A semantic modeling method and system for highway construction safety training

Abstract

The application relates to the technical field of semantic modeling, and discloses a semantic modeling method and system for highway construction safety training, which comprises the following steps: obtaining templates related to construction safety from a template source; abstracting a plurality of core concept classes and constructing an ontology framework containing the core concept classes and hierarchical relationships thereof; defining object attributes for connecting different concept classes and data attributes for describing the characteristics of the concept classes, so as to form a semantic relationship network of the ontology; constructing semantic rules for logical reasoning based on a semantic rule language; mapping structured knowledge clauses to instances in batches, generating instance triples, and filling the instance triples into the ontology framework; based on the constructed ontology, semantic rules and instance triples, performing dynamic risk identification on an input construction task, and outputting corresponding safety training recommended measures; and the application solves the problems of simple logical conditions and poor robustness of the existing semantic modeling method of safety training.

Description

Technical Field

[0001] This invention relates to the field of semantic modeling technology, and in particular to a semantic modeling method and system for highway construction safety training. Background Technology

[0002] Highway construction projects are large-scale, complex, and involve numerous risk factors. Safety training for construction workers directly impacts accident prevention and project quality. In recent years, the introduction of corpus and knowledge graph methods has led to significant progress in the digitization of construction safety knowledge. However, current research primarily focuses on corpus processing and shallow information extraction, with shortcomings in semantic modeling and reasoning applications. While existing corpus resources exist, most remain at the clause or sentence level, lacking a unified ontology hierarchy and failing to support a systematic expression of construction activities, risk sources, and mitigation measures. Existing knowledge graph construction methods often employ triple representations, describing "who-where-what," but cannot support reasoning under complex logical conditions. Furthermore, existing research often lacks instance-based input from real construction models, causing knowledge graphs to remain at the schema level, unable to map to specific training tasks, and struggling to provide differentiated, job-specific training recommendations. Current safety knowledge graph research largely focuses on structural display or static retrieval, lacking systematic validation based on reasoning engines and ontology quality evaluation tools, thus failing to demonstrate its effectiveness and generalizability in safety training scenarios. It is evident that existing semantic modeling methods for security training suffer from problems such as simple logical conditions and poor robustness. Summary of the Invention

[0003] This invention provides a semantic modeling method and system for highway construction safety training, which solves the problems of simple logical conditions and poor robustness in existing semantic modeling methods for safety training.

[0004] To achieve the above objectives, the present invention employs the following technical solution:

[0005] In a first aspect, the present invention provides a semantic modeling method for highway construction safety training, comprising: S1. Obtain templates related to construction safety, and perform optical character recognition, regular expression cleaning, and clause segmentation on the templates to form structured knowledge clauses; S2. Based on the three-dimensional logical framework of operation control, risk prevention and emergency response, multiple core concept classes are abstracted from structured knowledge clauses, and an ontology framework containing the core concept classes and their hierarchical relationships is constructed. S3. Define object attributes used to connect different concept classes and data attributes used to describe the characteristics of concept classes to form a semantic relationship network of the ontology; S4. Construct semantic rules for logical reasoning based on semantic rule language in semantic relation network; S5. Using semantic rules, the structured knowledge terms are mapped to instances in batches, and instance triples are generated and populated into the ontology framework. S6. Based on the filled ontology framework, semantic rules, and instance triples, dynamically identify risks in the input construction tasks and output corresponding safety training recommendations.

[0006] Optionally, in step S1, optical character recognition includes: performing optical character recognition on scanned documents in templates related to construction safety, converting the scanned documents into processable text; regular expression cleaning includes: removing symbols unrelated to the text content from the processable text; and clause segmentation includes: constructing clause feature words, and performing semantic segmentation on the regular expression-cleaned processable text based on the clause feature words.

[0007] Optionally, in step S2, the core concept classes include: construction location, construction activities, work requirements, hazard sources, potential risks, preventive measures, accident types, and handling measures; The ontology framework, which includes the core concept classes and their hierarchical relationships, satisfies the following relation: ; In the formula, This refers to the core of safety training for construction workers in highway construction. This represents ontology concept classes that have the same characteristics. This represents the relationship between object attributes of different classes. Data attributes that represent the relationship between classes and data values. Represent semantic constraint rules in domain knowledge. Represents a specific instance of a class in the ontology.

