Coal mine accident information learning method and system

Abstract

The embodiment of the invention provides a coal mine accident information learning method and system, and belongs to the technical field of knowledge maps. The method comprises the following steps: collecting historical coal mine accident information, and constructing a corresponding accident cause network model based on the historical coal mine accident information; executing corresponding information extraction based on the accident cause network model to construct a corresponding coal mine accident knowledge graph based on the extracted information; performing visualization processing on the coal mine accident knowledge graph, and storing the coal mine accident knowledge graph; and responding to a query request of a user, returning answer information corresponding to the request type based on the stored coal mine accident knowledge graph, and pushing the answer information to the user side. According to the scheme, the processing efficiency, the analysis depth and the query convenience of the coal mine accident information are effectively improved, and an intelligent auxiliary decision-making tool is provided for coal mine safety production.

Description

Technical Field

[0001] The present invention relates to the technical field of knowledge graphs, and specifically relates to a coal mine accident information learning method and a coal mine accident information learning system. Background Art

[0002] In the coal mining industry, work safety has always been the focus of attention. With the increase in the number of coal mine accidents and the complexity of their causes, how to effectively analyze and prevent coal mine accidents has become an urgent problem to be solved. Most of the existing coal mine accident analysis methods rely on traditional manual records and case analysis, and learn from experience by summarizing typical accidents. However, this method has obvious limitations. First, most of the existing coal mine accident data exists in the form of unstructured text. Information such as accident investigation reports, hidden danger investigation records, and hazard inspection records is often scattered, lacking a unified standardized format and systematic storage. This poses a huge information management challenge for coal mine managers when they need to quickly search for and analyze the causes of accidents. They often need to spend a lot of time searching and sorting, and it is difficult to achieve quick response and efficient decision-making. Second, due to the usually complex causal relationships between accident causes, traditional analysis methods often ignore the depth and complexity of these relationships. The root causes of accidents are often the result of the interweaving of multiple factors, and existing solutions are difficult to accurately model and analyze these complex causal relationships. Therefore, they can only stay on the surface and it is difficult to effectively prevent accidents from occurring.

[0003] In addition, the application of knowledge graphs in the field of coal mine safety is less. Although knowledge graphs have been widely used in other fields for accident cause analysis and decision support, their application in the field of coal mine accidents is still insufficient. Most of the existing coal mine accident data analysis methods rely on manual sorting and static records, lacking the ability of dynamic update and intelligent processing. At the same time, the knowledge graph construction methods in many existing solutions are too general and do not fully consider the particularity of the coal mining industry, resulting in poor practical application effects of knowledge graphs in coal mine accident cause analysis. Although some solutions attempt to introduce text analysis technology, these solutions have insufficient processing capabilities for the complex non-standard language and implicit causal relationships in coal mine accident reports, resulting in low-quality extracted information and being difficult to provide effective support for actual accident prevention and safety management. Therefore, there are still many deficiencies in the existing coal mine accident analysis systems in terms of information integration, causal relationship modeling, and intelligent decision support. There is an urgent need for a new technical solution to make up for these defects and improve the management efficiency and prevention ability of coal mine accident information. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide a coal mine accident information learning method and system to at least solve the problems that there are still many deficiencies in the existing coal mine accident analysis systems in terms of information integration, causal relationship modeling, and intelligent decision support.

[0005] To achieve the above object, a first aspect of the present invention provides a method for learning coal mine accident information, the method comprising: collecting historical coal mine accident information, and constructing a corresponding accident cause network model based on the historical coal mine accident information; performing corresponding information extraction based on the accident cause network model to construct a corresponding coal mine accident knowledge graph based on the extracted information; performing visualization processing on the coal mine accident knowledge graph and storing the coal mine accident knowledge graph; in response to a user's query request, returning answer information of a corresponding request type based on the stored coal mine accident knowledge graph and pushing the answer information to the user terminal.

[0006] Optionally, the performing corresponding information extraction based on the accident cause network model to construct a corresponding coal mine accident knowledge graph includes: performing corresponding information extraction based on the accident cause network model, and constructing a corresponding accident information knowledge graph database based on the information extraction result; in the accident information knowledge graph database, performing entity extraction, relationship extraction and attribute extraction respectively to complete the structuring process of each data and obtain structured data; constructing a corresponding knowledge graph based on the structured data as the coal mine accident knowledge graph.

