtain raw data and preset structured validation rules from external data sources and store them in the central data warehouse; Document data is extracted from the central data warehouse, semantic parsing and multi-dimensional classification are performed, unique semantic classification labels are generated and stored back; Information related to personnel and functions is extracted from the central data warehouse, and structured to construct a semantic model containing hierarchical and attribute relationships. Semantic tags are generated for each element in the semantic model and then stored back. By integrating classified documents and semantic models, and combining them with platform metadata, a full-element semantic association graph is constructed and stored through semantic vectorization and similarity calculation. Based on the verification rules and the semantic association graph of all elements, multiple verification models are invoked to perform multi-dimensional semantic verification on the policy documents, generating a verification result set containing non-compliant items; Based on the verification rules, risk assessments and classifications are performed on non-compliant items in the verification result set, and rectification suggestions and verification reports are automatically generated by combining the full-element semantic association graph.

[0007] As a preferred solution for a semantic mapping-based institutional document architecture compliance verification method, wherein: The process of acquiring raw data and preset structured validation rules from external data sources and storing them in a central data warehouse includes: Access multi-source enterprise data, including enterprise management needs, architecture information, and original policy data, and configure data interfaces with external business systems; Preset verification rules, the verification rules include: Semantic classification rules for multi-level fine-grained classification of files, dynamic semantic matching threshold rules for quantitative judgment during classification and verification, and multi-dimensional quantitative scoring rules for risk assessment of identified problems.

[0008] The beneficial effects of this preferred technical solution are as follows: By clearly defining the composition of the verification rules (classification, matching, and scoring) and integrating all necessary multi-source enterprise data and external systems at once, a unified, configurable benchmark and complete data foundation are established for the entire verification process. This not only ensures the consistency and standardization of all subsequent steps (such as classification, verification, and evaluation), avoiding verification deviations caused by missing rules or data silos, but also significantly improves the system's scalability and adaptability to different enterprise environments.

[0009] As a preferred solution for a semantic mapping-based institutional document architecture compliance verification method, wherein: The step of extracting document data from the central data warehouse, performing semantic parsing and multi-dimensional classification, generating unique semantic classification labels, and storing them back includes: The management system documents are obtained from the central data warehouse. Based on the document format and metadata, at least one of the following methods is used to extract the text content and metadata: native content extraction, optical character recognition, structure parsing, or direct interface reading. The extracted metadata is then converted into a standardized format. Semantic features are extracted from the parsed text content. Domain-knowledge-optimized word segmentation and semantic understanding techniques are used to obtain the deep semantic vector representation, core keywords, business scenarios, and management requirements of the text, and organize them into a standardized set of semantic features. Based on the set of semantic features and the preset classification rules, multi-level semantic matching is performed to obtain file association relationships. The matching includes: determining the business domain to which the file belongs based on semantic similarity, determining the management dimension based on scene feature matching within the determined business domain, and determining the file type based on functional semantic matching. Based on the matching results, a structured semantic classification label containing business domain, management dimension, file type and unique identifier information is generated, and the structured semantic classification label, semantic feature set and file association relationship are updated and stored in the central data warehouse.

[0010] The beneficial effects of this preferred technical solution are as follows: Through in-depth analysis of files of different formats and semantic feature extraction optimized based on domain knowledge, the content of files can be accurately understood. Combined with pre-defined multi-level classification rules for progressive semantic matching, automated and refined classification of institutional documents from business domain and management dimensions to file type is achieved, generating globally unique semantic tags. This fundamentally solves the problems of traditional methods' coarse classification and reliance on manual labor, greatly improving the standardization of file management and retrieval efficiency, and providing high-quality structured data for subsequent establishment of accurate semantic associations and verification.

[0011] As a preferred solution for a semantic mapping-based institutional document architecture compliance verification method, wherein: The process of extracting personnel and function-related information from the central data warehouse, performing structured processing, constructing a semantic model containing hierarchical and attribute relationships, generating semantic tags for each element in the semantic model, and then storing the data includes: Organizational structure and job responsibility information is obtained from the central data warehouse. Through parsing and information extraction technologies, elements such as organizational hierarchy, job positions, job responsibilities, and scope of authority are identified and separated. Standardized semantic tags are generated for each extracted element. Among them, the job responsibility tag is generated by matching and semantic analysis based on the preset business domain and management dimension rules, and the permission tag is generated based on permission keywords and hierarchical division. Based on the consistency of business logic and semantic similarity calculation, the semantic association between job positions and responsibility items is automatically established, and the rationality of the association is verified and corrected. Using organizational hierarchy, job positions, responsibilities, and scope of authority as core dimensions, a tree-structured organizational model with semantic relationships between elements is constructed, and the tree-structured organizational model and the semantic tags of the elements are updated and stored in the central data warehouse.

[0012] As a preferred solution for a semantic mapping-based institutional document architecture compliance verification method, wherein: The process involves fusing classified documents and semantic models, combining them with platform metadata, and constructing and storing a full-element semantic association graph through semantic vectorization and similarity calculation. Extract categorized institutional documents, structured organizational models, digital platform elements, and enterprise architecture elements from the central data warehouse, and convert all types of elements into standardized semantic vectors generated based on pre-trained semantic models; Based on the standardized semantic vector, candidate associations between different categories of elements are identified through semantic similarity calculation, wherein the semantic similarity calculation is modified by combining association importance weights, and candidate association pairs are selected according to a preset threshold. The candidate association pairs are verified according to multi-dimensional logical verification rules, which include business logic consistency verification, hierarchical relationship matching verification, and functional semantic association verification, in order to eliminate associations that do not conform to logic. Based on the candidate association pairs that have passed logical verification, a multi-level hierarchical mapping relationship is established according to the preset hierarchical structure; Based on the hierarchical mapping relationship, a full-element semantic association graph containing all element nodes and semantic mapping edges is constructed using a graph structure and stored in the central data warehouse.

