A Method for Constructing a Standard Ontology of Knowledge Graph

Abstract

The present invention provides a method for constructing a knowledge graph standard ontology, belonging to the technical field of power grid knowledge graph construction. The method includes the following steps: Step 1: Information processing: Process the standard specification documents of transformers and circuit breakers in the multi-source database, including: information extraction, extracting, indexing, producing, and integrating multi-source and heterogeneous data to complete its fragmentation, serialization, and semanticization; information fusion, integrating the extracted information; information processing, classifying and storing the fragmented knowledge generated from the fused information, and after quality evaluation, constructing a standard knowledge base for transformers and circuit breakers; Step 2: Ontology construction: Provide new, automatic, and intelligent methods and tools for knowledge organization, storage, extraction, and reasoning, aggregate a large number of discrete information points into a semantic network, and introducing a mature and stable graph construction function system in the industry can make this link achieve twice the result with half the effort.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid knowledge graph construction, and more specifically, to a method for constructing a standard ontology of a knowledge graph. Background Art

[0002] With the development of technology, the traditional knowledge organization and management methods can no longer meet the current needs of the power system. Currently, knowledge bases based on knowledge representation and knowledge reasoning have been widely applied in the power system, such as intelligent decision-making systems combined with traditional expert systems, fault location systems, and transmission network planning decisions.

[0003] However, most of these knowledge bases rely on the traditional knowledge management method of experts extracting, organizing, and storing data in the form of icons in a database. The knowledge structures they can store are relatively single, and each update requires a large amount of time from professional technical personnel. Especially in fields where knowledge such as power dispatching, equipment management, data interaction, and business query changes rapidly, the existing knowledge management methods have seriously lagged behind the development needs of the system. Summary of the Invention

[0004] To make up for the above deficiencies, the present invention provides a method for constructing a standard ontology of a knowledge graph, aiming to conduct research on the key technologies of a standard knowledge production platform (knowledge base). For the research and development of relevant key technologies in the field of standard digital transformation around processes such as standard structuring, fragmentation, indexing, modeling, knowledge elementization, graphing, and intelligentization, a company standard knowledge base and a main network equipment knowledge graph are constructed.

[0005] The present invention is implemented as follows: A method for constructing a standard ontology of a knowledge graph includes the following steps:

[0006] Step 1: Information processing: Process the standard specification documents of transformers and circuit breakers in a multi-source database, including:

[0007] Information extraction: Extract, index, produce, and integrate multi-source and heterogeneous data to complete its fragmentation, serialization, and semanticization work, and extract or learn entities, attributes, and the mutual relationships between entities from the multi-source database to form an ontological information expression;

[0008] Information fusion: Integrate the extracted information to eliminate contradictions and ambiguities, and produce fragment types including articles, terms, indicators, formulas, pictures, tables, and appendices;

[0009] Information processing: Classify and store the fragmented knowledge generated from the fused information, and after quality assessment, construct a standard knowledge base for transformers and circuit breakers;

[0010] Step 2: Ontology construction: Based on the knowledge base obtained in Step 1, construct a mapping layer for characterizing the mapping relationship between the meta-ontology and ontologies in the standard data. Here, the meta-ontology is a common and essential feature knowledge base extracted from multiple ontologies, used for abstract expression of the ontologies. Determine the meta-ontology model and the mapping function between the meta-ontology and ontologies. According to the mapping function, link each meta-ontology and each ontology correspondingly to construct the mapping layer;

[0011] The determination of the meta-ontology model and the mapping function between the meta-ontology and ontologies includes: determining the meta-ontology model; according to the meta-ontology model, determining the first mapping function for characterizing the mapping from the meta-ontology to the ontologies; according to the mapping function, link each meta-ontology and each ontology correspondingly to construct the mapping layer; the construction of the standard knowledge graph based on the mapping layer includes: based on the mapping layer, constructing the ontology model; extracting corresponding entities from the standard text according to the ontology model to construct the standard ontology of the knowledge graph.