[0007] Optionally, the rule of entity extraction is: performing word segmentation processing in the accident information knowledge graph database based on a natural language processing algorithm, and performing preset entity recognition based on the word segmentation result; obtaining an initial extracted entity based on the preset entity recognition; performing the same entity filtering on the initial extracted entity to obtain the entity extraction result.

[0008] Optionally, the rule of relationship extraction is: performing relationship recognition between each entity and other entities based on a preset entity relationship; combining two entities with a preset entity relationship into a relationship extraction result and indicating the corresponding entity relationship; wherein, the entity relationship includes any one or more of causal relationship, safety measure relationship, legal and regulatory relationship, precaution relationship and definition relationship.

[0009] Optionally, the rule of attribute extraction is: performing preprocessing on the data in the accident information knowledge graph database, and performing attribute annotation on the data after the preprocessing is completed; performing word segmentation processing on the data after the preprocessing is completed, and performing corpus annotation based on the word segmentation result; performing entity recognition on the data after the attribute annotation and corpus annotation are completed based on the Lattice-LSTM model, calculating the semantic similarity between the entities in the recognition result, and determining the relationship between the entities based on the semantic similarity; determining the attribute extraction result of each data based on the attribute annotation result, corpus annotation result and the relationship with other entities.

[0010] Optionally, performing word segmentation on the preprocessed data and performing corpus annotation based on the word segmentation results includes: performing word segmentation on the preprocessed data based on the HanLP toolkit, and constructing a corresponding word segmentation set based on the word segmentation structure; wherein, during the word segmentation process, performing corpus annotation on each word segmentation result based on the HanLP toolkit; the word segmentation set is represented as:

[0011] D = {O, S, B, M, E, I}

[0012] wherein, S represents an entity that is an independent word; I represents an internal character in a phrase; B represents the starting word of a combined word; M represents an internal word of a combined word; E represents the ending word of a combined word; O represents a specified part of speech.

[0013] Optionally, performing visualization processing on the coal mine accident knowledge graph and performing storage of the coal mine accident knowledge graph includes: performing visualization processing and storage of the coal mine accident knowledge graph based on the Neo4j graph database.

[0014] Optionally, the request types include any one or more of entity query, relationship query, and accident case reasoning; returning response information corresponding to the request type based on the stored coal mine accident knowledge graph and pushing the response information to the user side includes: in response to a user's query request, parsing the user's request signal to determine the corresponding request type; determining target-related information in the coal mine accident knowledge graph based on the request type and the corresponding request target, and performing result visualization based on the determined target-related information; pushing the visualization result to the user side.

[0015] In a second aspect of the present invention, a coal mine accident information learning system is provided, and the system includes: a collection unit for collecting historical coal mine accident information and constructing a corresponding accident cause network model based on the historical coal mine accident information; a graph construction unit for performing corresponding information extraction based on the accident cause network model to construct a corresponding coal mine accident knowledge graph based on the extracted information; a graph storage unit for performing visualization processing on the coal mine accident knowledge graph and performing storage of the coal mine accident knowledge graph; a query unit for, in response to a user's query request, returning response information corresponding to the request type based on the stored coal mine accident knowledge graph and pushing the response information to the user side.

[0016] On the other hand, the present invention provides a computer-readable storage medium, and instructions are stored on the computer-readable storage medium, and when the instructions run on a computer, the computer is caused to execute the above-mentioned coal mine accident information learning method.

[0017] Through the above technical solution, the solution of the present invention collects historical coal mine accident information and constructs an accident cause network model, effectively converting scattered unstructured accident data into structured information, and solving the problems of chaotic storage and inconvenient query of existing coal mine accident data. Based on this network model, information extraction is performed to construct a knowledge graph of coal mine accidents, which not only improves the accuracy of accident cause analysis, but also can deeply explore the causal relationships between accidents, providing more intuitive and systematic decision-making support for coal mine safety management. The visualization processing of the knowledge graph further enhances users' understanding and application of accident information, facilitating coal mine management personnel to quickly identify accident hazards and take measures. Through graph storage and response to user queries, flexible information retrieval can be achieved, quickly returning specific accident cases or relevant safety measures, providing real-time knowledge support for coal mine workers and management personnel. Generally speaking, this technical solution effectively improves the processing efficiency, analysis depth and query convenience of coal mine accident information, and provides an intelligent auxiliary decision-making tool for coal mine safety production.