[0013] As a preferred solution for a semantic mapping-based institutional document architecture compliance verification method, wherein: The method, based on verification rules and a comprehensive semantic association graph, calls multiple verification models to perform multi-dimensional semantic verification on the policy documents, generating a verification result set containing non-compliant items, including: By comparing the semantics of policy documents with the semantics of management needs and architectural blueprints, the completeness and content conflicts of the documents are verified; by comparing the semantics of job responsibilities with the semantics of requirements in related policy documents and analyzing permission settings, the compliance of organizational positions is verified; by comparing the semantics of the functions, data, and processes of the digital platform with the semantics of requirements in policy documents, the synergy of the digital platform is verified; by analyzing the matching degree and consistency between the semantics of policy documents and the semantics of enterprise architectural elements, architectural compliance is verified. By integrating the outputs of various verification models, each identified problem is organized into a structured and standardized verification result set based on related elements, problem type, semantic basis, and quantitative comparison results.

[0014] As a preferred solution for a semantic mapping-based institutional document architecture compliance verification method, wherein: The process of risk assessment and classification of non-compliant items in the verification result set according to verification rules, and automatic generation of rectification suggestions and verification reports based on the full-element semantic association graph, includes: For each non-compliant item in the verification result set, a risk assessment is conducted based on a preset multi-dimensional quantitative scoring rule. The risk assessment is based on a weighted calculation of a comprehensive risk value, including multiple dimensions such as the scope of impact, severity, rectification difficulty, and probability of occurrence. The risk level and priority are then classified according to the comprehensive risk value. Based on the aforementioned semantic association graph of all elements, the system documents, organizational positions and digital platform elements associated with non-compliant items are analyzed, and targeted rectification suggestions are automatically generated in combination with the semantic basis of non-compliant items. The suggestions are then broken down into specific implementation steps that include the responsible parties and time limits. Based on the risk assessment results and the generated rectification suggestions, a structured rectification tracking document and a standardized verification report are output.

[0015] Secondly, the present invention provides a compliance verification system for institutional document architecture based on semantic mapping, comprising: The data and rules configuration module is used to obtain raw data and preset structured validation rules from external data sources and store them in the central data warehouse. The document semantic classification module is used to extract document data from the central data warehouse, perform semantic parsing and multi-dimensional classification, generate unique semantic classification labels and store them back; The organization modeling module is used to extract information related to personnel and functions from the central data warehouse, perform structured processing, construct a semantic model containing hierarchical and attribute relationships, generate semantic labels for each element in the semantic model, and then save the data. The semantic graph construction module is used to integrate classified documents and semantic models, combine them with platform metadata, and construct and store a full-element semantic association graph through semantic vectorization and similarity calculation. The intelligent verification engine module is used to perform multi-dimensional semantic verification of institutional documents based on verification rules and the semantic association graph of all elements, and to generate a set of verification results containing non-compliant items. The rectification management module is used to conduct risk assessment and classification of non-compliant items in the verification result set according to the verification rules, and automatically generate rectification suggestions and verification reports by combining the full-element semantic association graph.

[0016] Thirdly, the present invention provides a computer device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of a semantic mapping-based institutional document architecture compliance verification method.

[0017] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of a semantic mapping-based regulatory document architecture compliance verification method.

[0018] The beneficial effects of this invention are as follows: By constructing an intelligent verification system based on semantic mapping, this invention effectively solves the long-standing problems in enterprise system document management, such as crude classification, single verification dimensions, insufficient semantic understanding, high reliance on manual labor, and difficulty in rectification and traceability. It realizes automated and accurate classification and tagging management of system documents based on content semantics, significantly improving document retrieval and governance efficiency. It establishes a full-scenario, multi-dimensional intelligent verification capability covering document content, organizational positions, digital platforms, and enterprise architecture, which can accurately discover semantic conflicts, coverage omissions, and collaboration gaps that are difficult to identify using traditional methods, ensuring a deep integration of the management system with business and digital construction. By automatically generating accurate rectification suggestions and implementation paths and establishing a closed-loop traceability archive throughout the entire lifecycle, it significantly reduces the cost of manual review and subjective errors, ensuring the effective rectification and implementation of non-compliance issues, and comprehensively improving the standardization, digitalization level, and compliance control efficiency of the enterprise management system. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is an overall flowchart of a semantic mapping-based institutional document architecture compliance verification method provided by the present invention. Detailed Implementation

[0021] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0022] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for verifying compliance with institutional document architecture based on semantic mapping, including: S1: Obtain raw data and preset structured validation rules from external data sources and store them in the central data warehouse; S2: Extract document data from the central data warehouse, perform semantic parsing and multi-dimensional classification, generate unique semantic classification labels and store them back; S3: Extract information related to personnel and functions from the central data warehouse, perform structured processing, construct a semantic model containing hierarchical and attribute relationships, generate semantic tags for each element in the semantic model, and then save the data. S4: Integrate classified documents and semantic models, combine them with platform metadata, and construct and store a full-element semantic association graph through semantic vectorization and similarity calculation; S5: Based on the verification rules and the semantic association graph of all elements, multiple verification models are called to perform multi-dimensional semantic verification on the policy documents and generate a verification result set containing non-compliant items; S6: Based on the verification rules, conduct risk assessment and classification of non-compliant items in the verification result set, and automatically generate rectification suggestions and verification reports by combining the full-element semantic association graph.

[0023] Example 2, refer to Figure 1 As an embodiment of the present invention, based on the previous embodiment, a method for verifying compliance with institutional document architecture based on semantic mapping is provided, including: In this embodiment, step S1 above, which involves obtaining raw data and preset structured validation rules from an external data source and storing them in a central data warehouse, includes: The data in the enterprise data source includes, but is not limited to, enterprise management requirements lists, business architecture blueprints, and enterprise architecture data; Configure external system interface parameters to complete the connection with digital platforms, human resources systems, etc.