[0012] In a preferred technical solution of the present invention, in Step 1, process no less than 1600 standard specifications of transformers and circuit breakers, including multi-source and heterogeneous data extraction, indexing and production, and integration, complete their fragmentation, serialization, and semanticization work. The output fragment types need to include chapters, terms, indicators, formulas, pictures, tables, appendices, etc. The processing accuracy requirement is above 95%; classify and store the generated fragmentary knowledge, construct the standard knowledge base of transformers and circuit breakers, support functions such as data warehousing, resource management, resource security management, version management, system management, and data storage and backup, and can form a shared service capability for external opening.

[0013] In a preferred technical solution of the present invention, the standard specifications of transformers and circuit breakers include national standards, enterprise standards, industry standards, group standards, technical specifications, operation guides, accident prevention measures, and typical designs.

[0014] In a preferred technical solution of the present invention, in Step 1, the processing standard specifications of transformers and circuit breaker equipment are as follows:

[0015] a. Chapters: Extract and process the chapters at each level, support the association and nesting of chapters, that is, hierarchical processing, and the parent-level chapters can contain sub-level chapters;

[0016] b. Terms: Support the extraction of terms, and the extraction results include term names, term definitions, etc.;

[0017] c. Indicators and indicator values: Extract the indicators in the standard and output them in the form of key-value;

[0018] d. Formulas: Extract the formulas in the standard, which need to include formula names and specific formulas;

[0019] e. Images: Extract non-pure text images within the standard and output image resources, which should include image names and image resource files.

[0020] f. Tables: Extract tables within the standard. The extraction results support two types: extraction as images and Excel tables, and include table names, table headers, row data, column data, etc.

[0021] In a preferred technical solution of the present invention, in step 1, construct a knowledge base model for the technical standards of transformers and circuit breakers:

[0022] a. Construct a knowledge base for the technical standards of transformers and circuit breakers. Starting from the processing and warehousing of resources, store information data in a digital resource library, manage metadata, digital objects, XML data, etc. of the resources, and construct a knowledge base for the technical standards of main network transformers.

[0023] b. Have functions such as data warehousing, resource management, resource security management, version management, system management, data storage, and backup, and abstract and share service capabilities for external access.

[0024] c. The knowledge base for the technical standards of transformers and circuit breakers needs to include multiple sub-libraries such as a standard file sub-library, a term sub-library, a chapter and article sub-library, an index sub-library, an image sub-library, a table sub-library, and a formula sub-library. Each sub-library needs to support adding, deleting, modifying, and querying data, and at the same time, a front-end graphical page is required for users to operate.

[0025] In a preferred technical solution of the present invention, in step 2, the constructed standard ontology of the knowledge graph includes functions for extracting and constructing the knowledge graph of transformer and circuit breaker equipment standards: Knowledge extraction provides knowledge extraction services for different data sources. All knowledge extraction services run periodically in the background in the form of tasks to ensure the continuous access of various external data. Through structured and unstructured data access, automate the construction of the data in the source library into the knowledge graph and provide the ability to input structured data into the graph.

[0026] In a preferred technical solution of the present invention, the ability to input structured data into the graph includes:

[0027] a) Display of the knowledge graph of transformer and circuit breaker equipment standards: Visualize graph data, including entity attributes, query of relationships between entities, query of entity attributes, etc.; classify and statistically analyze the knowledge graph data and visually manage the content of the graph data; support upper-layer applications and provide interfaces for querying entities, attributes, and relationships; support no less than two modeling methods such as lists and visual graphs, and support graphical editing of entities, relationships, and attributes.

[0028] b) Construction and management of knowledge graphs for transformer and circuit breaker equipment standards: support user authority allocation management; support visual management of graph storage; support manual intervention or automatic extraction and addition of graph data (including schema), visual management, and traceability of historical graphs;

[0029] c) Multimodal knowledge understanding of transformer and circuit breaker equipment standards: Supports knowledge extraction from PDF, WORD, TXT and other documents, and constructs knowledge graphs; for different data forms, uses text representation information of structural features as the analysis object, uses mature technical methods in the fields of machine learning, natural language processing, speech recognition, deep learning, etc., combines field problems and practical experience, and combines relevant databases to calculate and select text features of material content;

