Emergency inspection knowledge base construction method based on knowledge fusion driving

Abstract

The invention discloses an emergency inspection knowledge base construction method based on knowledge fusion driving, which belongs to the technical field of emergency management and information, and specifically comprises the following steps: collecting and preprocessing emergency inspection data from different data sources, extracting and representing emergency inspection knowledge by using a relationship extraction method and a semantic web technology, and constructing an emergency inspection knowledge base. Constructing an emergency inspection knowledge graph through a knowledge fusion strategy; knowledge in the emergency inspection knowledge graph is classified and encoded according to emergency inspection requirements, and a classification word list is formulated; knowledge reasoning is carried out based on the emergency inspection knowledge graph, an intelligent question answering system and an emergency inspection process are constructed, and an emergency inspection knowledge base is formed; optimizing and improving the knowledge base through function testing, performance testing and user feedback; according to the invention, the emergency inspection efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention belongs to the field of emergency management and information technology, and specifically is a method for constructing an emergency inspection knowledge base based on knowledge fusion drive. Background Art

[0002] In emergency inspection work, the knowledge base plays a vital role in guiding inspectors to carry out their work accurately and efficiently. However, the existing knowledge base construction methods have the following problems. On the one hand, the sources of knowledge are scattered, including laws and regulations, standards and specifications, enterprise safety risk data, hidden danger data and other types, which are difficult to integrate; on the other hand, traditional methods lack effective knowledge fusion means and cannot fully explore the relationship between different knowledge, resulting in insufficient ability of the knowledge base to deal with complex emergency scenarios and cannot provide comprehensive and accurate knowledge support for emergency inspections; in addition, the knowledge representation and update mechanism is not flexible enough, and it is difficult to adapt to the characteristics of rapid changes in knowledge in the emergency field, which affects the application effect of the knowledge base in actual emergency inspections.

[0003] For example, a Chinese patent with authorization announcement number CN111143448B discloses a knowledge base construction method, including: determining a data source; obtaining ontology data according to the data source; based on the ontology data; capturing ontology data according to the data source to generate data to be processed; determining whether the data to be processed is semi-structured data; if the data to be processed is not semi-structured data, extracting knowledge information from the data to be processed to obtain result data; if the data to be processed is semi-structured data, performing unified processing on the data to be processed to obtain unified processed data; extracting knowledge information from the unified processed data to obtain result data; if the data to be processed is not semi-structured data, extracting knowledge information from the data to be processed to obtain result data; adding the result data to the knowledge base.

[0004] For example, a Chinese patent with authorization announcement number CN111782825B discloses a knowledge base construction method and device, including: obtaining audit information and audit requirements, wherein the audit requirements correspond to multiple audit fields; grouping the audit information according to preset grouping rules to generate multiple audit data groups; constructing an audit field knowledge graph corresponding to each audit field based on the multiple audit data groups and the audit requirements; generating an audit knowledge base based on each audit field knowledge graph. The knowledge base construction method provided by this technical solution creates multiple audit field knowledge graphs in multiple audit fields based on audit information and audit requirements, and then generates an audit knowledge base from the multiple audit field knowledge graphs, which can timely and effectively obtain various audit information and facilitate timely correct and reasonable changes.

[0005] The above existing technologies all have the following problems: In CN111143448B, the knowledge information extraction method is ambiguous, and lacks knowledge fusion and graph construction; it does not involve the classification and coding of knowledge, which may make it difficult to effectively retrieve and utilize the knowledge in the knowledge base; there is a lack of intelligent application; in CN111782825B, it is mentioned that a knowledge graph of the audit field is constructed based on the audit data group and audit requirements, but the method of knowledge extraction and the form of knowledge representation are not specifically explained; there is no mention of the process of knowledge fusion, that is, how to identify and merge duplicate entities, and how to optimize the knowledge set; there is a lack of knowledge classification and coding. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention proposes a method for constructing an emergency inspection knowledge base driven by knowledge fusion, collects and preprocesses emergency inspection data from different data sources, uses relationship extraction methods and semantic web technology to extract and represent emergency inspection knowledge, and constructs an emergency inspection knowledge graph through a knowledge fusion strategy; classifies and encodes the knowledge in the emergency inspection knowledge graph according to emergency inspection needs, and formulates a classification vocabulary; performs knowledge reasoning based on the emergency inspection knowledge graph, constructs an intelligent question-and-answer system and emergency inspection process, and forms an emergency inspection knowledge base; optimizes and improves the knowledge base through functional testing, performance testing and user feedback. The present invention improves the efficiency and accuracy of emergency inspections.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The emergency inspection knowledge base construction method based on knowledge fusion drive includes:

[0009] Step S1: Collect emergency inspection data from structured, semi-structured and unstructured data sources, and perform preprocessing to obtain emergency inspection text data;

[0010] Step S2: extract emergency inspection knowledge from the emergency inspection text data using a relation extraction method combined with emergency inspection specifications, and use a knowledge representation method based on the semantic web to represent the extracted emergency inspection knowledge in the form of subject-predicate-object triples to form an emergency inspection knowledge set;

[0011] Step S3: Use the knowledge fusion strategy to identify and merge duplicate entities in the emergency inspection knowledge set. At the same time, use the rule engine to optimize the merged emergency inspection knowledge set, and build an emergency inspection knowledge graph based on the knowledge fusion to store the knowledge in the form of a graph structure.

[0012] Step S4: Classify the knowledge in the emergency inspection knowledge graph according to the needs and field characteristics of the emergency inspection, and establish a unique code for each emergency inspection knowledge under the classification to obtain the classified emergency inspection knowledge graph;

[0013] Step S5: Using natural language processing technology and the classified emergency inspection knowledge graph, a question-answering system is constructed, and based on the classified emergency inspection knowledge graph and the question-answering system, an emergency inspection process is constructed to form an emergency inspection knowledge base;

[0014] Step S6: Perform functional and performance tests on the emergency inspection knowledge base, collect user feedback, and optimize and improve the emergency inspection knowledge base based on the feedback.

[0015] Specifically, the specific steps of step S2 include:

[0016] S2.1: According to the emergency inspection specifications, clarify the type and scope of knowledge to be extracted;

[0017] S2.2: Obtain emergency inspection text data , and from Extract emergency inspection features from represents the nth emergency inspection text data, where n represents the number of emergency inspection text data;

[0018] S2.3: Load the pre-built machine learning model, and use the extracted emergency inspection features to train and evaluate the pre-built machine learning model to generate a relationship extraction model;

[0019] S2.4: Input into the relation extraction model to identify entities and relations, obtain the entities and relations between entities in the emergency inspection text, and generate emergency inspection knowledge , where K represents the generated emergency inspection knowledge, represents the relation extraction model, E represents the identified entities in the emergency inspection text, R represents the set of relations between entities in the emergency inspection text, Indicates a mapping relationship;

[0020] S2.5: Based on the generated emergency inspection knowledge K, all entities and their corresponding relationships are identified, and subject-predicate-object triples are constructed based on the identified entities and their corresponding relationships. The subject-predicate-object triples are ,in, and Represent two different entities. express and interaction or connection between them.

[0021] Specifically, the specific steps of step S2 also include:

[0022] S2.6: Verify the constructed subject-verb-object triples;

[0023] like , then the constructed subject-verb-object triple is wrong and needs to be reconstructed;

[0024] like , then judge Is it described? and interactions or connections between them;

[0025] like and and If the actual connection between them does not match, the constructed subject-verb-object triple is wrong and needs to be reconstructed;

[0026] like and and If the actual connection between Whether it belongs to the set of relations between entities in the emergency inspection text;

[0027] like , then judge Is there a conflict in R?

[0028] If there is a conflict, standard data in the field is used for screening;

[0029] If there is no conflict, the constructed subject-predicate-object triple is retained;

[0030] S2.7: Combine multiple constructed subject-verb-object triples to form an emergency inspection knowledge set;

[0031] S2.8: Use the semantic web method to represent the integrated emergency inspection knowledge set as a structured semantic network and store the structured semantic network.

[0032] Specifically, the specific steps of step S3 include:

[0033] S3.1: Obtain emergency inspection knowledge set and perform preprocessing;

[0034] S3.2: Use entity alignment algorithm to identify duplicate entities in the preprocessed emergency inspection knowledge set, and merge the identified duplicate entities;

[0035] S3.3: Use the rule engine to optimize the merged emergency inspection knowledge set;

[0036] S3.4: Based on the optimized emergency inspection knowledge set, construct an emergency inspection knowledge graph, and store the constructed emergency inspection knowledge graph in the database.