[0024] Specifically, the preset verification rules include: Classification rules: The first level (business domain) is divided into six core business domains: production, human resources, finance, R&D, supply chain, and quality. Each business domain is defined by 3-5 core semantic keywords (e.g., keywords for the production domain: production process, equipment operation, production safety, and capacity control). The second level (management dimensions): Each business domain has 5-8 management dimensions. For example, the production domain includes dimensions such as process management, equipment management, safety management, capacity management, and cost management. Management dimensions are defined by semantic scenario features (such as equipment management dimension features: equipment maintenance, equipment inspection, equipment repair, and equipment calibration). The third level (document type): Each management dimension is divided into 5 types of documents, including procedure documents, management system, operation specifications, technical standards, and record forms. These are defined by the semantic meaning of the document function (e.g., the semantic meaning of procedure document function: clearly define the business execution process, responsible party, and operation steps).

[0025] In another possible implementation, when setting multi-level refined semantic classification rules, the classification rules may not be completely statically preset. Some dimensions (such as the scene feature lexicon under the management dimension) support adaptive clustering and updating based on the semantic feature distribution of classified files during system operation, forming a dynamically evolving classification knowledge base.

[0026] In another possible implementation, when setting up semantic classification rules for multi-level refined classification, in addition to the core three levels of "business domain - management dimension - document type", an optional fourth level of "applicable region / factory area" can be added. The matching rules of this level are based on specific geographical location keywords or codes mentioned in the document to meet the differentiated management system classification needs of cross-regional group enterprises.

[0027] Semantic matching threshold: Basic matching threshold: When classifying and matching files, a semantic similarity of ≥85% is considered an exact match and is directly classified; 60%-84% is considered a fuzzy match and triggers manual review by the administrator; <60% is considered a mismatch, marked as "file to be classified" and prompts for supplementary semantic information. Verification matching thresholds: In the verification of the coverage and integrity of files and management requirements and architectural elements, semantic coverage ≥90% is judged as complete coverage, 70%-89% is judged as partial coverage, and <70% is judged as no coverage; In the verification of cross-file semantic conflicts, semantic contradiction ≥75% is judged as major conflict, 50%-74% is judged as general conflict, and <50% is judged as no conflict.

[0028] Risk level classification criteria: Risk scoring dimensions include four dimensions: scope of impact (core business / non-core business), severity (business disruption / efficiency impact / minor flaws), difficulty of rectification (high / medium / low), and probability of occurrence (high / medium / low). Scoring Rules: Each dimension is quantified with 1-3 points (e.g., scope of impact: 3 points for core business, 1 point for non-core business; severity: 3 points for business disruption, 2 points for efficiency impact, 1 point for minor flaws). The total score is calculated as: Scope of impact score × 2 + Severity score × 2 + Difficulty of rectification score + Probability of occurrence score. The total score ranges from 2 to 18 points. The risk levels are divided into: high risk (12-18 points), medium risk (7-11 points), and low risk (2-6 points), corresponding to different rectification time limits.

[0029] In this embodiment, step S2 above, which involves extracting document data from the central data warehouse, performing semantic parsing and multi-dimensional classification, generating unique semantic classification labels, and storing them back, includes: Obtain various management system documents submitted by enterprises from the central data warehouse, such as the "Production Process Management Method" and the "Human Resources Management Manual".

[0030] Different parsing schemes are employed for different file formats: For PDF files, a dual-mode fusion scheme combining OCR recognition and native text extraction is used. Specifically, for PDF files with copyable text, the native text content and metadata (such as publication date and version number) in document attributes are directly extracted. For scanned image-format PDF files, the text content is first extracted using an Optical Character Recognition (OCR) model based on a deep learning architecture (such as a combination of an optimized Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) network). Then, a keyword matching algorithm is used to retrieve keywords such as "publication date," "version number," and "effective date" from the text to extract corresponding metadata field values. For Word files (DOC / DOCX), a custom parsing interface is called to directly read the file body, header and footer information, and document attributes. Built-in metadata (such as author and modification date) is extracted by parsing XML tags. Simultaneously, regular expressions are used to match text in specific formats such as "publication date: XXXX year XX month XX day" and "version: Vx.x" to supplement the extraction of non-built-in metadata. Subsequently, all extracted metadata was uniformly converted into standardized formats such as "Release Date: YYYY-MM-DD", "Version Number: Vx.x", and "Effective Status: Effective / Ineffective", and stored in the central data warehouse.

[0031] After completing file parsing and metadata extraction, an optimized natural language processing (NLP) technology solution is adopted to extract semantic features from the parsed text content. Specifically, it includes: performing word segmentation on the text. This process is based on the jieba word segmentation tool and has been optimized, and a professional term dictionary for the industry where the enterprise is located is particularly supplemented to improve the accuracy of word segmentation; removing stop words. The stop word list used is a custom word list for the enterprise management field, including common meaningless words such as "of", "already", "according to", etc., as well as general and non-distinguishing words in this industry; performing词性标注 (it seems there is a mistake here, assuming it should be "词性标注" which means "pos tagging" in English). On this basis, by using a BERT pre-trained model fine-tuned to adapt to the characteristics of enterprise management texts to extract the deep semantic vector representation of the text, and at the same time combining the TF-IDF algorithm to calculate the keyword weights, the top 20 keywords with the highest weights are selected as the core semantic features. In addition, through semantic role labeling technology, the core business scenarios described in the text (such as "equipment inspection", "leave approval") and specific management requirements (such as "requiring approval by the department manager", "at least once a month") are identified. Finally, the extracted keywords and their weights, identified business scenarios and management requirements are organized in the structure of "keyword - weight - business scenario - management requirement" to form a standardized semantic feature set.