[0030] d) Other functional requirements are as follows: support model training for entity, attribute and other type extraction; support visualization of platform status, training process and result evaluation; provide basic word segmentation and entity recognition capabilities; support corpus annotation capabilities, users can customize annotation labels, and support annotation of multi-modal data; have a complete knowledge graph construction platform function, with full-stack construction capabilities such as knowledge representation, knowledge modeling, knowledge extraction, knowledge fusion, knowledge storage, and knowledge computing; have a complete knowledge application platform function, with graph-based knowledge retrieval, knowledge question and answer, and online relational reasoning knowledge application capabilities.

[0031] In a preferred technical solution of the present invention, in step 2, the constructed knowledge graph standard ontology also includes graph storage and query functions: supporting the storage, processing and data synchronization update of structured, semi-structured and other data sources in the graph construction process; supporting the relationship management between entities, including adding and deleting edge relationships, setting multiple relationship objects, etc.; realizing entity retrieval, entity relationship calculation, feature query services, etc. of the knowledge graph; supporting the retrieval and display of knowledge graph content through complete standard graph query statements.

[0032] The beneficial effects of the present invention are as follows: the entire project starts with the current situation and needs of digital transformation of standards, i.e., construction of digital standards. First, the top-level design is carried out, and then the standard specifications of digitization, fragmentation and indexation are formulated. At the same time, core key technologies are studied, including standard document digitization-related technologies, data processing and indexing technologies, and intelligent service technologies, etc., and then tool integration and development are carried out to realize the standard knowledge base of digitization and knowledge meta-ization, and finally an intelligent application platform is built to provide scenario-based services for business applications. New, automatic, and intelligent knowledge organization, storage, extraction, and reasoning methods and tools aggregate massive discrete information points into semantic networks, and the introduction of a mature and stable graph construction function system in the industry can make this link more efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and thus should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0034] Figure 1 is the flowchart of the method for constructing the knowledge graph standard ontology provided by the embodiment of the present invention;

[0035] Figure 2 is the structural schematic diagram of the knowledge graph standard ontology construction subsystem provided by the embodiment of the present invention;

[0036] Figure 3 is the structural schematic diagram of the knowledge production subsystem provided by the embodiment of the present invention;

[0037] Figure 4 is the structural schematic diagram of the knowledge graph question and answer system provided by the embodiment of the present invention;

[0038] Figure 5 is the structural schematic diagram of the knowledge production system of the knowledge middle platform provided by the embodiment of the present invention. Specific Embodiments

[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0040] Examples

[0041] Please refer to Figure 1 , the present invention provides a technical solution: a method for constructing a knowledge graph standard ontology, including the following steps:

[0042] Step 1: Information processing: Process the standard specification documents of transformers and circuit breakers in the multi-source database, including:

[0043] Information extraction, extracting, indexing, producing, and integrating multi-source and heterogeneous data, completing its fragmentation, serialization, and semanticization work, extracting or learning entities, attributes, and the mutual relationships between entities from the multi-source database, and forming an ontological information expression;

[0044] Information fusion, integrating the extracted information to eliminate contradictions and ambiguities, and producing fragment types including chapters, terms, indicators, formulas, pictures, tables, and appendices;

[0045] Information processing, classifying and storing the fragmented knowledge generated from the fused information. After quality assessment, a standard knowledge base for transformers and circuit breakers is constructed;

[0046] Specifically, process no less than 1,600 transformer and circuit breaker standards and specifications (including national standards, enterprise standards, industry standards, group standards, technical specifications, operation guides, accident prevention measures, typical designs, etc.), including multi-source and heterogeneous data extraction, indexing, production, and integration, complete their fragmentation, serialization, and semanticization work, and the output fragment types should include chapters, terms, indicators, formulas, pictures, tables, appendices, etc. The processing accuracy rate is required to be above 95%; classify and store the generated fragmented knowledge, construct a standard knowledge base for transformers and circuit breakers, support functions such as data warehousing, resource management, resource security management, version management, system management, data storage, and backup, and can form a shared service capability for external access.