[0008] Optionally, the ontology framework includes ontology concept classes with the same characteristics. Defined as three concept sets: operation control, risk prevention, and emergency response, the ontology concept class is expressed according to the following relation: ; In the formula, For work control, For risk prevention, For emergency response; Concept set A set of top-level concept classes used to construct the operation control dimension; concept set The top-level concept set used to construct the risk prevention dimension; concept set A set of top-level concept classes used to construct the emergency response dimension; concept set , , They respectively satisfy the following relations: ; ; ; In the formula, Construction location, For construction activities, As required by the assignment, As a source of danger, As a potential risk, As a precaution, Accident type As a measure to deal with the situation; Concept set , , Each class contains j second-level subclasses, and the second-level subclasses satisfy the following relationship: ; ; ; In the formula, This is a second-level concept class in the task control dimension. 'j' represents the top-level class number, and 'j' represents the second-level class number.

[0009] Optionally, in step S3, the data attributes include: name attributes, encoding attributes, description attributes, risk level and legal basis, and the object attributes are location attribute words, behavior attribute words and result-oriented words; Relationships between object attributes of classes The semantic relations in the equation are abstracted into triples, which satisfy the following relation: ; In the formula, Let r be the starting class of the relation, and r be the predicate attribute. For the target class, A collection of domain concept classes. It is in triplet form; And based on semantic function, classify the object attribute relationships between classes. The object attribute relationships are classified into three categories: spatial relationships, normative constraint relationships, and causal relationships. Based on the above classification of object attribute relationships, the triplet forms satisfy the following relationships: Formal expression of triples in spatial relations: ; In the formula, It is a triplet form under spatial relations. It is a spatial relation predicate. Words with location attributes; Formal representation of triples under normalized constraints: ; In the formula, To standardize the form of triples under constraints, To standardize the predicates of constraint relations, All are behavioral attribute terms; Formal representation of triples under causal relationships: ； In the formula, In the form of a triple under causal relationship, It is a predicate of cause and effect. All are result-oriented terms; Based on the above object attribute relationships, a semantic relationship network of the ontology is formed.

[0010] Optionally, in step S4, the semantic rules are constructed using the Semantic Web Rule Language, wherein the general form of the Semantic Web Rule Language rules satisfies the following relation: ; In the formula, This indicates the relationship between object attributes in the premise of the rule. Indicates the conclusion of the rule. For Semantic Web rule language rules, It represents the relationship between object attributes between classes.

[0011] Optionally, in step S5, generating instance triples includes: Specific instances of classes in the ontology An instance triple is represented as a triple structure consisting of a subject entity, a semantic relation predicate, and an object entity. The formal definition of an instance triple satisfies the following relation: ; In the formula, The semantic entities in the instance layer are derived from the ontology concept class. The instantiation result, Represents a semantic object attribute, used to define the semantic relationship between two entities.

[0012] Secondly, embodiments of this application provide a semantic modeling system for highway construction safety training, used to implement the semantic modeling method for highway construction safety training as described in the first aspect, the system comprising: The knowledge acquisition and preprocessing module is used to acquire and process construction safety templates from template sources, and to perform optical character recognition, regular expression cleaning and clause segmentation on the templates to form structured knowledge clauses. The semantic ontology construction module is used to abstract multiple core concept classes based on the three-dimensional logical framework of operation control-risk prevention-emergency response, and to construct an ontology framework containing the core concept classes and their hierarchical relationships. A semantic rule base is used to connect object attributes of different concept classes and data attributes used to describe the characteristics of concept classes to form a semantic relation network of the ontology, and to construct semantic rules for logical reasoning based on the semantic rule language in the semantic relation network. The reasoning engine is used to load and execute rules in the semantic rule base to perform logical reasoning. The instantiation module is used to batch map structured knowledge terms to instances and generate instance triples to populate the ontology framework. The training recommendation engine is used to dynamically identify risks in input construction tasks based on the constructed ontology, semantic rules, and instance triples, and output corresponding safety training recommendations.

