Text knowledge automatic extraction and mapping method and system based on deep learning

Abstract

The invention discloses a text knowledge automatic extraction and mapping method based on deep learning, which optimizes a semantic understanding algorithm by combining a deep learning model, enhances the perception ability of the model for the internal structure and external relationship of a text, realizes comprehensive extraction and detailed relationship analysis of equipment fault knowledge, and utilizes a knowledge mapping technology to realize automatic extraction and mapping of equipment fault knowledge. According to the method, an automatic graph construction process is implemented to improve the query efficiency and analysis capability of the graph, and the method comprises the following steps: S1, performing structured processing on an original document, segmenting the original document into independent case units, and extracting key information related to an equipment fault; s2, key entities in the document and mutual relations of the key entities are recognized and marked; and S3, based on the extracted entities and relationships, constructing and updating a knowledge graph of the fault case. The system has high automation, intellectualization, expandability and modular design, can be easily integrated into an existing industrial informatization system, and assists an enterprise in realizing intellectualized transformation and upgrading.

Description

Technical Field

[0001] The present invention belongs to the intersection of artificial intelligence and industrial information technology, and specifically relates to a method for automatic extraction and graphing of text knowledge based on deep learning. Background Art

[0002] With the continuous deepening of industrialization and the extensive application of mechanical equipment, the stable operation of equipment has become a key factor in the sustainable development of enterprises. However, in the actual production process, equipment failures occur frequently, which not only affects the production progress, but also increases the maintenance costs of enterprises. Traditional equipment failure analysis methods often rely on manual experience and paper records, and there are problems such as scattered data, time-consuming sorting, and inaccurate analysis. With the rapid development of technologies such as big data and artificial intelligence, the industrial Internet and intelligent manufacturing have gradually emerged, and enterprises have put forward higher requirements for the intelligent level of equipment maintenance management. How to use these advanced technologies to realize the automatic extraction and efficient use of equipment failure knowledge has become a problem that enterprises need to solve urgently.

[0003] When the inventors conducted in-depth research on natural language processing (NLP) and knowledge graph construction technology in the field of industrial equipment failure analysis, they found that the application of existing technologies in this field has significant defects. Traditional methods often only focus on the surface information of the text, such as keyword matching or simple syntactic analysis, but ignore the deep semantic structure and contextual relationship of the text. This processing method results in a significant reduction in accuracy and comprehensiveness when extracting key information such as equipment name, fault phenomenon, fault cause and treatment measures. The technical difficulty lies in how to build a model that can deeply understand the semantics of the text and accurately capture complex contexts and implicit relationships.

[0004] Further research shows that the semantic understanding ability of existing technologies is limited by algorithm design and model structure. Although some methods use machine learning or statistical methods, they lack the ability of deep neural networks to extract deep features of text, making it difficult to accurately understand the complex semantic relationships in equipment fault descriptions. Therefore, another technical difficulty lies in how to design a deep learning model that can automatically learn high-level features of text and achieve effective semantic understanding in different fields and contexts.

[0005] In addition, in practical applications, the inventors found that traditional methods have low automation and poor scalability when processing large-scale and diverse text data. This is mainly because traditional methods are highly dependent on manual intervention and rule definition, and are difficult to adapt to rapidly changing data environments. Therefore, the third technical difficulty lies in how to achieve a highly automated text processing process that can automatically adapt to text data of different sizes and types while maintaining efficiency and accuracy. Summary of the invention

[0006] The technical problem to be solved by the present invention is that the text knowledge processing efficiency is low and the accuracy is insufficient in the field of industrial equipment fault analysis. To this end, the present invention proposes a method for automatic extraction and graphing of text knowledge based on deep learning, which includes the following steps:

[0007] Step S1, document analysis and information extraction: Use the deep learning model to structure the original document, segment it into independent case units, identify key paragraphs, and automatically extract text knowledge related to equipment failures, laying the foundation for knowledge graph construction;

[0008] Step S2, entity relationship extraction: through named entity recognition (NER) and relationship extraction, identify and annotate key entities and their relationships in the document to provide structured basic data for knowledge graph construction;

[0009] Step S3, knowledge graph construction: Based on the entities and relationships extracted in steps S1 and S2, build and update the knowledge graph of fault cases to ensure the accuracy and scalability of the graph.