[0018] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent specific embodiment part. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings are used to provide a further understanding of the embodiments of the present invention, and constitute a part of the specification, and are used to explain the embodiments of the present invention together with the following specific embodiments, but do not constitute a limitation to the embodiments of the present invention. In the drawings:

[0020] Figure 1 is a flowchart of the steps of a method for learning coal mine accident information provided by an embodiment of the present invention;

[0021] Figure 2 is a schematic diagram of the construction process of a knowledge database provided by an embodiment of the present invention;

[0022] Figure 3 is a schematic diagram of entity Chinese recognition based on Lattice-LSTM provided by an embodiment of the present invention;

[0023] Figure 4 is a schematic diagram of Lattice entity recognition provided by an embodiment of the present invention;

[0024] Figure 5 is a schematic diagram of causal relationship display based on IA01 entity nodes provided by an embodiment of the present invention;

[0025] Figure 6 is a schematic diagram of the display interface of an accident case reasoning module provided by an embodiment of the present invention;

[0026] Figure 7It is the system structure diagram of the coal mine accident information learning system provided by an embodiment of the present invention. Specific embodiments

[0027] The following further elaborates on the specific embodiments of the present invention with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not intended to limit the present invention.

[0028] Figure 1 It is the method flow diagram of the coal mine accident information learning method provided by an embodiment of the present invention. As Figure 1 shown, an embodiment of the present invention provides a coal mine accident information learning method, and the method includes:

[0029] Step S10: Collect historical coal mine accident information, and construct a corresponding accident cause network model based on the historical coal mine accident information.

[0030] Specifically, by collecting historical coal mine accident information and using modern data processing technology, a large amount of scattered coal mine accident data is gathered into a unified database. Historical coal mine accident information usually includes the time, location, accident type, losses caused, and investigation reports of the accident, etc. Most of this information is stored in an unstructured or semi-structured form, such as text reports, inspection records, etc., and there are problems such as scattered information, difficult retrieval, and processing. Therefore, first through data collection and cleaning, the accident information from different sources is transformed into standardized data for subsequent analysis and processing.

[0031] Based on the sorted historical accident information, an accident cause network model is further constructed. The accident cause network model systematically sorts out various reasons for the occurrence of accidents, revealing the causal relationship and its hierarchical structure among accident causes. This model not only considers the direct effect of a single factor, but also can reveal the interaction and linkage effect among multiple factors, thus providing a more comprehensive understanding of the occurrence of coal mine accidents. For example, the safety management system of the coal mine, the operation behavior of miners, the technical state of equipment, and environmental factors, etc. may all be intertwined to form a complex causal chain. By constructing an accident cause network model, these factors can be effectively summarized and organized according to the causal relationship, providing a scientific basis for further accident prevention.

[0032] Based on the solution of the present invention, this process significantly improves the efficiency of sorting and analyzing coal mine accident data, and solves the problems of difficult unified processing of historical accident information and low query efficiency. By establishing an accident cause network model, the key factors leading to accidents and their interrelationships can be intuitively displayed, thus providing more accurate decision-making support for the safety management of mining enterprises. This model can not only help coal mine managers discover potential safety hazards, but also provide a solid foundation for the subsequent construction of knowledge graphs, promoting the intelligent and systematic development of the coal mine accident prevention system.

[0033] Step S20: Perform corresponding information extraction based on the accident cause network model to construct a corresponding coal mine accident knowledge graph based on the extracted information.

[0034] Specifically, perform corresponding information extraction based on the accident cause network model, and construct a corresponding accident information knowledge graph database based on the information extraction results; in the accident information knowledge graph database, perform entity extraction, relationship extraction, and attribute extraction respectively to complete the structured processing of each data and obtain structured data; construct a corresponding knowledge graph based on the structured data as the coal mine accident knowledge graph.

[0035] In the embodiment of the present invention, as Figure 2 , information extraction is performed on the key nodes and connection relationships identified from the accident cause network model, and the extraction content usually includes the specific cause of the accident, the relevant responsible parties, the environmental factors of the accident, the preventive measures of the accident, etc. Through this information, a more comprehensive accident information knowledge graph database can be constructed, providing more accurate data support for subsequent decision-making and analysis.

[0036] In the accident information knowledge graph database, entity extraction, relationship extraction, and attribute extraction must be performed first. Entity extraction is to identify the key entities in the accident report through natural language processing techniques (such as word segmentation and named entity recognition), such as accident types, accident causes, affected objects, etc. Relationship extraction is used to determine the relevance between different entities, such as causal relationships and responsibility relationships, which can help establish the connections between accident causes and form structured data. Attribute extraction is to identify the specific characteristics of entities and relationships, such as the occurrence time of the accident, the scope of influence of the accident, safety measures, etc. This series of information extraction steps finally transform irregular and scattered text information into standardized and structured data, ensuring the efficiency and accuracy of subsequent data storage, query, and analysis.