[0037] Specifically, the specific steps of S3.2 include:

[0038] S3.21: Obtain the preprocessed emergency inspection knowledge set B, identify and extract the named entities in B ;in, represents the i-th named entity in B, and m represents the number of named entities in B;

[0039] S3.22: Extract features from each entity and use the entity alignment algorithm to calculate the similarity between entities based on the extracted features. The formula is:

[0040] ;

[0041] in, Indicates two different entities in B and The similarity of and Respectively and The neighbor entity set of , C represents the damping coefficient, and , Representing Entities The weight of the jth neighbor of Representing Entities The weight of the lth neighbor of represents the similarity based on neighbor entities, and Respectively The jth neighbor entity and The lth neighbor entity of Representing Entities and The similarity of its own attributes;

[0042] S3.23: Set similarity threshold h;

[0043] like , then it means Corresponding entity and is a repeating entity;

[0044] S3.24: Merge duplicate entities and output a list of duplicate entities.

[0045] Specifically, the specific steps of step S4 include:

[0046] S4.1: Obtain the knowledge graph of emergency inspection, set mutually exclusive classification principles according to the needs of emergency inspection, and build a preliminary knowledge classification system based on the set mutually exclusive classification principles;

[0047] S4.2: Refine and label the preliminary knowledge classification system, generate a knowledge classification system, and classify the knowledge in the emergency inspection knowledge graph according to the knowledge classification system;

[0048] S4.3: Design a unique code for each category of emergency inspection knowledge, and add the designed unique code to the corresponding position of each emergency inspection knowledge in the emergency inspection knowledge graph. At the same time, update the emergency inspection knowledge graph.

[0049] Specifically, the specific steps of step S5 include:

[0050] S5.1: Design the overall architecture of the question-answering system, which includes a user interface, a preprocessing module, a semantic understanding module, a knowledge retrieval module, an answer generation module, and a post-processing module;

[0051] S5.2: Connect the classified emergency inspection knowledge graph to the question-answering system as the basis for knowledge retrieval and answer generation;

[0052] S5.3: Design the emergency inspection process based on the emergency inspection knowledge graph and the functions of the question-answering system, and integrate the emergency inspection process into the management system;

[0053] S5.4: Optimize and adjust the emergency inspection process according to actual usage;

[0054] S5.5: Integrate the classified emergency inspection knowledge graph, question-and-answer system, and optimized and adjusted emergency inspection process to form an emergency inspection knowledge base.

[0055] Specifically, the mutually exclusive classification principle in S4.1 means that there is no overlap between classifications.

[0056] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a method for constructing an emergency inspection knowledge base driven by knowledge fusion are implemented.

[0057] A computer-readable storage medium stores computer instructions, which, when executed, execute the steps of a method for building an emergency inspection knowledge base driven by knowledge fusion.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] 1. The present invention proposes a method for constructing an emergency inspection knowledge base driven by knowledge fusion. By collecting and preprocessing data from diverse data sources, and then using natural language processing and semantic web technology to extract and structure knowledge representation, it effectively integrates a wide range of information in the field of emergency inspections. The knowledge fusion strategy and the construction of the knowledge graph further ensure the uniqueness and accuracy of the data.

[0060] 2. The present invention proposes a method for constructing an emergency inspection knowledge base based on knowledge fusion drive, and also improves the intelligence level of emergency inspection. Through the classification and coding of knowledge, knowledge reasoning, and the construction of an intelligent question-and-answer system and inspection process, it not only improves the speed of emergency response, but also makes the inspection process more scientific and standardized; at the same time, continuous functional testing, performance testing and user feedback collection ensure the continuous optimization and improvement of the emergency inspection knowledge base, so as to better serve emergency management work. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a schematic diagram of the method for constructing an emergency inspection knowledge base based on knowledge fusion drive of the present invention;

[0062] Figure 2 This is a principle flow chart of the method for building an emergency inspection knowledge base based on knowledge fusion drive of the present invention;

[0063] Figure 3 It is a flow chart of implementing the emergency inspection knowledge set of the emergency inspection knowledge base construction method based on knowledge fusion drive of the present invention;

[0064] Figure 4 This is a flow chart for implementing the emergency inspection knowledge graph of the emergency inspection knowledge base construction method driven by knowledge fusion in the present invention. DETAILED DESCRIPTION

[0065] Example 1

[0066] See also Figure 1 and Figure 2 The present invention provides an embodiment: a method for constructing an emergency inspection knowledge base based on knowledge fusion drive, comprising the following steps:

[0067] Step S1: Collect emergency inspection data from structured, semi-structured and unstructured data sources, and perform preprocessing to obtain emergency inspection text data;

[0068] Among them, the emergency inspection data collected in the form of text, including plain text, documents, and web page content, is text data; structured data sources include databases, tables, etc.; semi-structured data sources include web pages, logs, etc.; unstructured data sources include text files, images, videos, etc.