[0047] Among them, the processing standards and specifications for transformer and circuit breaker equipment are as follows: a. Chapters: Extract and process the chapters at each level, support the association and nesting of chapters, that is, hierarchical processing, and the parent-level chapters can contain sub-level chapters.

[0048] b. Terms: Support the extraction of terms, and the extraction results include term names, term definitions, etc.

[0049] c. Indicators and indicator values: Extract the indicators (including text-based clauses and numerical values) in the standard and output them in the key-value form.

[0050] d. Formulas: Extract the formulas in the standard, which should include formula names and specific formulas.

[0051] e. Pictures: Extract non-pure text pictures in the standard and output picture resources, which should include picture names and picture resource files.

[0052] f. Tables: Extract the tables in the standard, and the extraction results support two types of extraction as pictures and excel tables, including table names (if any), table headers (if any), row data, column data, etc.

[0053] (2) Construct a technical standard knowledge base model for transformers and circuit breakers:

[0054] a. Construction of the technical standard knowledge base for transformers and circuit breakers. Starting from the processing and warehousing of resources, data is stored in the digital resource library, and the metadata, digital objects, XML data, etc. of the resources are managed to construct the technical standard knowledge base for main network transformers;

[0055] b. It has functions such as data warehousing, resource management, resource security management, version management, system management, data storage, and backup, and abstracts the shared service capabilities for external access.

[0056] c. The technical standard knowledge base for transformers and circuit breakers needs to include multiple sub-libraries such as the standard document sub-library, term sub-library, chapter and article sub-library, index sub-library, picture sub-library, table sub-library, formula sub-library, etc. Each sub-library needs to support adding, deleting, modifying, and querying data, and at the same time, a front-end graphical page is required for users to operate.

[0057] In addition, purchase the standard knowledge graph construction tool model and the support service for the construction process of the transformer and circuit breaker equipment standard knowledge graph, including the transformer and circuit breaker equipment standard knowledge graph construction and incremental iteration tool model, as well as the whole-process technical support, maintenance, and training in aspects such as solution formulation, knowledge modeling, knowledge extraction and review, knowledge disambiguation, graph construction, and graph application, to assist in completing the graph construction and update. The purchased model and service need to run through the business chain from the standard production to the knowledge application of transformers and circuit breakers.

[0058] The specific construction of the technical standard knowledge base model for transformers and circuit breakers includes:

[0059] a. Construction of the technical standard knowledge base for transformers and circuit breakers. Starting from the processing and warehousing of resources, information data is stored in the digital resource library, and the metadata, digital objects, XML data, etc. of the resources are managed to construct the technical standard knowledge base for main network transformers;

[0060] b. It has functions such as data warehousing, resource management, resource security management, version management, system management, data storage, and backup, and abstracts the shared service capabilities for external access;

[0061] c. The technical standard knowledge base for transformers and circuit breakers needs to include multiple sub-libraries such as the standard document sub-library, term sub-library, chapter and article sub-library, index sub-library, picture sub-library, table sub-library, formula sub-library, etc. Each sub-library needs to support adding, deleting, modifying, and querying data, and at the same time, a front-end graphical page is required for users to operate.

[0062] Step 2: Ontology construction: Based on the knowledge base obtained in Step 1, construct a mapping layer for characterizing the mapping relationship between the meta-ontology and ontologies in the standard data. Here, the meta-ontology is a common and essential feature knowledge base extracted from multiple ontologies, used for abstract expression of the ontologies. Determine the meta-ontology model and the mapping function between the meta-ontology and ontologies. According to the mapping function, link each meta-ontology and each ontology correspondingly to construct the mapping layer;

[0063] The determination of the meta-ontology model and the mapping function between the meta-ontology and ontologies includes: determining the meta-ontology model; according to the meta-ontology model, determining the first mapping function for characterizing the mapping from the meta-ontology to the ontologies; the linking of each meta-ontology and each ontology correspondingly according to the mapping function to construct the mapping layer; the construction of the standard knowledge graph based on the mapping layer includes: constructing the ontology model based on the mapping layer; extracting corresponding entities from the standard text according to the ontology model to construct the standard ontology of the knowledge graph.