[0010] Furthermore, the deep learning model is pre-trained on a large-scale general corpus, and is further trained on a dedicated training data set containing professional terms related to water conservancy equipment maintenance, complex technical descriptions, and specific document structures, so as to improve the model's semantic understanding and information extraction capabilities in the field of water conservancy equipment maintenance.

[0011] Furthermore, the deep learning model is pre-trained on a large-scale general corpus and is further trained on a special corpus containing relevant data in the field of industrial equipment failure analysis, so that it is particularly suitable for text processing in the field of industrial equipment failure analysis, and can accurately identify and extract key information such as equipment name, fault phenomenon, fault cause and treatment measures.

[0012] Furthermore, step S1 includes the following steps:

[0013] Step S101, document reading and segmentation: read the original document, use regular expressions to match the title tags and paragraph separators in the document, segment the document into multiple units, each unit corresponds to a case, and further analyze the content coherence and contextual relationship, adjust the segmentation boundaries, and ensure the independence and integrity of each unit;

[0014] Step S102, keyword identification: using natural language processing (NLP) technology based on a deep learning model to segment and tag the text, and construct a keyword dictionary containing "fault phenomenon", "fault cause analysis", "fault solution", "fault solution effect" and their synonyms and phrase variants related to equipment failure;

[0015] Step S103, paragraph marking and extraction: by text matching and semantic analysis, paragraphs or sentences containing keywords or related semantics are identified in the document, and these key paragraphs are marked and extracted to prepare for subsequent structured information extraction;

[0016] Step S104, structured extraction: based on keyword tags, automatically extract key information of each case including title, fault phenomenon description, cause analysis, solution and solution effect evaluation.

[0017] Further, step S2 includes the following steps:

[0018] Step S201, named entity recognition: using the NER model in NLP, perform entity recognition on the text in the document, define entity types, including equipment, fault phenomenon, fault cause and treatment measures, and train or adjust the NER model to identify these types of entities, perform entity annotation on the document, and output text containing entity types and their boundaries;

[0019] Step S202, relationship extraction: design relationship extraction templates or rules, define relationship types based on entity types and context information, including "occurrence", "attribution" and "recommendation" and other relevant phrases that define relationship types, identify and extract relationships between entities, construct relationship triples (entity 1-relationship-entity 2) and store them.

[0020] Further, step S3 includes the following steps:

[0021] Step S301, node and edge creation: the extracted entities are used as nodes, each node contains entity ID, type and attributes, the extracted relationships are used as edges, the corresponding entity nodes are connected, each edge contains attributes such as relationship type and direction, and a knowledge graph of fault cases is constructed, and each case is regarded as a subgraph in the graph;

[0022] Step S302, graph fusion: Check the existing knowledge graph, identify the nodes and edges that match the entities and relationships in the newly extracted cases, merge or update the matching nodes and edges to ensure the accuracy and consistency of the graph, and add the newly emerging entities and relationships to the knowledge graph to expand the coverage and depth of the graph.

[0023] Furthermore, the entities and relations extracted in steps S1 and S2 support manual review, quality assessment of the extracted fault cases, and correction or elimination of cases that do not meet the quality standards to ensure the high quality of the final data set.

[0024] Furthermore, in step S3, after extracting key information, the information is represented and stored in the form of a graph using knowledge graph technology, and in the process of constructing the knowledge graph, a graph database management system with high query speed and large-scale data processing capabilities is used to support the organization and management of equipment fault knowledge and ensure the storage and query performance of the graph.

[0025] Furthermore, the method supports parallel processing of large amounts of documents and complex data through a distributed computing architecture or multi-threaded processing, has good scalability and adaptability, is suitable for efficiently completing the extraction of equipment failure-related information and the construction of a knowledge graph in a short period of time, and is adaptable to water conservancy equipment maintenance documents of different sizes and a variety of data types.

[0026] Furthermore, the present invention adopts the following technical solutions:

[0027] A text knowledge automatic extraction and graphing system based on deep learning. The system adopts a modular design and includes the following independent modules:

[0028] Semantic analysis module: used to perform the semantic analysis function of the deep learning model. This module uses the deep learning model trained on a large-scale general corpus and a special domain corpus to perform deep semantic understanding of the input text and identify key information in the text, such as device name, fault phenomenon, fault cause, and treatment measures.