[0032] Based on the standardized semantic feature set and the pre-configured three-level classification rules, multi-dimensional classification matching is carried out. The matching process is specifically as follows: First, perform business domain matching. Calculate the semantic similarity (using the cosine similarity algorithm) between the keywords in the semantic features and the core keywords defined in each business domain, and determine the candidate domain as the business domain with the highest similarity; then perform management dimension matching. Under the candidate business domain, calculate the matching degree between the identified core business scenarios and the semantic scenario features defined in each management dimension to determine the candidate management dimension; then perform file type matching. Based on the identified management requirements and the functional semantics defined for various file types, perform matching analysis to determine the final file type. The matching result is determined according to the preset semantic matching threshold: If the matching degree reaches the precise matching threshold, it directly enters the label generation link; if it is in the fuzzy matching range, a system prompt is triggered and the administrator conducts manual review; if the matching degree is lower than the threshold, the file is marked as a "file to be classified".

[0033] To generate and synchronize unique semantic classification tags for matched files, the system employs a four-level structure: "Business Domain - Management Dimension - File Type - Unique Identifier". The unique identifier is composed of the release date extracted from the file's standardized metadata (converted to YYYYMMDD format) and a randomly generated 6-digit number, ensuring global uniqueness across the entire domain. For example, "Production Domain - Equipment Management - Operation Specifications - 20250115 - 678901". The tag generation process includes: automatically concatenating and generating three levels of basic tags ("Business Domain - Management Dimension - File Type") based on the classification matching results; extracting the file release date and combining it with random numbers to generate a unique identifier; and concatenating the basic tags with the unique identifier to form a complete classification tag. The uniqueness of the generated tags is automatically verified; if duplicates are found, the random number portion is regenerated. Administrators can view the automatic classification results and generated tags through a visual interface and can manually adjust the classification dimensions based on actual business needs. Tags will be regenerated based on the adjustments, and all manual correction records are synchronously archived. Finally, the classified files, their corresponding standardized semantic feature sets, and the generated unique classification labels are synchronously updated and stored in the central data warehouse, and an association mapping relationship between files and labels is established.

[0034] In another possible implementation, when performing parsing and semantic analysis, for online documents generated through collaborative editing in the cloud, the open application programming interface provided by the document platform can be called to directly obtain its plain text content, version history, and collaborative metadata, thereby completing the parsing; in the semantic analysis stage, a dedicated pre-trained model based on the Transformer architecture and trained on a massive amount of Chinese official documents is mainly used to extract semantic vectors and identify key information.

[0035] In another possible implementation, when performing parsing and semantic analysis, for binary or proprietary format files unique to the enterprise, a parser or conversion tool customized for that format can be used to convert them into an intermediate format text before further processing; during semantic analysis, rule template matching and a lightweight neural network model are integrated to quickly identify and classify frequently occurring fixed phrases or clauses in the file.

[0036] In another possible implementation, when generating a unique identifier, a one-way hash operation (such as MD5 or SHA-256) can be performed on the combined string of "file release date + file title", and the first few characters of the hash value can be extracted as the identifier to ensure uniqueness and enhance anti-collision capability.

[0037] In another possible implementation, the standardized code of the department to which the document belongs, the document security classification code, and a sequence number assigned by the central data warehouse can be combined to form a unique identification code, so that the label itself also carries some management attribute information.

[0038] In this embodiment, step S3 above involves extracting information related to personnel and functions from the central data warehouse, performing structured processing, constructing a semantic model containing hierarchical and attribute relationships, generating semantic tags for each element in the semantic model, and then storing the data back. Process organizational charts (supporting Visio, Excel, PPT, and other formats) and job descriptions (supporting Word and PDF formats) obtained from the central data warehouse. Parse the organizational charts to extract four-level organizational hierarchy information: "Company-Department-Team-Position," and use an entity recognition model to identify department and position names. Parse the job descriptions, using a semantic segmentation algorithm to break down the job descriptions into independent job items, and use keyword matching technology to extract permission scope information.

[0039] Standardized semantic tags are generated for the extracted elements. Organizational hierarchy tags are generated using a "hierarchy-name" structure. Job responsibility tags are generated using a "business domain-management dimension-responsibility type-core action" structure: the split responsibility items are matched with predefined business domain and management dimension rules to determine the business domain and management dimension, and then the responsibility type and core action are identified through semantic analysis and assembled into a complete tag. Permission scope tags are generated using a "permission type-permission object-permission level" structure: the permission type is determined by identifying permission keywords, the permission object is determined by combining contextual semantics, and the permission level is determined according to the preset hierarchy. Tag validation is performed to ensure that each element corresponds to a unique semantic tag.

[0040] Establish semantic associations between job titles and responsibilities. Initial associations are automatically established based on the consistency of job title and responsibility tags across business domains and management dimensions. The similarity between the semantic vectors of job titles and responsibilities is calculated, and the reasonableness of the associations is verified according to preset thresholds, eliminating associations with insufficient similarity. A visual interface is provided for administrators to view the association results, supporting manual addition or deletion of associations; all corrections are recorded.

[0041] A tree-structured organizational model is constructed using "organizational hierarchy - position - responsibility - authority" as the core dimensions. The model data is organized according to the association structure of "hierarchy ID - position ID - responsibility ID - authority ID - semantic tag" and is synchronously updated and stored in the central data warehouse, enabling retrieval of related data by any dimension.

[0042] In another possible implementation, when constructing a tree-structured organizational model with semantic relationships between elements, the tree model can not only exist in the form of an association table during physical storage, but its nodes and relationships can be mapped to an attribute graph database. The native characteristics of the graph database can be used to efficiently store and traverse the complex tree and network relationships between "organization-position-responsibility".

[0043] In another possible implementation, when constructing a tree-structured organizational model with semantic relationships between elements, in addition to extracting static job descriptions, the actual job data, part-time information, and temporary roles in project teams in the employee human resources system can be synchronized through an interface. These dynamic job information can be incorporated into the overall organizational model as subtrees or special nodes to make it more in line with the actual operating status of the enterprise.