[0069] Furthermore, the preprocessing process includes:

[0070] (1) Clean the collected emergency inspection data to remove irrelevant information, advertisements, tags, HTML, JS and other codes and comments;

[0071] (2) Segment the text into smaller language units for subsequent processing;

[0072] (3) Part-of-speech tagging: label each word or phrase with a word class, such as adjective, verb, noun, etc., to incorporate more useful language information;

[0073] (4) Remove stop words, that is, words that do not contribute to the text features, such as punctuation marks, modal particles, etc.

[0074] Step S2: extract emergency inspection knowledge from the emergency inspection text data using a relation extraction method combined with emergency inspection specifications, and use a knowledge representation method based on the semantic web to represent the extracted emergency inspection knowledge in the form of subject-predicate-object triples to form an emergency inspection knowledge set;

[0075] Step S3: Use the knowledge fusion strategy to identify and merge duplicate entities in the emergency inspection knowledge set. At the same time, use the rule engine to optimize the merged emergency inspection knowledge set, and build an emergency inspection knowledge graph based on the knowledge fusion to store the knowledge in the form of a graph structure.

[0076] Step S4: Classify the knowledge in the emergency inspection knowledge graph according to the needs and field characteristics of the emergency inspection, and establish a unique code for each emergency inspection knowledge under the classification to obtain the classified emergency inspection knowledge graph;

[0077] Step S5: Using natural language processing technology and the classified emergency inspection knowledge graph, a question-answering system is constructed, and based on the classified emergency inspection knowledge graph and the question-answering system, an emergency inspection process is constructed to form an emergency inspection knowledge base;

[0078] Step S6: Perform functional and performance tests on the emergency inspection knowledge base, collect user feedback, and optimize and improve the emergency inspection knowledge base based on the feedback.

[0079] Example 2

[0080] See also Figure 3 and Figure 4 In this embodiment, the specific steps of step S2 include:

[0081] S2.1: According to the emergency inspection specifications, clarify the type and scope of knowledge to be extracted;

[0082] S2.2: Obtain emergency inspection text data , and from Extract emergency inspection features from represents the nth emergency inspection text data, where n represents the number of emergency inspection text data;

[0083] S2.3: Load a pre-built machine learning model, and use the extracted emergency inspection features to train and evaluate the pre-built machine learning model to generate a relationship extraction model, wherein the machine learning model is the prior art content in this field, is not the inventive solution of the present application, and is not described in detail here;

[0084] S2.4: Input into the relation extraction model to identify entities and relations, obtain the entities and relations between entities in the emergency inspection text, and generate emergency inspection knowledge , where K represents the generated emergency inspection knowledge, represents the relation extraction model, E represents the identified entities in the emergency inspection text, R represents the set of relations between entities in the emergency inspection text, Indicates a mapping relationship;

[0085] It should be noted that the expression A relation extraction process is described, in which the relation extraction model M extracts entities E and relations R from the input emergency inspection text data X, and finally generates structured emergency inspection knowledge K.

[0086] S2.5: Based on the generated emergency inspection knowledge K, all entities and their corresponding relationships are identified, and subject-predicate-object triples are constructed based on the identified entities and their corresponding relationships. The subject-predicate-object triples are ,in, and Represent two different entities. express and interactions or connections between them;

[0087] S2.6: Verify the constructed subject-verb-object triples;

[0088] like , then the constructed subject-verb-object triple is wrong and needs to be reconstructed;

[0089] like , then judge Is it described? and interactions or connections between them;

[0090] like and and If the actual connection between them does not match, the constructed subject-verb-object triple is wrong and needs to be reconstructed;

[0091] like and and If the actual connection between Whether it belongs to the set of relations between entities in the emergency inspection text;

[0092] like , then judge Is there a conflict in R?

[0093] If there is a conflict, standard data in the field is used for screening;

[0094] If there is no conflict, the constructed subject-predicate-object triple is retained;

[0095] S2.7: Combine multiple constructed subject-verb-object triples to form an emergency inspection knowledge set;

[0096] S2.8: Use the semantic web method to represent the integrated emergency inspection knowledge set as a structured semantic network and store the structured semantic network.

[0097] Furthermore, the specific steps of S2.8 include:

[0098] (1) Obtain emergency inspection knowledge set and perform preprocessing;

[0099] (2) Define a semantic model based on the characteristics and requirements of emergency inspection knowledge. The semantic model includes entities, attributes, relationships, and their hierarchical structures and relationship networks. In the semantic model, entities can represent key elements in emergency inspections, such as equipment, personnel, and locations; attributes can describe the characteristics or status of entities; and relationships are used to represent the interactions or connections between entities.