[0029] Information extraction module: Based on the output of the semantic analysis module, it automatically extracts key information related to equipment failure and generates a data structure for subsequent knowledge graph construction. This module supports multi-threaded processing to improve information extraction efficiency and adapts to multiple data formats;

[0030] Knowledge graph construction module: Using knowledge graph technology, the data generated by the information extraction module is represented and stored in the form of a graph. This module uses a graph database management system that supports memory indexing and distributed storage technology, which is suitable for efficient query and processing of large-scale data, ensuring the storage and query performance of the knowledge graph;

[0031] Export module: supports exporting the results of knowledge extraction and graph construction into multiple formats to meet users' diverse data needs;

[0032] Integration interface module: provides an integration interface with existing industrial information systems, supports modular configuration and deployment, and enables the system to be flexibly integrated with existing systems.

[0033] Compared with the prior art, the advantages of the present invention are:

[0034] 1. Saving of raw materials and costs: By introducing deep learning and automated processing technology, the present invention can automatically extract key information from documents, significantly reducing the workload of manual reading, annotation and organization of documents, thereby saving human resources and reducing cost losses caused by human errors.

[0035] 2. Batch processing capability: The present invention supports batch processing of a large number of documents, and can complete the extraction and graph construction of a large amount of information in a short time, thereby improving overall work efficiency.

[0036] 3. The embodiment of better performance:

[0037] (1) Intelligent analysis and decision support: The knowledge graph constructed not only intuitively displays the correlation between equipment failures, but also supports complex query operations. This enables users to have a deeper understanding of the causes and trends of equipment failures, providing strong support for the formulation of more scientific and reasonable maintenance strategies and preventive maintenance plans.

[0038] (2) Scalability and adaptability: The methods and techniques of the present invention are highly scalable and adaptable, and can easily handle maintenance documents of water conservancy equipment of different sizes. This helps to maintain the advancement and long-term effectiveness of the technology.

[0039] 4. Modular design: The present invention separates the semantic analysis, information extraction and knowledge graph construction functions into independent modules. This design not only simplifies the deployment and configuration process of each module, but also allows each module to be updated and optimized independently, thereby improving the flexibility and scalability of the system. Since these modules can be individually adjusted and integrated according to specific needs, enterprises can more easily integrate the present invention with existing industrial information systems, which helps to quickly realize intelligent transformation and upgrading and reduce implementation costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A schematic diagram of the process flow of the text knowledge automatic extraction and graphing method provided by the present invention.

[0041] Figure 2 An example flowchart of the operation steps of a method for automatic extraction and graphing of text knowledge provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the following will be combined with the drawings in the embodiments of the present invention, and first explain several key terms involved in this application, and then clearly and completely describe the technical scheme in the embodiments of the present invention. It should be understood that the embodiments described below are only part of the embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0043] A method for automatic extraction and graphing of text knowledge based on deep learning, comprising the following steps:

[0044] Step S1, document analysis and information extraction: Use the deep learning model to structure the original document, segment it into independent case units, identify key paragraphs, and automatically extract text knowledge related to equipment failures, laying the foundation for knowledge graph construction. Step S1 specifically includes the following sub-steps:

[0045] Step S101, document reading and segmentation: read the original document, use regular expressions to match the title tags and paragraph separators in the document, segment the document into multiple units, each unit corresponds to a case, and further analyze the content coherence and contextual relationship, adjust the segmentation boundaries, and ensure the independence and integrity of each unit;

[0046] Step S102, keyword identification: using natural language processing (NLP) technology based on a deep learning model to segment and tag the text, and construct a keyword dictionary containing "fault phenomenon", "fault cause analysis", "fault solution", "fault solution effect" and their synonyms and phrase variants related to equipment failure;

[0047] Step S103, paragraph marking and extraction: by text matching and semantic analysis, paragraphs or sentences containing keywords or related semantics are identified in the document, and these key paragraphs are marked and extracted to prepare for subsequent structured information extraction;

[0048] Step S104, structured extraction: based on keyword tags, automatically extract key information of each case including title, fault phenomenon description, cause analysis, solution and solution effect evaluation.

[0049] Step S2, entity relationship extraction: Through named entity recognition (NER) and relationship extraction, key entities in the document and their relationships are identified and annotated to provide structured basic data for knowledge graph construction. Step S2 specifically includes the following sub-steps:

[0050] Step S201, named entity recognition: using the NER model in NLP, perform entity recognition on the text in the document, define entity types, including equipment, fault phenomenon, fault cause and treatment measures, and train or adjust the NER model to identify these types of entities, perform entity annotation on the document, and output text containing entity types and their boundaries;

[0051] Step S202, relationship extraction: design relationship extraction templates or rules, define relationship types based on entity types and context information, including "occurrence", "attribution" and "recommendation" and other relevant phrases that define relationship types, identify and extract relationships between entities, construct relationship triples (entity 1-relationship-entity 2) and store them.