[0037] Based on these structured data, the coal mine accident knowledge graph will contain multiple levels of information, such as accident type, cause, related factors, accident prevention measures and other nodes, and will be presented in the form of a graph database. This graph can not only intuitively show the overall picture of the causes of coal mine accidents, but also effectively show the causal relationship, time sequence and impact level between each accident, and enhance the depth of analysis of accident data.

[0038] Preferably, the entity extraction rules are as follows: perform word segmentation processing in the accident information knowledge graph database based on a natural language processing algorithm, and perform preset entity recognition based on the word segmentation results; obtain initial extracted entities based on preset entity recognition; perform identical entity filtering on the initial extracted entities to obtain entity extraction results.

[0039] In an embodiment of the present invention, the text data in the accident information database is segmented based on the natural language processing (NLP) algorithm. Word segmentation is a basic step in text preprocessing, which splits the coherent text into separate words or phrases so that subsequent entity recognition and relationship extraction can be performed more accurately. Through word segmentation, the system can effectively identify potential entities in the text, such as the type of accident, the time and place of the accident, the equipment involved, the personnel, and other information.

[0040] Next, based on the word segmentation results, the preset entity recognition process is performed. The preset entity recognition uses deep learning algorithms, such as named entity recognition (NER), to analyze the word segmentation results according to the set entity types (such as accident causes, responsible persons, preventive measures, etc.), and identify possible entities in the text. Through this process, the system can extract preliminary entities such as "gas explosion", "miners", and "safety measures" from the original text. For the entities initially extracted, perform identical entity filtering. The purpose of this process is to remove duplicate or irrelevant entities, merge multiple different text forms representing the same entity into a unified entity, and ensure the accuracy and consistency of the final entity set. After this filtering process, the entity extraction results obtained can effectively represent the key elements in the accident information, providing an important foundation for building an accurate and effective knowledge graph.

[0041] In a possible implementation, the entity categories of the coal mine gas explosion accident information knowledge graph are diverse and feature differentiated. After determining the composition of the entity node, it is necessary to expand the attributes of the entity, and extract the common attribute features obtained based on text mining as the attributes of the entity. The entity attributes of the coal mine gas explosion accident information knowledge graph are shown in Table 1.

[0042] Table 1 Entity attributes of coal mine gas explosion accident information knowledge graph

[0043]

[0044]

[0045] As shown in Table 1, the entity attributes of the coal mine gas explosion accident information knowledge graph can be summarized into six categories: definition, potential hazards, laws and regulations, safety measures, precautions, and causes. At the same time, considering the readability of the knowledge graph visualization display, the entity nodes are distributed in the form of a gravity map, and the entities are distinguished according to their nature and category colors.

[0046] Preferably, the rule for relationship extraction is: based on a preset entity relationship, identify the relationship between each entity and other entities; combine two entities with a preset entity relationship into a relationship extraction result and indicate the corresponding entity relationship; wherein, the entity relationship includes any one or more of causal relationship, safety measure relationship, laws and regulations relationship, precautions relationship, and definition relationship.

[0047] In the embodiment of the present invention, in the process of constructing the coal mine accident knowledge graph, relationship extraction is a key step in connecting various entities and revealing their interactions. Based on the preset entity relationship, relationship extraction first analyzes the identified entities to identify the relationship between an entity and other entities. These entity relationships include causal relationship, safety measure relationship, laws and regulations relationship, precautions relationship, and definition relationship, etc. This process accurately extracts various connections between different entities through natural language processing technology and machine learning algorithms, revealing the deep - seated causes of accidents and their prevention measures.

[0048] Specifically, relationship extraction first determines whether there is a specific relationship between two entities by identifying keywords and grammatical structures in the text. For example, in the analysis of coal mine accidents, the causal relationship between "poor mine ventilation" and "rising gas concentration" can be identified, or the safety measure relationship between "safety equipment" and "preventing accidents". In this way, the system can automatically identify various relationships in the original accident data and combine the relevant entities to form accurate relationship pairs.