[0044] In this embodiment, step S4 above integrates classified documents and semantic models, combines platform metadata, and constructs and stores a full-element semantic association graph through semantic vectorization and similarity calculation, including: The data is extracted from the central data warehouse, including categorized policy documents and their semantic features and classification labels, structured organizational job models and their element semantic labels and relationships, digital platform elements (including functional semantics, process semantics, and data semantics), and enterprise architecture elements (including business architecture semantics and application architecture semantics). Semantic standardization preprocessing is performed on each element, converting them into standardized semantic vectors generated based on a pre-trained BERT model. The converted elements and their semantic vectors are then stored in four categories: policy documents, organizational job models, digital platform elements, and enterprise architecture elements.

[0045] When calculating semantic similarity, the cosine similarity algorithm is used to calculate the basic similarity between semantic vectors of elements from different categories. To more accurately reflect the importance of association, a preset element association weight is introduced to weight and correct the basic similarity, resulting in a corrected similarity. Based on a preset semantic matching threshold, element pairs with a corrected similarity reaching a specified value (e.g., ≥60%) are selected as candidate association pairs.

[0046] Logical correlation analysis is performed on candidate association pairs to verify their rationality. Business logic validation is conducted to analyze whether candidate association pairs conform to the enterprise's business logic. For example, it is ensured that "production domain policy documents" are only associated with "production department positions" and "production-related digital platform functions," eliminating logically unrelated pairs across business domains. Hierarchical logic validation is performed, verifying the hierarchical matching of candidate association pairs based on the preset hierarchical relationship of "enterprise architecture - policy documents - organizational positions - digital platform." Functional logic validation is performed, verifying the rationality of associations based on the functional semantics of elements, eliminating candidate association pairs with no substantial functional overlap.

[0047] Based on the candidate association pairs that have passed logical verification, establish a clear hierarchical mapping relationship. Establish a first-level mapping between enterprise architectural elements and policy document elements. Establish a second-level mapping between policy document elements and organizational position elements. Establish a third-level mapping between policy document elements and digital platform elements. Establish a fourth-level mapping between organizational position elements and digital platform elements.

[0048] A full-element semantic association graph is constructed using a graph database, with each element as a node and the defined semantic mapping relationships as edges. Node attributes include element ID, semantic label, and semantic vector; edge attributes include corrected similarity, association type, and logical verification results. A community detection algorithm is used to cluster closely related element nodes in the graph, forming dedicated sub-graphs divided by business domain to improve retrieval efficiency. The completed full-element semantic association graph is synchronized to a central data warehouse, supporting real-time updates and multi-dimensional retrieval.

[0049] In this embodiment, step S5 above, based on the verification rules and the full-element semantic association graph, calls multiple verification models to perform multi-dimensional semantic verification on the policy document, generating a verification result set containing non-compliant items, including: Multiple pre-defined verification models are invoked to perform verification based on data in the central data warehouse and the constructed full-element semantic association graph. The models used include a document-based verification model, an organizational job compliance verification model, a digital platform association verification model, and an architecture compliance verification model.

[0050] The document basic verification model performs the following operations: First, it performs coverage integrity verification by extracting the semantic feature vectors of the management requirements list and business architecture blueprint, which serve as the verification benchmark. These vectors are compared with the semantic feature vectors of the policy documents. The semantic coverage is calculated using the formula "Coverage = Intersection of semantic features / Total number of semantic features × 100%", and determined as complete, partial, or no coverage based on a preset threshold. Second, it performs content conflict verification by using a conflict detection algorithm based on a pre-trained model to analyze the semantic consistency of policy documents belonging to the same business domain and management dimension. The semantic contradiction degree is calculated to identify regulatory conflicts, and the conflict level is determined based on a preset threshold. Third, it performs timeliness and validity verification by extracting the publication date and effective status from the policy document's metadata. This information is then combined with preset policy timeliness rules for calculation and comparison to determine the document's validity.

[0051] The organizational job compliance verification model performs the following operations: First, it verifies the completeness of responsibility coverage by obtaining all relevant policy documents associated with a specific job through a semantic mapping graph. The semantic features of management requirements in these documents are extracted and compared with the semantic features of job responsibilities. The responsibility coverage is calculated, and a threshold is used to determine completeness. Second, it verifies the matching degree between responsibilities and policies by using a semantic matching algorithm to calculate the similarity between the semantic vector of job responsibilities and the semantic vector of corresponding policy requirements. A threshold is used to determine sufficiency of the matching degree. Third, it verifies permission conflicts and reasonableness by analyzing the permission tags in the organizational job model to determine whether there are logical conflicts between different permissions within the same job, whether there are permission overlaps between different jobs, and whether the configured permissions are semantically related to the job responsibilities.

[0052] The digital platform association verification model performs the following operations: Functional coverage verification: It obtains the digital platform functional elements associated with policy documents through a semantic mapping graph, extracts the semantic features of policy requirements, compares them with the semantic features of platform functions, calculates functional coverage, and determines sufficiency based on a threshold. Data consistency verification: It uses a semantic alignment algorithm to compare the data field definitions of the digital platform with the data record requirements of the policy documents to determine semantic consistency. Process matching verification: It analyzes the semantic matching degree between the business processes implemented in the digital platform and the processes stipulated in the policy documents, and determines whether they match based on a threshold.

[0053] The architecture compliance verification model performs the following operations: It conducts a semantic matching assessment, using a semantic similarity algorithm to calculate the similarity between the semantic vector of the policy document and the semantic vector of the enterprise architecture elements, and determines whether the matching degree is sufficient based on a threshold. It then conducts a consistency assessment, analyzing whether the management requirements and business processes embodied in the policy document are consistent with the planning goals and design principles of the enterprise architecture. Finally, it combines the results of the matching and consistency assessments to determine whether the policy document meets the enterprise architecture requirements.

[0054] By integrating the outputs of all verification models, each identified issue is organized according to the structure of "non-compliance item ID - associated element - verification model - non-compliance type - semantic basis - similarity / coverage value" to form a standardized verification result set.