[0052] Step S3, knowledge graph construction: Based on the entities and relationships extracted in steps S1 and S2, construct and update the knowledge graph of fault cases to ensure the accuracy and scalability of the graph. Step S3 specifically includes the following sub-steps:

[0053] Step S301, node and edge creation: the extracted entities are used as nodes, each node contains entity ID, type and attributes, the extracted relationships are used as edges, the corresponding entity nodes are connected, each edge contains attributes such as relationship type and direction, and a knowledge graph of fault cases is constructed, and each case is regarded as a subgraph in the graph;

[0054] Step S302, graph fusion: Check the existing knowledge graph, identify the nodes and edges that match the entities and relationships in the newly extracted cases, merge or update the matching nodes and edges to ensure the accuracy and consistency of the graph, and add the newly emerging entities and relationships to the knowledge graph to expand the coverage and depth of the graph.

[0055] Step S4, quality assessment: Support manual review, conduct quality assessment on the extracted fault cases, correct or eliminate unqualified cases, and ensure the high quality of the final data set.

[0056] Reference below Figure 2 The operation steps of the embodiment of the present invention are introduced.

[0057] Figure 2The operational flow chart of the intelligent platform based on the text knowledge automatic extraction and graphing system of the present invention is shown. The intelligent platform is a preferred embodiment of the present invention. Users can easily fill in the name of the extraction project, upload attachments containing multiple fault cases, and fill in the relevant descriptions in the platform. Inside the platform, advanced deep learning models including semantic analysis modules are first pre-trained on a large-scale general corpus to establish basic semantic understanding capabilities for a wide range of corpora. Subsequently, the model is further optimized and trained on a special training data set, which covers professional terms and complex technical descriptions in the field of water conservancy equipment maintenance, as well as relevant data in the field of industrial equipment failure analysis. Through this process, the semantic analysis module can accurately identify and extract key information including equipment name, fault phenomenon, fault cause and treatment measures, make full use of the contextual semantic associations of the text, and ensure the accuracy and comprehensiveness of the extraction results.

[0058] After the semantic analysis is completed, the information extraction module automatically generates a data structure for knowledge graph construction. The knowledge graph construction module of the preferred embodiment uses knowledge graph technology to intuitively represent and store the extracted key information in the form of a graph, thereby ensuring the systematic organization and management of equipment failure knowledge. This knowledge graph not only facilitates users to quickly understand the overall situation of equipment failures, but also provides strong support for subsequent complex relationship queries and analyses. In the process of constructing a knowledge graph, the present invention adopts an optimized graph database management system that supports memory indexing and distributed storage technology to achieve efficient query and has the ability to process large-scale data, thereby significantly improving the storage and query performance of the graph. In addition, the platform performs tasks such as document parsing, information extraction, and knowledge graph construction in parallel through multi-threaded processing, thereby significantly reducing processing time and improving the overall efficiency of the system.

[0059] The export module of this preferred embodiment also supports exporting the extraction results to TXT or EXCEL format to meet the diverse data needs of users. The modular design of the system ensures its high degree of automation, intelligence and scalability. By continuously optimizing the deep learning model, the system can continuously improve the accuracy and efficiency of knowledge extraction as the amount of data increases, and can be easily integrated into the existing industrial information system to help enterprises achieve intelligent transformation and upgrading.

[0060] In the specific operation process, the user first logs in to the intelligent platform. On the main interface of the platform, the user clicks the directory and clicks the "Add" button to enter the project initialization page.

[0061] On this page, users need to fill in the name of the project, which should concisely describe the fault case, such as "XX fault case extraction".

[0062] Optionally, the user can also enter a brief description of the fault case.

[0063] The user uploads an attachment containing one or more fault cases (in PDF, WORD, or TXT format), each of which should include a description of the fault phenomenon and keywords such as "fault cause analysis," "fault solution," and "fault resolution effect." The platform preprocesses the uploaded attachments, including text extraction and word segmentation, to prepare for subsequent knowledge extraction.

[0064] After uploading and confirming that the attachment is correct, the user clicks the "Extract" button to start the knowledge extraction process.