[0100] (3) Using the entities, attributes, and relationships in the semantic model, a structured semantic network is constructed. This network is usually represented in the form of a graph, where nodes represent entities or attributes and edges represent relationships. At the same time, when constructing a semantic network, it is necessary to ensure that each entity and relationship has a clear definition and description to facilitate subsequent storage and query.

[0101] (4) Convert the data in the semantic network into the storage format of the graph database and store it in the corresponding storage medium;

[0102] (5) Verify the stored semantic network to ensure the accuracy and completeness of the data, and optimize the semantic network based on the verification results, such as correcting erroneous data and completing missing information.

[0103] The specific steps of step S3 include:

[0104] S3.1: Obtain emergency inspection knowledge set and perform preprocessing;

[0105] S3.2: Use entity alignment algorithm to identify duplicate entities in the preprocessed emergency inspection knowledge set, and merge the identified duplicate entities;

[0106] S3.3: Use a rule engine to optimize the merged emergency inspection knowledge set, wherein the rule engine is a prior art in this field and is not an inventive solution of the present application, and will not be described in detail here;

[0107] S3.4: Based on the optimized emergency inspection knowledge set, construct an emergency inspection knowledge graph, and store the constructed emergency inspection knowledge graph in the database.

[0108] Furthermore, the specific steps of S3.4 include:

[0109] (1) Based on the application attributes, knowledge characteristics, and actual needs of emergency inspections, business abstraction and business modeling are performed according to the knowledge graph model, mainly entity definition and relationship definition;

[0110] (2) Obtain an optimized set of emergency inspection knowledge to ensure the accuracy, completeness, and consistency of the knowledge;

[0111] (3) Based on the characteristics and requirements of the optimized emergency inspection knowledge, the basic architecture of the emergency inspection knowledge graph is designed, including the definitions of entities, relationships, attributes, and the ways in which they are associated;

[0112] (4) Knowledge extraction: extracting structured information such as entities, relationships, and attributes from the optimized emergency inspection knowledge set;

[0113] (5) Entity extraction: Identify and extract key entities in the emergency inspection knowledge set, such as equipment, fault type, and inspection standard;

[0114] (6) Relationship extraction: Determine the relationship between entities, such as the correspondence between equipment and fault types, the association between equipment and inspection standards, etc.

[0115] (7) Attribute extraction: extracting the attribute information of entities, such as the model of the equipment, the description of the fault type, and the specific content of the inspection standard;

[0116] (8) Knowledge fusion: Integrate and fuse the extracted knowledge, eliminate contradictions and ambiguities, and form a unified knowledge system;

[0117] (9) Knowledge storage: The constructed emergency inspection knowledge graph is stored in the database for subsequent query and application.

[0118] The specific steps of S3.2 include:

[0119] S3.21: Obtain the preprocessed emergency inspection knowledge set B, identify and extract the named entities in B ;in, represents the i-th named entity in B, and m represents the number of named entities in B;

[0120] S3.22: Extract features from each entity and use the entity alignment algorithm to calculate the similarity between entities based on the extracted features. The formula is:

[0121] ;

[0122] in, Indicates two different entities in B and The similarity of and Respectively and The neighbor entity set of , C represents the damping coefficient, and , Representing Entities The weight of the jth neighbor of Representing Entities The weight of the lth neighbor of represents the similarity based on neighbor entities, and Respectively The jth neighbor entity and The lth neighbor entity of Representing Entities and The similarity of its own attributes;

[0123] It should be noted that the formula in the present invention introduces weights and combines neighborhood similarity and attribute similarity, so that the similarity calculation is more accurate and comprehensive, and can reflect the complex relationship between two objects in more detail.

[0124] S3.23: Set similarity threshold h;

[0125] like , then it means Corresponding entity and is a repeating entity;

[0126] S3.24: Merge duplicate entities and output a list of duplicate entities.

[0127] The specific steps of S4 include:

[0128] S4.1: Obtain the knowledge graph of emergency inspection, set mutually exclusive classification principles according to the needs of emergency inspection, and build a preliminary knowledge classification system based on the set mutually exclusive classification principles;

[0129] Among them, the principle of mutually exclusive classification means that there is no overlap between classifications.