[0055] In this embodiment, step S6 above involves risk assessment and level classification of non-compliant items in the verification result set according to verification rules, and automatically generating rectification suggestions and verification reports based on the full-element semantic association graph, including: For each non-compliant item in the verification result set, a quantitative score and classification are performed according to a preset risk level classification rule. A quantitative score of 1-3 points is assigned based on four dimensions: the scope of impact, severity, rectification difficulty, and probability of occurrence of the non-compliant item. A comprehensive risk score is then calculated using a preset weighted formula. Based on the preset range of the total score, the non-compliant item is classified into high-risk, medium-risk, or low-risk levels, and a corresponding risk level label and rectification priority indicator are attached to each non-compliant item. Simultaneously, each non-compliant item is marked with a status of "Pending Rectification," "In Progress," "Completed," or "No Rectification Required," with the initial status defaulting to "Pending Rectification." The marked status can be synchronously updated via an external system interface based on the actual rectification progress.

[0056] Based on a comprehensive semantic association graph, the system automatically generates rectification suggestions and specific implementation steps for non-compliance items. The graph identifies all elements associated with the non-compliance item (such as specific policy documents, job positions, and digital platform functions), and, combined with the specific type and semantic basis of the non-compliance item, generates targeted and precise rectification suggestions. These suggestions are further broken down into specific implementation steps under the logic of "requirements analysis - solution formulation - implementation - verification," clearly defining the responsible entity (department or position determined based on organizational job model matching), planned completion deadline, and key milestones for each step. The generated rectification suggestions and implementation steps are optimized by combining them with the enterprise's historical rectification case library stored in the central data warehouse to improve their feasibility and effectiveness.

[0057] Provides rectification tracking and report generation functions. It supports exporting rectification progress information, including key fields such as non-compliance item ID, risk level, rectification suggestions, implementation steps, responsible party, completion deadline, rectification status, and review results, into a standard format (such as Excel) rectification tracking table to facilitate offline tracking and management. It automatically generates standardized verification reports, including a verification overview (such as the total number of verification documents, the total number of non-compliance items, and the percentage of each risk level), a detailed list of non-compliance items sorted by risk level, a summary of rectification suggestions, and a rectification tracking plan. The report supports exporting in PDF or Word format and contains a unique identifier, allowing traceability of all associated verification records through a central data warehouse.

[0058] Furthermore, all data generated during steps S2 to S6, including file classification results, organizational modeling data, verification process records, rectification suggestions, and status records, are uniformly archived and managed to form a full lifecycle verification archive. This archive supports users in searching and tracing historical data by file type, business domain, rectification status, and other dimensions. Administrators can track the rectification progress in real time and review completed rectification items, thus forming a complete "verification-rectification-review-archiving" management closed loop.

[0059] Example 3: The above is an illustrative scheme of a semantic mapping-based institutional document architecture compliance verification method according to this embodiment. It should be noted that the technical solution of a semantic mapping-based institutional document architecture compliance verification system belongs to the same concept as the above-described semantic mapping-based institutional document architecture compliance verification method. Details not described in detail in the semantic mapping-based institutional document architecture compliance verification system in this embodiment can be found in the description of the above-described semantic mapping-based institutional document architecture compliance verification method.

[0060] This embodiment also provides a compliance verification system for institutional document architecture based on semantic mapping, including: The data and rules configuration module is used to obtain raw data and preset structured validation rules from external data sources and store them in the central data warehouse. The document semantic classification module is used to extract document data from the central data warehouse, perform semantic parsing and multi-dimensional classification, generate unique semantic classification labels and store them back; The organization modeling module is used to extract information related to personnel and functions from the central data warehouse, perform structured processing, construct a semantic model containing hierarchical and attribute relationships, generate semantic labels for each element in the semantic model, and then save the data. The semantic graph construction module is used to integrate classified documents and semantic models, combine them with platform metadata, and construct and store a full-element semantic association graph through semantic vectorization and similarity calculation. The intelligent verification engine module is used to perform multi-dimensional semantic verification of institutional documents based on verification rules and the semantic association graph of all elements, and to generate a set of verification results containing non-compliant items. The rectification management module is used to conduct risk assessment and classification of non-compliant items in the verification result set according to the verification rules, and automatically generate rectification suggestions and verification reports by combining the full-element semantic association graph.

[0061] This embodiment also provides an electronic device applicable to a semantic mapping-based institutional document architecture compliance verification method, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes the computer-executable instructions to implement a semantic mapping-based compliance verification method for institutional document architecture as proposed in the above embodiments.

[0062] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements a semantic mapping-based institutional document architecture compliance verification method as proposed in the above embodiments.

[0063] The storage medium proposed in this embodiment belongs to the same inventive concept as the semantic mapping-based institutional document architecture compliance verification method proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0064] Example 4 is an embodiment of the present invention, which provides a method for verifying compliance of institutional document architecture based on semantic mapping. In order to verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.

[0065] Taking a manufacturing company's "Management System Document Structure Compliance Verification" project as an example, the application process of the present invention is explained in detail: 1. Data initialization and resource configuration: Enter the enterprise's management requirements list (including core requirements such as "intelligent equipment inspection" and "employee attendance management"), business architecture blueprint (including 6 major business domains such as production domain and human resources domain), enterprise architecture data and preset verification rules into the central data warehouse; complete the interface connection with the enterprise's existing ERP digital platform and human resources management system.

[0066] 2. Semantic Classification of Management System Documents: Enterprises submit various management system documents, including "Production Process Management Methods" (Word format), "Human Resources Management Manual" (PDF format), "Employee Attendance Management System" (Word format), and "Production Equipment Safety Specifications" (PDF format). Multi-format file parsing: Parsing the "Production Process Management Measures" (Word) to extract the main text and metadata (Publication date: 2023-01-15, Version: V1.0); Parsing the "Production Equipment Safety Specifications" (scanned PDF) to extract text through OCR recognition and matching keywords to extract metadata (Publication date: 2021-03-20, Version: V1.0, Effective status: Expired); Semantic feature extraction: After preprocessing, the core keywords (production process, equipment operation, process control) of the "Production Process Management Measures" are extracted, with the core business scenario being "equipment operation process" and the management requirement being "equipment maintenance once a month"; Multi-dimensional classification matching: The keywords have a 92% semantic similarity to the core keywords of the production domain (exact match), the business scenario is matched to the "equipment management" dimension, and the management requirement is matched to the "program file" type; Generate classification tags: "Production domain - equipment management - program file - 20230115-123456", which are verified and confirmed by the administrator and synchronized to the central data warehouse.