[0064] The constructed standard ontology of the knowledge graph includes the functions of knowledge extraction and graph construction for transformers and circuit breaker equipment: Knowledge extraction provides knowledge extraction services for different data sources. All knowledge extraction services run periodically in the background in the form of tasks to ensure the continuous access of various external data. Through data access such as structured and unstructured data, the automated construction from the source library data to the knowledge graph is completed, providing the ability to input structured data into the graph.

[0065] The ability to input structured data into the graph includes:

[0066] a) Display of the standard knowledge graph of transformers and circuit breaker equipment: Visualize graph data, including entity attributes, query of relationships between entities, query of entity attributes, etc.; classification and statistics of knowledge graph data and visualization management of graph data content; support upper-layer applications and provide interfaces for querying entities, attributes, and relationships; support no less than two modeling methods such as lists and visual graphs, and support graphical editing of entities, relationships, and attributes;

[0067] b) Construction management of the standard knowledge graph of transformers and circuit breaker equipment: Support user permission allocation management; support visual management of graph storage; support manual intervention or automatic extraction, addition, visual management, and traceability of historical graphs for graph data (including schema);

[0068] c) Multimodal knowledge understanding of transformer and circuit breaker equipment standards: Supports knowledge extraction from PDF, WORD, TXT and other documents, and constructs knowledge graphs; for different data forms, uses text representation information of structural features as the analysis object, uses mature technical methods in the fields of machine learning, natural language processing, speech recognition, deep learning, etc., combines field problems and practical experience, and combines relevant databases to calculate and select text features of material content;

[0069] d) Other functional requirements are as follows: support model training for entity, attribute and other type extraction; support visualization of platform status, training process and result evaluation; provide basic word segmentation and entity recognition capabilities; support corpus annotation capabilities, users can customize annotation labels, and support annotation of multi-modal data; have a complete knowledge graph construction platform function, with full-stack construction capabilities such as knowledge representation, knowledge modeling, knowledge extraction, knowledge fusion, knowledge storage, and knowledge computing; have a complete knowledge application platform function, with graph-based knowledge retrieval, knowledge question and answer, and online relational reasoning knowledge application capabilities.

[0070] The constructed knowledge graph standard ontology also includes graph storage and query functions: it supports the storage, processing and synchronous update of structured, semi-structured and other data sources in the graph construction process; it supports the relationship management between entities, including adding and deleting edge relationships, setting multiple relationship objects, etc.; it can realize entity retrieval, entity relationship calculation, feature query services, etc. of the knowledge graph; it supports the retrieval and display of knowledge graph content through complete standard graph query statements.

[0071] See also Figure 2 In some specific implementation schemes, the ontology construction subsystem is the skeleton layer of the knowledge graph, which defines the basic structure of knowledge, including entity classes, attribute classes, hierarchical relationships between entity classes, and ownership relationships between entity attributes. The system uses a top-down approach to visually construct the knowledge graph schema, supports low-cost custom addition of field attribute information corresponding to each category, supports presetting a large number of general knowledge graph schemas for system reference, and supports selecting data from the production source database directly to quickly generate the Schema.

[0072] The system supports three ways of creating categories: manual addition, Excel import, and synchronization of structured data structures. The schemas created in the three ways are uniformly stored and managed in the "schema storage and management" module. The system supports the creation of subcategories under categories, and subcategories will automatically inherit the attributes of the parent category, thereby saving administrators time managing categories with subordinate management.

[0073] Form-based modeling, which supports manually adding categories in the way of interactive form operations, adding attributes to this category, adding attribute types and constraints, adding relationships, and adding relationship types and constraints.

[0074] Mapping-based modeling, which supports synchronizing the data structures of structured data and directly generating the targets of knowledge modeling in a fast mapping way.

[0075] The management and display module provides unified reference, query, and modification interfaces for the schema constructed by the system externally.

[0076] Supports defining complex schemas, including nested expressions of attribute values and defining attributes on edge relationships.