[0130] S4.2: Refine and label the preliminary knowledge classification system, generate a knowledge classification system, and classify the knowledge in the emergency inspection knowledge graph according to the knowledge classification system;

[0131] Furthermore, the specific steps of S4.2 include:

[0132] (1) Obtain a preliminary knowledge classification system, add necessary sub-categories to the preliminary classification according to the characteristics of the field and user needs, refine the classification system, and ensure that the logical relationship between the sub-categories and the parent categories is clear and accurate;

[0133] (2) Based on the analysis results, adjust the classification structure, such as merging duplicate categories and splitting overly general categories, to ensure that the classification system is hierarchical, logical, and complete;

[0134] (3) Design clear, concise, and easy-to-understand labels for each category and subcategory. The labels should accurately reflect the content and characteristics of the category and facilitate user identification and retrieval;

[0135] (4) Organize the designed tags according to the classification system to form a tag system. At the same time, ensure that the tag system is consistent with the knowledge classification system to facilitate subsequent knowledge classification and retrieval;

[0136] (5) Combine the refined classification system and label system to generate a knowledge classification system. The knowledge classification system should be able to comprehensively cover the knowledge in the field and be easy for users to retrieve and use.

[0137] (6) Obtain the emergency inspection knowledge graph, classify the knowledge in the emergency inspection knowledge graph according to the generated knowledge classification system, classify each knowledge entity into the corresponding category or subcategory, and mark it with the corresponding label;

[0138] (7) Verify the classification results to ensure the accuracy and rationality of the classification, and identify and correct existing classification errors or omissions.

[0139] S4.3: Design a unique code for each category of emergency inspection knowledge, and add the designed unique code to the corresponding position of each emergency inspection knowledge in the emergency inspection knowledge graph. At the same time, update the emergency inspection knowledge graph.

[0140] Furthermore, the specific steps of S4.3 include:

[0141] (1) Design reasonable coding rules based on the classification system, knowledge characteristics and management requirements of emergency inspection knowledge. The coding rules should ensure that each category of emergency inspection knowledge can obtain a unique code;

[0142] (2) Assign a unique code range to each category to avoid coding conflicts;

[0143] (3) Generate a unique code for each emergency inspection knowledge according to the designed coding rules and the assigned coding range. The code contains elements such as classification information, knowledge sequence number, and timestamp;

[0144] (4) Use the code management system to verify whether the generated code is unique. If there are duplicate codes, they need to be regenerated until each code is ensured to be unique;

[0145] (5) Find the corresponding position of each emergency inspection knowledge in the emergency inspection knowledge graph through a local traversal method, and add the generated unique code to the metadata or attributes of the corresponding emergency inspection knowledge. At the same time, ensure that the addition of the code does not destroy the structure and integrity of the knowledge graph. The local traversal method is the prior art content in this field and is not the inventive solution of this application, so it will not be described in detail here.

[0146] (6) After adding the code, the emergency inspection knowledge graph shall be verified to ensure that the addition of the code does not introduce errors or omissions, and the updated emergency inspection knowledge graph shall be published to the corresponding storage system or service platform to ensure that users can access the latest version of the emergency inspection knowledge graph.

[0147] Exemplarily, assume that D is the emergency inspection knowledge set, Q is the classification set, and G is the coding set;

[0148] For each emergency check knowledge , there is a classification , and a unique code ;

[0149] Encoding generation function Generate a unique code g based on emergency inspection knowledge d and classification q;

[0150] Encoding add function The generated code g is added to the metadata or attributes of the emergency inspection knowledge d.

[0151] The specific steps of step S5 include:

[0152] S5.1: Design the overall architecture of the question-answering system, which includes a user interface, a preprocessing module, a semantic understanding module, a knowledge retrieval module, an answer generation module, and a post-processing module;

[0153] Furthermore, the steps of designing the overall architecture of the question answering system include:

[0154] (1) User interface: Design a simple and clear user interface to facilitate users to enter questions and view results;

[0155] (2) Question processing: Determine how the question-answering system processes questions input by users, including steps such as word segmentation, semantic understanding, knowledge retrieval, and answer generation;

[0156] (3) Result feedback: Design the way the question-answering system provides feedback to users, including text, charts, voice, and other forms.

[0157] S5.2: Connect the classified emergency inspection knowledge graph to the question-answering system as the basis for knowledge retrieval and answer generation;

[0158] S5.3: Design the emergency inspection process based on the emergency inspection knowledge graph and the functions of the question-answering system, and integrate the emergency inspection process into the management system;

[0159] Furthermore, the specific steps for designing the emergency inspection process include:

[0160] (1) Use the question-and-answer system to quickly obtain basic knowledge about emergency inspections, such as inspection standards, methods, and precautions, and determine the focus and priority of inspections based on the relationships in the emergency inspection knowledge graph;

[0161] (2) Conduct detailed on-site inspections according to the determined focus and priority. During the inspection process, use the question-answering system to obtain relevant information and answer questions in real time. Based on the inspection results, use the relationships in the emergency inspection knowledge graph to reason and analyze to identify potential risks and problems.