[0049] For each pair of entities with a preset relationship, the system combines them into a relationship and indicates the type of the relationship. For example, if the causal relationship between "equipment failure" and "gas leakage" is found in the accident description, the system combines the two and labels it as "causal relationship". This relationship extraction process can cover multiple types of relationships at the same time, providing support for the subsequent construction of a more complex coal mine accident knowledge graph.

[0050] In a possible implementation, in the knowledge graph related to coal mine accidents, entities can be accidents. In the knowledge graph of coal mine gas explosion accident information, relationships can include multiple types, which are used to describe different connections or associations between entities. The main relationships include:

[0051] 1) Causal relationship: Describes the relationship between a coal mine accident and its potential causes. For example, poor mine ventilation may lead to an increase in gas concentration.

[0052] 2) Safety measure relationship: The connection between an entity and safety measures, indicating that specific measures can reduce the likelihood of an accident, such as specific safety equipment or safety management procedures.

[0053] 3) Laws and regulations relationship: The connection between an entity and applicable laws and regulations, including regulations and policies related to coal mine safety management.

[0054] 4) Precautions relationship: Describes the matters that coal mine workers or relevant parties should pay attention to, such as preventive measures to be taken in specific situations.

[0055] 5) Definition relationship: The connection between an entity and its definition or description, such as the definition of a specific term or concept.

[0056] Preferably, the rules for attribute extraction are as follows: Perform preprocessing on the data in the accident information knowledge graph database, and perform attribute annotation on the data after the preprocessing is completed; perform word segmentation on the data after the preprocessing is completed, and perform corpus annotation based on the word segmentation results; perform entity recognition on the data after the attribute annotation and corpus annotation are completed based on the Lattice-LSTM model, calculate the semantic similarity between the entities in the recognition results, and determine the relationship between the entities based on the semantic similarity; determine the attribute extraction results of each data based on the attribute annotation results, corpus annotation results, and the relationship with other entities. Specifically, it includes the following steps:

[0057] 1) Data preprocessing: The corpus mainly comes from coal mine accident investigation reports and the main coal mine safety regulations. At the same time, relevant information related to entities is obtained through web crawling of data literature and Baidu Encyclopedia. Remove special symbols in the text data, unify the text format, and add attribute tags to the relevant explanations or specific parts of the collected entities to describe their characteristics or properties. For example, the three elements of a gas explosion accident are labeled as "occurrence conditions of the accident", etc., and corresponding tags are added to each category.

[0058] 2) Word Segmentation: Information extraction is a crucial step in extracting entities, attributes, and relationships from web pages or texts. Word segmentation is usually the first step in information extraction, which decomposes a continuous text string into distinct words or sentences. The natural language processing toolkit HanLP, which is oriented towards Chinese text processing, is used to perform word segmentation on the expected text and construct a set D to represent.

[0059] D = {O, S, B, M, E, I}

[0060] During the word segmentation process, HanLP uses a method similar to the BIOES annotation method to annotate words, which can handle Chinese texts well and identify phrases and word groups as accurately as possible. This annotation method uses the marking rules of "S", "B", "M", and "E" when processing Chinese sequences, and the specific representations are as follows: S (Single): Represents an entity that is an independent word. I (Inside): Represents the internal characters in a word group. B (Begin): Represents the starting word of a combined word. M (Middle): Represents the internal word of a combined word. E (End): Represents the ending word of a combined word. In addition, in addition to marking the start, middle, and end positions of combined words, HanLP also annotates some specified parts of speech (such as auxiliary words, prepositions, etc.), usually represented by "O".

[0061] 3) Corpus Annotation: When it comes to word segmentation and corpus annotation in specific vertical fields such as coal mine accident texts, due to the lack of a dedicated and publicly available training dataset in this field. The particularity and professionalism of the coal mine field make the current standardized general corpus unable to fully meet the needs of word segmentation and corpus annotation. Therefore, when processing texts related to coal mine accidents, it is necessary to rely on the current crawled data format and supplement it with manual annotation to adjust and correct the annotation results.

[0062] 4) Entity Recognition with Lattice-LSTM Model: The Lattice-LSTM model, as an improved BiLSTM-CRF model based on the character level, aims to optimize the fusion between character-level information and word-level information. Compared with traditional word-based models, it makes better use of character-level features and combines character subsequences matched in the self-constructed dictionary, such as Figure 3 "coal mine gas" exemplified in. The input of the model includes the character sequence c_1, c_2,..., c_m and the character subsequences matched with the dictionary. This fusion method enables the model to more effectively capture the association between character-level and word-level information and improves the entity recognition ability for complex sequences in languages such as Chinese.