[0077] Please refer to Figure 3 , the knowledge production subsystem, whose input is various forms of raw data introduced by the data access subsystem and the knowledge production goals defined by the ontology construction subsystem, and whose output is graph knowledge.

[0078] In some specific implementation schemes, the knowledge production subsystem provides a basic offline data processing architecture and corresponding support mechanisms. Each type of knowledge production task can be abstracted into two parts: 1) the support of the unified data processing architecture; 2) a series of strategies or algorithms related to specific knowledge types. The knowledge production subsystem provides unified distributed file storage, distributed state storage, distributed result storage, distributed caching, as well as computing scheduling capabilities, batch processing capabilities, streaming processing capabilities, and heterogeneous computing capabilities for these knowledge production strategies or algorithms.

[0079] The graph knowledge production module converts structured and unstructured data into knowledge graph data and establishes relationships between entities and entities. Specific functions include: supporting the identification of entities, relationships, and attributes from free text, and optimizing the accuracy of free text extraction by manually intervening in the model; supporting directly converting data from structured data sources, mapping and aligning with the knowledge graph schema, and automatically generating knowledge graph data; supporting customizing knowledge graph extraction models, including tuning models, optimizing vocabulary, and defining templates; supporting machine learning models, machine rules, and manual methods for entity, attribute, and relationship mapping, cleaning, fusion, normalization, edge building, and completion; the entire process of graph knowledge production supports visualization, white-boxing, and audit intervention.

[0080] The graph knowledge production module relies on the knowledge production subsystem to provide overall architecture support, and relies on the model strategy hosting system to provide overall algorithm training, execution, and prediction capabilities, and completes the serial execution of four major sub-modules: knowledge extraction, knowledge processing, knowledge fusion, and knowledge association.

[0081] Knowledge processing involves performing schema-based attribute mapping on the results of knowledge extraction to make the extracted attribute names conform to the equivalent attributes defined in the schema, and cleaning the attribute values of knowledge extraction based on regular expressions to make the extracted attribute values conform to the attribute constraint conditions defined in the schema.

[0082] Knowledge fusion: Knowledge graph data often comes from multiple sources. The same entities extracted from different sources need to be unified and disambiguated at the instance level, and the same-named attributes extracted from different sources need to be uniformly optimized for attributes at the instance level. The disambiguation strategies of this module mainly implement three types of attribute comparison algorithms: text similarity comparison, semantic similarity comparison, and comparison of various types of attribute values (such as addresses, phone numbers, dates, numerical units). At the upper layer, the bayes model, XGBoost model, and XGRank model are used to perform machine learning and fitting scoring on the results of various comparison algorithms.

[0083] Knowledge association: The attributes defined in the knowledge graph schema definition stage are relationship types, and edge building needs to be performed in the final stage of graph knowledge production. This module implements an edge building strategy based on rule configuration. Users can determine whether two entities should be connected by an edge based on attributes of types such as strings and numerical values. For an already established graph, two methods, rule configuration and knowledge representation learning inference, are supported to complement and discover the potential relationships between the entities in the current knowledge graph.

[0084] Knowledge storage: The knowledge storage subsystem includes a graph storage engine and a text storage engine. The graph storage engine constructs a super-large-scale, high-performance distributed graph index and storage engine. It supports common graph models such as Property Graph and a Turing-complete graph query language similar to Gremlin, provides a native graph storage engine, supports multiple storage media / systems in the storage architecture, such as memory or direct SSD, has distributed storage capabilities, meets the storage requirements of massive graph data, and has multi-active instances and fast failover to achieve high availability of services. The text knowledge storage engine integrates elasticsearch that has undergone in-depth effect and performance optimization, provides a storage and retrieval system for large-scale text data, the system capacity is scalable, and a series of optimizable configurations are provided.

[0085] Graph storage engine: The graph database BGraph is a high-performance commercial graph database independently developed by Baidu and is suitable for application scenarios where data is highly correlated and in-depth analysis is required. The core of BGraph is a high-performance graph database engine that has been applied and practiced in Baidu's knowledge graph system for many years. It can support hundreds of millions of entities and has a response latency in milliseconds, providing distributed and high-availability capabilities to meet the needs of enterprise-level applications.