[0162] (3) For the problems found, use the question-and-answer system to obtain solutions and suggestions, and formulate specific rectification measures and timetables based on the solutions and suggestions. At the same time, record the rectification measures and timetable in the knowledge graph for subsequent tracking and verification;

[0163] (4) Use computer technology to automate the above emergency inspection steps to form a complete emergency inspection process, and optimize and adjust the emergency inspection process according to actual application conditions to improve its efficiency and accuracy.

[0164] S5.4: Optimize and adjust the emergency inspection process according to actual usage;

[0165] S5.5: Integrate the classified emergency inspection knowledge graph, question-and-answer system, and optimized and adjusted emergency inspection process to form an emergency inspection knowledge base.

[0166] Example 3

[0167] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a method for constructing an emergency inspection knowledge base driven by knowledge fusion are implemented. For details, please refer to the above method embodiment, which will not be repeated here.

[0168] A computer-readable storage medium stores computer instructions, which, when executed, execute the steps of a method for building an emergency inspection knowledge base driven by knowledge fusion, wherein the storage medium may be a volatile or non-volatile computer-readable storage medium.

[0169] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the purpose and scope of protection of the present invention, and all of these are within the protection of the present invention.

[0170] If the disclosed technical solution involves personal information, the product using the disclosed technical solution has clearly informed the personal information processing rules and obtained the individual's voluntary consent before processing the personal information. If the disclosed technical solution involves sensitive personal information, the product using the disclosed technical solution has obtained the individual's separate consent before processing the sensitive personal information, and at the same time meets the "explicit consent" requirement. For example, on personal information collection devices such as cameras, clear and prominent signs are set to inform that the personal information collection scope has been entered and personal information will be collected. If the individual voluntarily enters the collection scope, it is deemed that he or she agrees to the collection of his or her personal information; or on the device that processes personal information, when the personal information processing rules are notified by obvious signs / information, the individual's authorization is obtained through pop-up information or by asking the individual to upload his or her personal information; among them, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the type of personal information processed.

Claims

1. The method for constructing an emergency inspection knowledge base based on knowledge fusion drive is characterized by: include: Step S1: Collect emergency inspection data from structured, semi-structured and unstructured data sources, and perform preprocessing to obtain emergency inspection text data; Step S2: extract emergency inspection knowledge from the emergency inspection text data using a relation extraction method combined with emergency inspection specifications, and use a knowledge representation method based on the semantic web to represent the extracted emergency inspection knowledge in the form of subject-predicate-object triples to form an emergency inspection knowledge set; Step S3: Use the knowledge fusion strategy to identify and merge duplicate entities in the emergency inspection knowledge set. At the same time, use the rule engine to optimize the merged emergency inspection knowledge set, and build an emergency inspection knowledge graph based on the knowledge fusion to store the knowledge in the form of a graph structure. Step S4: Classify the knowledge in the emergency inspection knowledge graph according to the needs and field characteristics of the emergency inspection, and establish a unique code for each emergency inspection knowledge under the classification to obtain the classified emergency inspection knowledge graph; Step S5: Using natural language processing technology and the classified emergency inspection knowledge graph, a question-answering system is constructed, and based on the classified emergency inspection knowledge graph and the question-answering system, an emergency inspection process is constructed to form an emergency inspection knowledge base; Step S6: Perform functional and performance tests on the emergency inspection knowledge base, collect user feedback, and optimize and improve the emergency inspection knowledge base based on the feedback.

2. The method for constructing an emergency inspection knowledge base based on knowledge fusion drive according to claim 1, characterized in that: The specific steps of step S2 include: S2.1: According to the emergency inspection specifications, clarify the type and scope of knowledge to be extracted; S2.2: Obtain emergency inspection text data , and from Extract emergency inspection features from represents the nth emergency inspection text data, where n represents the number of emergency inspection text data; S2.3: Load the pre-built machine learning model, and use the extracted emergency inspection features to train and evaluate the pre-built machine learning model to generate a relationship extraction model; S2.4: Input into the relation extraction model to identify entities and relations, obtain the entities and relations between entities in the emergency inspection text, and generate emergency inspection knowledge , where K represents the generated emergency inspection knowledge, represents the relation extraction model, E represents the identified entities in the emergency inspection text, R represents the set of relations between entities in the emergency inspection text, Indicates a mapping relationship; S2.5: Based on the generated emergency inspection knowledge K, all entities and their corresponding relationships are identified, and subject-predicate-object triples are constructed based on the identified entities and their corresponding relationships. The subject-predicate-object triples are ,in, and Represent two different entities. express and interaction or connection between them.