[0063] In the Lattice-LSTM model, the word sequence not only contains character-level information but also propagates this information to each character vector through multiple cyclic paths Through Figure 4 the structure shown, the information sources carried by the character vectors cover multiple levels of feature representations, such as ("Si"), ("gas"), and ("coal mine gas"). This design allows the model to effectively capture word-level and phrase-level information and provides more comprehensive and richer feature representations during entity recognition. Partial entity recognition results are shown in Table 2.

[0064] Table 2 Recognition Results of Entities in Example Sentences

[0065]

[0066] 5) Calculate semantic similarity to determine relationships: In the entity recognition of coal mine accident information, the accurate recognition of relationships between entities is achieved by analyzing the semantic similarity of the respective senses of the entities. This study uses the HowNet knowledge base as a tool for calculating semantic similarity. For two entities w1 and w2, the relationship between the entities is determined by calculating the similarity between their respective different senses.

[0067] Step S30: Perform visualization processing on the coal mine accident knowledge graph and store the coal mine accident knowledge graph.

[0068] Specifically, perform the visualization processing and storage of the coal mine accident knowledge graph based on the Neo4j graph database.

[0069] In the implementation of the present invention, in view of the fact that the gas explosion accident information knowledge graph requires frequent queries and updates for entity and attribute relationships, the present invention selects the Neo4j graph database as the underlying storage and query platform. Nodes and relationships can both carry one or more labels, which exist in the form of key-value pairs of "name: value" for distinguishing and identifying different instances. Neo4j uses Cypher as a declarative query language for processing graph data, mainly including operations such as the creation, query, update, and deletion of nodes and relationships. Based on the previous steps of the gas explosion accident cause network model and Lattice-LSTM text knowledge extraction, a gas explosion accident information knowledge graph is constructed. The Neo4j Browser visually displays the connections of these nodes and relationships so that users can more intuitively understand and analyze the relevance of accident information. When querying the causal relationship of a certain node, the causal relationship instruction "MATCH p = ()-[r:`causal relationship`]->() RETURN p LIMIT 25" can be executed. For example, when querying all first-level causal relationships based on the IA01 (failure to check gas concentration) entity, the node range is controlled to display 25 entities, such asFigure 5 As shown. The 10 accident causes directly leading to IA01.

[0070] Step S40: In response to a user's query request, return answer information corresponding to the request type based on the stored coal mine accident knowledge graph, and push the answer information to the user terminal.

[0071] Specifically, the request types include any one or more of entity query, relationship query, and accident case reasoning; returning answer information corresponding to the request type based on the stored coal mine accident knowledge graph and pushing the answer information to the user terminal includes: in response to the user's query request, parsing the user's request signal to determine the corresponding request type; based on the request type and the corresponding request target, determining target-related information in the coal mine accident knowledge graph, and performing result visualization based on the determined target-related information; pushing the visualization result to the user terminal.

[0072] In the embodiment of the present invention, by inputting a query request, the system can quickly return the required answer information according to the stored coal mine accident knowledge graph and push it to the user terminal, ensuring that managers and coal miners can conveniently obtain the required safety knowledge and accident-related data.

[0073] Specifically, the system supports multiple query types, including entity query, relationship query, and accident case reasoning. Entity query mainly helps users quickly find information related to specific accident types, accident causes, accident responsible persons, etc. For example, users can query all relevant entities of a certain type of accident, such as information on the accident type, victims, responsible persons, etc. of a "gas explosion accident". Relationship query can reveal the relationships between different entities, helping users understand the associations between a certain entity and other factors. For example, the causal relationship between a certain accident cause and the occurrence of an accident or its association with preventive measures. Accident case reasoning is that after the user inputs a specific accident type or cause, the system infers the occurrence path and results of similar accidents through the causal relationships and historical cases in the graph, assisting users in understanding the accident occurrence mechanism and preventive measures to be taken.

[0074] When the system receives a query request from a user, it first analyzes the user's request signal to clarify the request type and target. For example, the user may request "analysis of the possibility of an accident caused by a certain accident cause" or query "whether a certain safety measure can effectively prevent a specific accident". The system will retrieve relevant information in the knowledge graph according to the request type, extract the node and relationship information related to the query target, and perform corresponding visualization processing. Visualization processing can display the query results in a graphical interface, such as showing accident entities and their causal relationships through a gravity graph, or showing various prevention measures and their effects through an accident chain. Finally, the system pushes these visualization results to the user side, enabling the user to intuitively understand the accident information and make corresponding safety decisions based on the query results.