[0086] Please refer to Figure 4, Knowledge Q&A, Knowledge Graph Question Answering (KB-QA) is one of the most important applications based on the knowledge graph. It refers to a knowledge base oriented to the knowledge graph. By inputting a natural language question, through semantic understanding and parsing of the question, it automatically finds the answer from the knowledge graph through querying, calculating, and reasoning, directly meeting the needs of users.

[0087] Considering the knowledge storage forms such as SQL databases and tables within enterprises, the system has been further abstracted and integrated in the technical solution to meet the Q&A scenarios of common structured data knowledge bases in enterprises.

[0088] Please refer to Figure 5 , the knowledge production system of the knowledge middle platform is the core support connecting the underlying data middle platform (data governance, computing, and storage) and the upper-layer application service platform. It mainly realizes functions such as knowledge-based expression, extraction, construction, and management of massive multi-source heterogeneous data, providing knowledge data production support and storage management support for applications such as semantic search, knowledge Q&A, computing analysis, and reasoning decision-making.

[0089] The knowledge production system of the knowledge middle platform consists of various core subsystems and support guarantee subsystems that run through the entire knowledge life cycle.

[0090] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for constructing a standard ontology of a knowledge graph, characterized in that It includes the following steps: Step 1: Information processing: Process the standard specification documents of transformers and circuit breakers in the multi-source database, including: Information extraction: Extract, index, produce, and integrate multi-source and heterogeneous data to complete its fragmentation, serialization, and semanticization. Extract or learn entities, attributes, and the mutual relationships between entities from the multi-source database to form an ontological information expression; Information fusion: Integrate the extracted information to eliminate contradictions and ambiguities, and output fragment types including articles, terms, indicators, formulas, pictures, tables, and appendices; Information processing: Classify and store the fragmented knowledge generated from the fused information. After quality evaluation, construct a standard knowledge base for transformers and circuit breakers; Among them, the standard specification documents are as follows: a. Articles: Extract and process articles at each level, support the association and nesting of articles, that is, hierarchical processing, and the parent-level articles can contain sub-level articles; b. Terms: Support the extraction of terms, and the extraction results include term names and term definitions; c. Indicators and indicator values: Extract the indicators in the standard and output them in the form of key-value; d. Formulas: Extract the formulas in the standard, which need to include formula names and specific formulas; e. Pictures: Extract non-pure text pictures in the standard and output picture resources, which need to include picture names and picture resource files; f. Tables: Extract the tables in the standard, and the extraction results support being extracted into two types: pictures and excel tables, including table names, table headers, row data, and column data; Step 2: Ontology construction: Based on the knowledge base obtained in Step 1, construct a mapping layer for characterizing the mapping relationship between the meta-ontology and the ontology in the standard data. Among them, the meta-ontology is a common and essential feature knowledge base extracted from multiple ontologies, used for abstract expression of the ontology. Determine the meta-ontology model and the mapping function between the meta-ontology and the ontology. According to the mapping function, link each meta-ontology and each ontology correspondingly to construct the mapping layer; The determination of the meta-ontology model and the mapping function between the meta-ontology and the ontology includes: determining the meta-ontology model; according to the meta-ontology model, determining the first mapping function for characterizing the mapping from the meta-ontology to the ontology; according to the mapping function, link each meta-ontology and each ontology correspondingly to construct the mapping layer; based on the mapping layer, construct a standard knowledge graph, including: based on the mapping layer, construct an ontology model; according to the ontology model, extract corresponding entities from the standard text to construct the standard ontology of the knowledge graph.

2. The method for constructing a knowledge graph standard ontology according to claim 1, wherein In Step 1, process no less than 1,600 standards and specifications for transformers and circuit breakers, including extraction, indexing, production, and integration of multi-source and heterogeneous data, and complete their fragmentation, serialization, and semanticization. The output fragment types shall include chapters and articles, terms, indicators, formulas, pictures, tables, and appendices. The processing accuracy rate shall be above 95%. Classify and store the generated fragmented knowledge, build a standards knowledge base for transformers and circuit breakers, support functions such as data warehousing, resource management, resource security management, version management, system management, data storage, and backup, and form a shared service capability for external access.