[0067] 3. Organizational Structure Modeling: Import the company's organizational chart (Company - Production Department - Equipment Maintenance Group - Equipment Administrator) and job descriptions ("Equipment Administrator" responsibilities: daily equipment maintenance, inspection records, and fault reporting). Extract core information: Extract the four-level organizational hierarchy, job titles, three responsibilities, and "Equipment Operation Permissions" and "Inspection Record Permissions"; Generate semantic tags: Job tag "Job - Production Department - Equipment Administrator", responsibility tags "Production Domain - Equipment Management - Maintenance Responsibilities - Daily Maintenance" and "Production Domain - Equipment Management - Inspection Responsibilities - Inspection Records", permission tag "Operation Permissions - Equipment Data - Input / Modification"; Establish semantic associations: Automatically associate the job with the two responsibilities (semantic similarity 85%), and associate the job with permissions; Construct a structured model: Form a tree structure of "Production Department - Equipment Maintenance Group - Equipment Administrator - Maintenance Responsibilities - Inspection Responsibilities - Operation Permissions", and synchronize it to the central data warehouse.

[0068] 4. Construct a full-element semantic mapping graph: Establish relationships: "Enterprise Architecture - Production Execution Process" → "Production Process Management Measures" → "Equipment Administrator" → "ERP Equipment Management Module", forming a production domain sub-graph and synchronizing it to the central data warehouse.

[0069] 5. Multi-dimensional semantic intelligent verification: Utilizes four major AI verification models to perform verification. Document basic verification: The semantics of "Intelligent Equipment Inspection" are not covered when comparing the "Production Process Management Regulations" with the management requirements list (coverage rate 65%, judged as not covered); the semantics of "Employee Attendance Management System" and "Human Resources Management Manual" are contradictory at a rate of 82% (judged as a major conflict); "Production Equipment Safety Specifications" have expired (judged as invalid).

[0070] Organizational job compliance verification: For the "Equipment Administrator" position, the job responsibility label "Production Domain - Equipment Management - Inspection Responsibility - Inspection Record" matches the semantic requirements of the "Production Process Management Measures" by 90% (judged as a match); however, the "Intelligent Equipment Inspection" responsibility that it should possess is not covered in the corresponding system (match rate 45%, judged as a mismatch).

[0071] Digital platform association verification: The functional semantic coverage of the "Equipment Management Module" in the ERP system with the "Production Process Management Method" is 75% (below the threshold, judged as insufficient functional coverage), which lacks the functional semantics related to "intelligent inspection data entry".

[0072] Architecture compliance verification: The semantic similarity between the "Production Process Management Measures" and the enterprise's production domain architecture is 92% (determined to meet the architecture requirements).

[0073] 6. Non-compliance handling and rectification management: Risk level classification: "Equipment intelligent inspection not covered by the system" total score 13 points (high risk), "Leave approval process conflict" total score 14 points (high risk), "Insufficient equipment management module function" total score 10 points (medium risk); Generate rectification suggestions and steps: For "not covered by the system", it is recommended to supplement the system chapters and platform functions, break down 4 implementation steps, and the responsible entities are the production department, technical department, etc.; Export rectification tracking table and verification report: Export Excel tracking table and PDF verification report.

[0074] 7. Full lifecycle archiving and traceability: The system archives all data, and managers can track the rectification progress through the archive module. After the rectification is completed, the verification model is called to review and confirm that all non-compliant items have been rectified, forming a closed-loop management.

[0075] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for verifying compliance with institutional document architecture based on semantic mapping, characterized in that, include: Obtain raw data and preset structured validation rules from external data sources and store them in the central data warehouse; Document data is extracted from the central data warehouse, semantic parsing and multi-dimensional classification are performed, unique semantic classification labels are generated and stored back; Information related to personnel and functions is extracted from the central data warehouse, and structured to construct a semantic model containing hierarchical and attribute relationships. Semantic tags are generated for each element in the semantic model and then stored back. By integrating classified documents and semantic models, and combining them with platform metadata, a full-element semantic association graph is constructed and stored through semantic vectorization and similarity calculation. Based on the verification rules and the semantic association graph of all elements, multiple verification models are invoked to perform multi-dimensional semantic verification on the policy documents, generating a verification result set containing non-compliant items; Based on the verification rules, risk assessments and classifications are performed on non-compliant items in the verification result set, and rectification suggestions and verification reports are automatically generated by combining the full-element semantic association graph.

2. The method for verifying compliance with institutional document architecture based on semantic mapping as described in claim 1, characterized in that, The process of acquiring raw data and preset structured validation rules from external data sources and storing them in a central data warehouse includes: Access multi-source enterprise data, including enterprise management needs, architecture information, and original policy data, and configure data interfaces with external business systems; Preset verification rules, the verification rules include: Semantic classification rules for multi-level fine-grained classification of files, dynamic semantic matching threshold rules for quantitative judgment during classification and verification, and multi-dimensional quantitative scoring rules for risk assessment of identified problems.

3. The method for verifying compliance with institutional document architecture based on semantic mapping as described in claim 2, characterized in that, The step of extracting document data from the central data warehouse, performing semantic parsing and multi-dimensional classification, generating unique semantic classification labels, and storing them back includes: The management system documents are obtained from the central data warehouse. Based on the document format and metadata, at least one of the following methods is used to extract the text content and metadata: native content extraction, optical character recognition, structure parsing, or direct interface reading. The extracted metadata is then converted into a standardized format. Semantic features are extracted from the parsed text content. Domain-knowledge-optimized word segmentation and semantic understanding techniques are used to obtain the deep semantic vector representation, core keywords, business scenarios, and management requirements of the text, and organize them into a standardized set of semantic features. Based on the set of semantic features and the preset classification rules, multi-level semantic matching is performed to obtain file association relationships. The matching includes: determining the business domain to which the file belongs based on semantic similarity, determining the management dimension based on scene feature matching within the determined business domain, and determining the file type based on functional semantic matching. Based on the matching results, a structured semantic classification label containing business domain, management dimension, file type and unique identifier information is generated, and the structured semantic classification label, semantic feature set and file association relationship are updated and stored in the central data warehouse.