[0065] The platform uses the built model to extract knowledge from the preprocessed text. The model automatically extracts key information such as device name, fault phenomenon, fault cause and treatment measures by identifying keywords in the text, analyzing sentence structure and contextual semantic associations, and stores the extraction results in a structured form.

[0066] After the extraction is completed, the platform provides a knowledge review interface for users to proofread the extraction results. The interface supports comparison between the original text and the extraction results.

[0067] Users can review the extraction results of each case one by one to confirm its accuracy and completeness.

[0068] If necessary, users can also manually add or delete erroneous entries to improve the extraction results.

[0069] After the review is passed, users can choose to import the results into the graph. The platform uses knowledge graph technology to represent and store the extracted entities and relationships in the form of a graph, and generates a visual graph for users to view.

[0070] Among them, the equipment name is used as a node (entity), and the fault phenomenon, fault cause and treatment measures are used as edges (relationships) to connect each node to form an intuitive knowledge network.

[0071] In the graph, users can clearly see the correlation between various fault cases and how they together form a complete fault knowledge network.

[0072] At the same time, in addition to importing graphs, users can choose to export the extracted and reviewed knowledge in different formats. On the "Knowledge Export" page, users can choose to export in TXT or EXCEL format.

[0073] If you select TXT or EXCEL format, the platform will generate the corresponding file for users to download.

[0074] In the knowledge export interface, the platform automatically generates a bar chart to count the number of entities and relationships, as well as a ring chart to count the number of reviewed and unreviewed cases.

[0075] The core advantage of the present invention lies in its high degree of automation, intelligence and scalability. By continuously optimizing the deep learning model, the present invention can cope with the processing needs of large-scale text data, and continuously improve the accuracy and efficiency of knowledge extraction as the amount of data increases. At the same time, the modular design makes the present invention easy to integrate into the existing industrial information system, helping enterprises to achieve intelligent transformation and upgrading.

[0076] It should be emphasized that the specific implementation of the present invention is not limited to the above description, and professionals can make appropriate adjustments and extensions according to actual needs. For example, a more complex deep learning model can be introduced to improve the extraction effect, or a new functional module can be added to meet the application requirements in a specific scenario. In addition, terms such as "first", "second", etc. in the present invention are only used to distinguish different entities or operations, and do not require or imply that there is an actual order or relationship between these entities or operations.

[0077] Finally, the methods, devices and systems of the present invention are all innovative achievements in the field, and their protection scope shall be based on the claims. Any modification, equivalent replacement or improvement of the present invention, as long as it falls within the spirit and principle of the present invention, shall be deemed to be within the protection scope of the present invention. In summary, the content of this specification should not be understood as a limitation of the present invention, but a comprehensive explanation of the technical ideas of the present invention.

Claims

1. A method for automatic extraction and graphing of text knowledge based on deep learning, characterized by: The following steps are involved: Step S1, document analysis and information extraction: Use the deep learning model to structure the original document, segment it into independent case units, identify key paragraphs, and automatically extract text knowledge related to equipment failures, laying the foundation for knowledge graph construction; Step S2, entity relationship extraction: through named entity recognition (NER) and relationship extraction, identify and annotate key entities and their relationships in the document to provide structured basic data for knowledge graph construction; Step S3, knowledge graph construction: Based on the entities and relationships extracted in steps S1 and S2, build and update the knowledge graph of fault cases to ensure the accuracy and scalability of the graph.

2. The method for automatic extraction and graphing of text knowledge based on deep learning according to claim 1 is characterized in that: The deep learning model is pre-trained on a large-scale general corpus and is further trained on a dedicated training dataset containing professional terms related to water conservancy equipment maintenance, complex technical descriptions, and specific document structures to improve the model's semantic understanding and information extraction capabilities in the field of water conservancy equipment maintenance.

3. The method for automatic extraction and graphing of text knowledge based on deep learning according to claim 1 is characterized in that: The deep learning model is pre-trained on a large-scale general corpus and is further trained on a special corpus containing relevant data in the field of industrial equipment failure analysis, so that it is particularly suitable for text processing in the field of industrial equipment failure analysis, and can accurately identify and extract key information such as equipment name, failure phenomenon, failure cause and treatment measures.