3. The method for constructing a knowledge graph standard ontology according to claim 2, wherein The standards and specifications for transformers and circuit breakers include national standards, enterprise standards, industry standards, group standards, technical specifications, operation guides, accident prevention measures, and typical designs.

4. The method for constructing a knowledge graph standard ontology according to claim 1, characterized in that In Step 1, build a knowledge base model for the technical standards of transformers and circuit breaker equipment: a. Build a knowledge base for the technical standards of transformers and circuit breaker equipment. Starting from the processing and warehousing of resources, store information data in the digital resource library, manage the metadata, digital objects, and XML data of resources, and build a knowledge base for the technical standards of main network transformers. b. Have functions such as data warehousing, resource management, resource security management, version management, system management, data storage, and backup, and abstract a shared service capability for external access. c. The knowledge base for the technical standards of transformers and circuit breaker equipment shall include multiple sub-libraries such as a standard document sub-library, a term sub-library, a chapter and article sub-library, an indicator sub-library, a picture sub-library, a table sub-library, and a formula sub-library. Each sub-library shall support adding, deleting, modifying, and querying data, and at the same time, a front-end graphical page shall be provided for users to operate.

5. The method for constructing a knowledge graph standard ontology according to claim 1, wherein In Step 2, the constructed standard ontology of the knowledge graph includes functions for extracting and constructing the standard knowledge of transformers and circuit breaker equipment: Knowledge extraction provides knowledge extraction services for different data sources. All knowledge extraction services run periodically in the background to ensure the continuous access of various external data. Through structured and unstructured data access, automate the construction of the source library data into the knowledge graph and provide the ability to input structured data into the graph.

6. The method for constructing a knowledge graph standard ontology according to claim 5, wherein The ability to input structured data into the graph includes: a) Display of the standard knowledge graph of transformers and circuit breaker equipment: Visualize the graph data, including entity attributes, query of relationships between entities, and query of entity attributes; classify and statistically analyze the knowledge graph data and visually manage the content of the graph data; support upper-layer applications and provide interfaces for querying entities, attributes, and relationships; support no less than two modeling methods such as list and visualization graph, and support graphical editing of entities, relationships, and attributes. b) Construction and management of the standard knowledge graph of transformers and circuit breaker equipment: Support user permission allocation management; support visual management of graph storage; support manual intervention or automatic extraction and addition of graph data, visual management, and traceability of historical graphs. c) Multimodal knowledge understanding of transformer and circuit breaker equipment standards: Supports knowledge extraction from PDF, WORD, and TXT documents, and builds knowledge graphs; for different data forms, uses text representation information of structural features as the analysis object, uses mature technical methods in the fields of machine learning, natural language processing, speech recognition, and deep learning, combines domain problems and practical experience, and combines relevant databases to calculate and select text features of material content; d) Other functional requirements are as follows: support model training for entity and attribute type extraction; support visualization of platform status, training process and result evaluation; provide basic word segmentation and entity recognition capabilities; support corpus annotation capabilities, users can customize annotation labels, and support annotation of multi-modal data; have a complete knowledge graph construction platform function, with full-stack construction capabilities for knowledge representation, knowledge modeling, knowledge extraction, knowledge fusion, knowledge storage, and knowledge computing; have a complete knowledge application platform function, with knowledge application capabilities based on graph knowledge retrieval, knowledge question and answer, and online relational reasoning.

7. The method for constructing a knowledge graph standard ontology according to claim 1, wherein In step 2, the knowledge graph standard ontology constructed also includes graph storage and query functions: supporting the storage, processing and data synchronization update of structured and semi-structured data sources in the graph construction process; supporting the relationship management between entities, including adding and deleting edge relationships and setting multiple relationship objects; and realizing entity retrieval, entity relationship calculation and feature query services of knowledge graphs; Supports retrieving and displaying knowledge graph content through complete standard graph query statements.

Images