[0075] In a possible implementation manner, the accident cause reasoning and accident information learning are realized through three main functions: front-end information management, knowledge query, and path display.

[0076] 1) Entity query function: The entity query module displays all entities that have a causal relationship with the entity by inputting the keyword of the query entity, that is, all factors that caused the accident in the historical data of coal mine accidents, and performs visual front-end display in the form of a three-dimensional gravity graph. For example, in an entity query interface, the upper part is the query condition area, which provides reference examples for querying, and the lower part is the display area of the query results. The right side is the system principle of the entity query module. The figure shows the entity query information of "poor awareness of observing regulations", where the pink node is the query node, and the green nodes are other entity nodes that have a causal relationship with the query node, a total of 10. In addition, left-clicking on an entity can arbitrarily drag the position of the entity, and right-clicking on an entity can expand the display range of the entity to display the first-level causal relationship network centered on this node.

[0077] 2) Relationship query: The relationship query module displays all attribute relationships starting from a certain node. This module provides three query methods for attribute relationships: specifying a certain entity node for query, such as the "gas concentration not checked" entity shown in the figure. In this way, all attribute relationships related to this node will be presented. Specifying a certain entity and a certain relationship for query, such as specifying the "gas concentration not checked" entity and the "causal relationship". In this way, only the causal relationship network under this entity will be displayed. Specifying two entities to query whether there is an attribute relationship between them, such as whether there is a connection between "gas concentration not checked" and "gas inspector absent from post".

[0078] 3) Accident case reasoning: Based on the constructed entity causal network, the case reasoning module queries all accident cause paths starting from a certain entity. This module provides two query methods for attribute relationships: (1) Single related node query. For example, the entity "gas concentration not checked" shown in the figure. In this way, all accident cause paths starting from this node will be presented. (2) Dual-node query, to query whether there is a connected path between the two. For example, Figure 6 shows the display interface of the accident case reasoning system, Figure 6 which is the case reasoning system interface with the "gas concentration not checked" as the reasoning node, divided into upper and lower parts. The upper part is the information input part, and the lower part is the visualization display part of related nodes. Among them, the blue nodes are input nodes, the red nodes are first-level relationship nodes, the green nodes are second-level relationship nodes, and the gray nodes are third-level relationship nodes. Different levels of nodes can be manually selected to be displayed or hidden. For the convenience of accident path display and analysis, the secondary accident network path regarding "gas concentration not checked" is shown here.

[0079] Based on the solution of the present invention, the present invention has established a set of process-based accident information extraction method and developed a coal mine accident information learning system for coal mine workers to understand the accident cause mechanism. This system can be applied to the safety training of coal mine enterprises and accident information learning, enabling coal mine workers to more conveniently understand the propagation path of coal mine gas explosion accidents and more systematically master the knowledge related to coal mine gas explosion accidents. It provides a tool for the intelligent application of network analysis and the accident information learning of coal mine workers.

[0080] Figure 7 is the system structure diagram of the coal mine accident information learning system provided by an embodiment of the present invention. As Figure 7 shown, an embodiment of the present invention provides a coal mine accident information learning system, which includes: a collection unit for collecting historical coal mine accident information and constructing a corresponding accident cause network model based on the historical coal mine accident information; a graph construction unit for performing corresponding information extraction based on the accident cause network model to construct a corresponding coal mine accident knowledge graph based on the extracted information; a graph storage unit for performing visualization processing on the coal mine accident knowledge graph and storing the coal mine accident knowledge graph; a query unit for responding to a user's query request, returning answer information of the corresponding request type based on the stored coal mine accident knowledge graph, and pushing the answer information to the user terminal.

[0081] An embodiment of the present invention also provides a computer-readable storage medium, which stores instructions thereon that, when running on a computer, cause the computer to execute the above-mentioned coal mine accident information learning method.

[0082] Those skilled in the art can understand that all or part of the steps in the methods of the above-described embodiments can be completed by instructing relevant hardware through a program, which is stored in a storage medium and includes several instructions to enable a single-chip microcomputer, a chip, or a processor to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs.

[0083] The optional embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the technical concept scope of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention. Additionally, it should be noted that, in the above specific embodiments, the various specific technical features described can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention do not separately describe various possible combination manners.

[0084] Furthermore, any combination can be made between various different embodiments of the present invention as long as it does not violate the idea of the embodiments of the present invention, and it should also be regarded as the content disclosed by the embodiments of the present invention.