4. The method for automatic extraction and graphing of text knowledge based on deep learning according to claim 1 is characterized in that: Step S1 includes the following steps: Step S101, document reading and segmentation: read the original document, use regular expressions to match the title tags and paragraph separators in the document, segment the document into multiple units, each unit corresponds to a case, and further analyze the content coherence and contextual relationship, adjust the segmentation boundaries, and ensure the independence and integrity of each unit; Step S102, keyword identification: natural language processing (NLP) technology based on a deep learning model is used to segment and tag the text, and a keyword dictionary containing "fault phenomenon", "fault cause analysis", "fault solution", "fault resolution effect" and their synonyms and phrase variants related to equipment failure is constructed; Step S103, paragraph marking and extraction: by text matching and semantic analysis, paragraphs or sentences containing keywords or related semantics are identified in the document, and these key paragraphs are marked and extracted to prepare for subsequent structured information extraction; Step S104, structured extraction: based on keyword tags, automatically extract key information of each case including title, fault phenomenon description, cause analysis, solution and solution effect evaluation.

5. The method for automatic extraction and graphing of text knowledge based on deep learning according to claim 1 is characterized in that: Step S2 includes the following steps: Step S201, named entity recognition: using the NER model in NLP, perform entity recognition on the text in the document, define entity types, including equipment, fault phenomenon, fault cause and treatment measures, and train or adjust the NER model to identify these types of entities, perform entity annotation on the document, and output text containing entity types and their boundaries; Step S202, relationship extraction: design relationship extraction templates or rules, define relationship types based on entity types and context information, including "occurrence", "attribution" and "recommendation" and other relevant phrases that define relationship types, identify and extract relationships between entities, construct relationship triples, and store them as a combination of entity 1-relationship-entity 2.

6. The method for automatic extraction and graphing of text knowledge based on deep learning according to claim 1 is characterized in that: Step S3 includes the following steps: Step S301, node and edge creation: the extracted entities are used as nodes, each node contains entity ID, type and attributes, the extracted relationships are used as edges, the corresponding entity nodes are connected, each edge contains attributes such as relationship type and direction, and a knowledge graph of fault cases is constructed, and each case is regarded as a subgraph in the graph; Step S302, graph fusion: Check the existing knowledge graph, identify the nodes and edges that match the entities and relationships in the newly extracted cases, merge or update the matching nodes and edges to ensure the accuracy and consistency of the graph, and add the newly emerging entities and relationships to the knowledge graph to expand the coverage and depth of the graph.

7. The method for automatic extraction and graphing of textual knowledge based on deep learning according to claim 1, characterized in that: The entities and relationships extracted in steps S1 and S2 support manual review, quality assessment of the extracted fault cases, and correction or elimination of cases that do not meet the quality standards to ensure the high quality of the final data set.

8. The method for automatic extraction and graphing of text knowledge based on deep learning according to claim 1 is characterized in that: In step S3, after extracting key information, the information is represented and stored in the form of a graph using knowledge graph technology. In the process of constructing the knowledge graph, a graph database management system with high query speed and large-scale data processing capabilities is used to support the organization and management of equipment fault knowledge and ensure the storage and query performance of the graph.

9. The method for automatic extraction and graphing of text knowledge based on deep learning according to claim 1, characterized in that: The method supports parallel processing of large amounts of documents and complex data through a distributed computing architecture or multi-threaded processing, has good scalability and adaptability, is suitable for efficiently completing the extraction of equipment failure-related information and the construction of a knowledge graph in a short period of time, and is adaptable to water conservancy equipment maintenance documents of different sizes and a variety of data types.

10. A text knowledge automatic extraction and graphing system based on deep learning, characterized in that: The system adopts modular design and consists of the following independent modules: Semantic analysis module: used to perform the semantic analysis function of the deep learning model. This module uses the deep learning model trained on a large-scale general corpus and a special domain corpus to perform deep semantic understanding of the input text and identify key information in the text, such as device name, fault phenomenon, fault cause, and treatment measures. Information extraction module: Based on the output of the semantic analysis module, it automatically extracts key information related to equipment failure and generates a data structure for subsequent knowledge graph construction. This module supports multi-threaded processing to improve information extraction efficiency and adapts to multiple data formats; Knowledge graph construction module: Using knowledge graph technology, the data generated by the information extraction module is represented and stored in the form of a graph. This module uses a graph database management system that supports memory indexing and distributed storage technology, which is suitable for efficient query and processing of large-scale data, ensuring the storage and query performance of the knowledge graph; Export module: supports exporting the results of knowledge extraction and graph construction into multiple formats to meet users' diverse data needs; Integration interface module: provides an integration interface with existing industrial information systems, supports modular configuration and deployment, and enables the system to be flexibly integrated with existing systems.

Images