3. The method for constructing an emergency inspection knowledge base based on knowledge fusion drive according to claim 2, characterized in that: The specific steps of step S2 also include: S2.6: Verify the constructed subject-verb-object triples; like , then the constructed subject-verb-object triple is wrong and needs to be reconstructed; like , then judge Is it described? and interactions or connections between them; like and and If the actual connection between them does not match, the constructed subject-verb-object triple is wrong and needs to be reconstructed; like and and If the actual connection between Whether it belongs to the set of relations between entities in the emergency inspection text; like , then judge Is there a conflict in R? If there is a conflict, standard data in the field is used for screening; If there is no conflict, the constructed subject-predicate-object triple is retained; S2.7: Combine multiple constructed subject-verb-object triples to form an emergency inspection knowledge set; S2.8: Use the semantic web method to represent the integrated emergency inspection knowledge set as a structured semantic network and store the structured semantic network.

4. The method for constructing an emergency inspection knowledge base based on knowledge fusion drive according to claim 3 is characterized in that: The specific steps of step S3 include: S3.1: Obtain emergency inspection knowledge set and perform preprocessing; S3.2: Use entity alignment algorithm to identify duplicate entities in the preprocessed emergency inspection knowledge set, and merge the identified duplicate entities; S3.3: Use the rule engine to optimize the merged emergency inspection knowledge set; S3.4: Based on the optimized emergency inspection knowledge set, construct an emergency inspection knowledge graph, and store the constructed emergency inspection knowledge graph in the database.

5. The method for constructing an emergency inspection knowledge base based on knowledge fusion drive according to claim 4, characterized in that: The specific steps of S3.2 include: S3.21: Obtain the preprocessed emergency inspection knowledge set B, identify and extract the named entities in B ;in, represents the i-th named entity in B, and m represents the number of named entities in B; S3.22: Extract features from each entity and use the entity alignment algorithm to calculate the similarity between entities based on the extracted features. The formula is: ； in, Indicates two different entities in B and The similarity of and Respectively and The neighbor entity set of , C represents the damping coefficient, and , Representing Entities The weight of the jth neighbor of Representing Entities The weight of the lth neighbor of represents the similarity based on neighbor entities, and Respectively The jth neighbor entity and The lth neighbor entity of Representing Entities and The similarity of its own attributes; S3.23: Set similarity threshold h; like , then it means Corresponding entity and is a repeating entity; S3.24: Merge duplicate entities and output a list of duplicate entities.

6. The method for constructing an emergency inspection knowledge base based on knowledge fusion drive according to claim 5, characterized in that: The specific steps of step S4 include: S4.1: Obtain the knowledge graph of emergency inspection, set mutually exclusive classification principles according to the needs of emergency inspection, and build a preliminary knowledge classification system based on the set mutually exclusive classification principles; S4.2: Refine and label the preliminary knowledge classification system, generate a knowledge classification system, and classify the knowledge in the emergency inspection knowledge graph according to the knowledge classification system; S4.3: Design a unique code for each category of emergency inspection knowledge, and add the designed unique code to the corresponding position of each emergency inspection knowledge in the emergency inspection knowledge graph. At the same time, update the emergency inspection knowledge graph.

7. The method for constructing an emergency inspection knowledge base based on knowledge fusion drive according to claim 6, characterized in that: The specific steps of step S5 include: S5.1: Design the overall architecture of the question-answering system, which includes a user interface, a preprocessing module, a semantic understanding module, a knowledge retrieval module, an answer generation module, and a post-processing module; S5.2: Connect the classified emergency inspection knowledge graph to the question-answering system as the basis for knowledge retrieval and answer generation; S5.3: Design the emergency inspection process based on the emergency inspection knowledge graph and the functions of the question-answering system, and integrate the emergency inspection process into the management system; S5.4: Optimize and adjust the emergency inspection process according to actual usage; S5.5: Integrate the classified emergency inspection knowledge graph, question-and-answer system, and optimized and adjusted emergency inspection process to form an emergency inspection knowledge base.

8. The method for constructing an emergency inspection knowledge base based on knowledge fusion drive according to claim 6, characterized in that: The mutually exclusive classification principle in S4.1 means that there is no overlap between classifications.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for constructing an emergency inspection knowledge base based on knowledge fusion drive described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, the steps of the method for constructing an emergency inspection knowledge base based on knowledge fusion drive as described in any one of claims 1 to 8 are executed.

Images