Claims

1. A method for learning coal mine accident information, characterized in that: The method comprises: Collect historical coal mine accident information, and build a corresponding accident cause network model based on the historical coal mine accident information; Perform corresponding information extraction based on the accident cause network model to construct a corresponding coal mine accident knowledge graph based on the extracted information; Performing visualization processing on the coal mine accident knowledge graph and performing coal mine accident knowledge graph storage; In response to the user's query request, answer information corresponding to the request type is returned based on the stored coal mine accident knowledge graph, and the answer information is pushed to the user end.

2. The method according to claim 1, characterized in that The performing corresponding information extraction based on the accident cause network model to construct a corresponding coal mine accident knowledge graph based on the extracted information includes: Perform corresponding information extraction based on the accident cause network model, and construct a corresponding accident information knowledge graph database based on the information extraction result; In the accident information knowledge graph database, entity extraction, relationship extraction and attribute extraction are respectively performed to complete structured processing of each data to obtain structured data; The corresponding knowledge graph is constructed based on the structured data as the knowledge graph of coal mine accidents.

3. The method according to claim 2, characterized in that The rules for entity extraction are: Perform word segmentation in the accident information knowledge graph database based on the natural language processing algorithm, and perform preset entity recognition based on the word segmentation results; Obtaining initial extracted entities based on preset entity recognition; Perform the same entity filtering on the initial extracted entities to obtain the entity extraction results.

4. The method according to claim 2, characterized in that: The relationship extraction rules are: Identify the relationship between each entity and other entities based on the preset entity relationship; Combine two entities with a preset entity relationship into one relationship collection result and indicate the corresponding entity relationship; wherein, The entity relationships include: Any one or more of the causal relationship, safety measure relationship, legal and regulatory relationship, precaution relationship and definition relationship.

5. The method according to claim 2, characterized in that: The attribute extraction rules are: Preprocess the data in the accident information knowledge graph database and perform attribute labeling on the preprocessed data; Perform word segmentation on the preprocessed data and perform corpus annotation based on the word segmentation results; Based on the Lattice-LSTM model, entity recognition is performed on the data with completed attribute annotation and corpus annotation, and the semantic similarity between the entities in the recognition results is calculated, and the relationship between the entities is determined based on the semantic similarity; The attribute extraction results of each data are determined based on the attribute annotation results, corpus annotation results and the relationship with other entities.

6. The method according to claim 5, characterized in that The performing word segmentation processing on the pre-processed data and performing corpus annotation based on the word segmentation results includes: Based on the HanLP toolkit, the preprocessed data is segmented and the corresponding segmentation set is constructed based on the segmentation structure. During the word segmentation process, corpus annotation is performed on each word segmentation result based on the HanLP toolkit; The word segmentation set is expressed as: D={O,S,B,M,E,I} Among them, S represents an entity that is an independent word; I represents an internal character in a phrase; B represents the starting word of the compound word; M represents the internal word of the compound word; E represents the ending word of the combination; O indicates the specified part of speech.

7. The method according to claim 1, characterized in that The performing visualization processing on the coal mine accident knowledge graph and performing coal mine accident knowledge graph storage includes: The coal mine accident knowledge graph visualization processing and coal mine accident knowledge graph storage are performed based on the Neo4j graph database.

8. The method according to claim 1, characterized in that Request types include: Any one or more of entity query, relationship query and accident case reasoning; The method of returning answer information corresponding to the request type based on the stored coal mine accident knowledge graph and pushing the answer information to the user end includes: In response to a query request from a user, parsing the user's request signal to determine a corresponding request type; Based on the request type and the corresponding request target, target related information is determined in the coal mine accident knowledge graph, and result visualization is performed based on the determined target related information; Push the visualization results to the user end.

9. A coal mine accident information learning system, characterized in that: The system comprises: A collection unit, used to collect historical coal mine accident information and construct a corresponding accident cause network model based on the historical coal mine accident information; A graph construction unit, used to perform corresponding information extraction based on the accident cause network model, so as to construct a corresponding coal mine accident knowledge graph based on the extracted information; A graph storage unit, used for performing visualization processing on the coal mine accident knowledge graph and performing coal mine accident knowledge graph storage; The query unit is used to respond to the user's query request, return answer information corresponding to the request type based on the stored coal mine accident knowledge graph, and push the answer information to the user end.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the coal mine accident information learning method described in any one of claims 1-8.

Images