4. The method for verifying compliance with institutional document architecture based on semantic mapping as described in claim 3, characterized in that, The process of extracting personnel and function-related information from the central data warehouse, performing structured processing, constructing a semantic model containing hierarchical and attribute relationships, generating semantic tags for each element in the semantic model, and then storing the data includes: Organizational structure and job responsibility information is obtained from the central data warehouse. Through parsing and information extraction technologies, elements such as organizational hierarchy, job positions, job responsibilities, and scope of authority are identified and separated. Standardized semantic tags are generated for each extracted element. Among them, the job responsibility tag is generated by matching and semantic analysis based on the preset business domain and management dimension rules, and the permission tag is generated based on permission keywords and hierarchical division. Based on the consistency of business logic and semantic similarity calculation, the semantic association between job positions and responsibility items is automatically established, and the rationality of the association is verified and corrected. Using organizational hierarchy, job positions, responsibilities, and scope of authority as core dimensions, a tree-structured organizational model with semantic relationships between elements is constructed, and the tree-structured organizational model and the semantic tags of the elements are updated and stored in the central data warehouse.

5. The method for verifying compliance with institutional document architecture based on semantic mapping as described in claim 4, characterized in that, The process involves fusing classified documents and semantic models, combining them with platform metadata, and constructing and storing a full-element semantic association graph through semantic vectorization and similarity calculation. Extract categorized institutional documents, structured organizational models, digital platform elements, and enterprise architecture elements from the central data warehouse, and convert all types of elements into standardized semantic vectors generated based on pre-trained semantic models; Based on the standardized semantic vector, candidate associations between different categories of elements are identified through semantic similarity calculation, wherein the semantic similarity calculation is modified by combining association importance weights, and candidate association pairs are selected according to a preset threshold. The candidate association pairs are verified according to multi-dimensional logical verification rules, which include business logic consistency verification, hierarchical relationship matching verification, and functional semantic association verification, in order to eliminate associations that do not conform to logic. Based on the candidate association pairs that have passed logical verification, a multi-level hierarchical mapping relationship is established according to the preset hierarchical structure; Based on the hierarchical mapping relationship, a full-element semantic association graph containing all element nodes and semantic mapping edges is constructed using a graph structure and stored in the central data warehouse.

6. The method for verifying compliance with institutional document architecture based on semantic mapping as described in claim 5, characterized in that, The method, based on verification rules and a comprehensive semantic association graph, calls multiple verification models to perform multi-dimensional semantic verification on the policy documents, generating a verification result set containing non-compliant items, including: By comparing the semantics of policy documents with the semantics of management needs and architectural blueprints, the completeness and content conflicts of the documents are verified; by comparing the semantics of job responsibilities with the semantics of requirements in related policy documents and analyzing permission settings, the compliance of organizational positions is verified; by comparing the semantics of the functions, data, and processes of the digital platform with the semantics of requirements in policy documents, the synergy of the digital platform is verified; by analyzing the matching degree and consistency between the semantics of policy documents and the semantics of enterprise architectural elements, architectural compliance is verified. By integrating the outputs of various verification models, each identified problem is organized into a structured and standardized verification result set based on related elements, problem type, semantic basis, and quantitative comparison results.

7. The method for verifying compliance with institutional document architecture based on semantic mapping as described in claim 6, characterized in that, The process of risk assessment and classification of non-compliant items in the verification result set according to verification rules, and automatic generation of rectification suggestions and verification reports based on the full-element semantic association graph, includes: For each non-compliant item in the verification result set, a risk assessment is conducted based on a preset multi-dimensional quantitative scoring rule. The risk assessment is based on a weighted calculation of a comprehensive risk value, including multiple dimensions such as the scope of impact, severity, rectification difficulty, and probability of occurrence. The risk level and priority are then classified according to the comprehensive risk value. Based on the aforementioned semantic association graph of all elements, the system documents, organizational positions and digital platform elements associated with non-compliant items are analyzed, and targeted rectification suggestions are automatically generated in combination with the semantic basis of non-compliant items. The suggestions are then broken down into specific implementation steps that include the responsible parties and time limits. Based on the risk assessment results and the generated rectification suggestions, a structured rectification tracking document and a standardized verification report are output.

8. A compliance verification system for institutional document architecture based on semantic mapping, employing the method described in any one of claims 1 to 7, characterized in that, include: The data and rules configuration module is used to obtain raw data and preset structured validation rules from external data sources and store them in the central data warehouse. The document semantic classification module is used to extract document data from the central data warehouse, perform semantic parsing and multi-dimensional classification, generate unique semantic classification labels and store them back; The organization modeling module is used to extract information related to personnel and functions from the central data warehouse, perform structured processing, construct a semantic model containing hierarchical and attribute relationships, generate semantic labels for each element in the semantic model, and then save the data. The semantic graph construction module is used to integrate classified documents and semantic models, combine them with platform metadata, and construct and store a full-element semantic association graph through semantic vectorization and similarity calculation. The intelligent verification engine module is used to perform multi-dimensional semantic verification of institutional documents based on verification rules and the semantic association graph of all elements, and to generate a set of verification results containing non-compliant items. The rectification management module is used to conduct risk assessment and classification of non-compliant items in the verification result set according to the verification rules, and automatically generate rectification suggestions and verification reports by combining the full-element semantic association graph.

9. An electronic device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores computer-executable instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 7.

Images