[0013] Beneficial effects: This invention provides a semantic modeling method for highway construction safety training, constructing a unified semantic ontology framework covering a three-dimensional logic of "operation control - risk prevention - emergency response" and containing eight core concepts. It elevates the clause-level knowledge scattered across over a hundred templates from its original, unstructured textual form into a machine-understandable model with a clear hierarchical structure and rich semantic relationships. This fundamentally changes the traditional situation where safety training knowledge remains only at the corpus or scattered triplet level, achieving a high degree of abstraction and systematic organization of human professional knowledge, laying a solid data foundation for intelligent applications. By introducing SWRL semantic rules and the HermiT inference engine, this invention endows the system with powerful logical reasoning capabilities. The system can simulate the thinking of domain experts, performing chain-like deductions on multiple constraints and causal chains. This makes the system no longer a passive knowledge base, but an "intelligent assistant" capable of proactively identifying risks, deriving consequences, and recommending measures, significantly outperforming traditional training systems that only support keyword matching and static retrieval. Using Protégé and Cellfie plugins, it batch maps regulatory clauses into 1,557 instance triples. This process transforms the ontology from an abstract "schema layer" into a grounded "instance layer" knowledge graph containing knowledge of real-world scenarios such as bridges, tunnels, and foundation pits. This large-scale instantiation ensures a close correspondence between the semantic model and real construction tasks, greatly enhancing the system's usability and relevance in practical training applications. The ONTOQA framework and HermiT inference engine are used to systematically verify the ontology's quality and consistency, addressing the logical simplicity issue present in existing semantic modeling methods. Evaluation results show excellent performance in key indicators such as relation richness, attribute richness, and class richness, indicating that the knowledge system constructed in this invention is structurally balanced and reasonable, information-rich and complete, and logically consistent. This provides a fundamental guarantee for the system's stable and reliable operation in practical applications and improves the robustness of the semantic modeling method. Attached Figure Description

[0014] Figure 1 This is a flowchart of a semantic modeling method for highway construction safety training according to a preferred embodiment of the present invention; Figure 2 This is a schematic diagram of class and subclass hierarchy modeling in a preferred embodiment of the present invention; Figure 3 This is a schematic diagram of the object attributes and semantic relationships of a top-level conceptual class provided in a preferred embodiment of the present invention. Detailed Implementation

[0015] The technical solution of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms "an" or "a" and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms "connected" or "linked" and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up," "down," "left," "right," etc., are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship also changes accordingly.

[0017] Please see Figure 1 This application provides a semantic modeling method for highway construction safety training, including: S1. Obtain templates related to construction safety, and perform optical character recognition, regular expression cleaning, and clause segmentation on the templates to form structured knowledge clauses; S2. Based on the three-dimensional logical framework of operation control, risk prevention and emergency response, multiple core concept classes are abstracted, and an ontology framework containing the core concept classes and their hierarchical relationships is constructed. S3. Define object attributes used to connect different concept classes and data attributes used to describe the characteristics of concept classes to form a semantic relationship network of the ontology; S4. Construct semantic rules for logical reasoning based on semantic rule language; S5. Map the structured knowledge terms to instances in batches, generate instance triples, and populate them into the ontology framework; S6. Based on the constructed ontology, semantic rules, and instance triples, dynamically identify risks in the input construction tasks and output corresponding safety training recommendations.

[0018] Optionally, in step S1, the template source includes authoritative templates obtained from the Ministry of Transport, the Ministry of Housing and Urban-Rural Development, the Ministry of Emergency Management, and the standards service platform.

[0019] In the above embodiments, the templates are mainly obtained from template sources. Authoritative templates are obtained from the Ministry of Transport, the Ministry of Housing and Urban-Rural Development, the Ministry of Emergency Management, and the standards service platform. They cover laws, administrative regulations, and implementation standards. Among them, legal texts have the highest legal effect and are applicable to safety production activities; administrative regulations provide detailed requirements for safety production management in specific industries or fields; departmental regulations provide detailed requirements for safety production management in specific industries or fields; administrative normative documents are documents with universal binding force and are applicable in specific regions or industries; industry standards are implemented in specific industries; and local standards are applicable to engineering construction activities under specific conditions in the region.

[0020] The authoritative templates cited above are for illustrative purposes only and are not intended to be limiting. Any template related to construction safety can be applied in this embodiment.

[0021] Optionally, in step S2, the core concepts include: construction location, construction activities, work requirements, hazard sources, potential risks, preventive measures, accident types, and handling measures; The ontology framework, which includes the core concept classes and their hierarchical relationships, satisfies the following relation: (1) In the formula, This refers to the core of safety training for construction workers in highway construction. This represents ontology concept classes that have the same characteristics. This represents the relationship between object attributes of different classes. Data attributes that represent the relationship between classes and data values. Represent semantic constraint rules in domain knowledge. Represents a specific instance of a class in the ontology.

[0022] In the above embodiments, based on the three-dimensional logical framework of "operation control – risk prevention – emergency response", several core concepts are abstracted from the authoritative template. These core concepts include location, activity, requirement, hazard, risk, prevention, accident, and response.

[0023] The number of core concepts listed above is for illustrative purposes only and is not intended to be limiting. In other embodiments, the number of core concepts may be increased or decreased according to actual needs.

[0024] This embodiment names the ontology model for highway construction worker safety training as STO-HCW (Safety Training Ontology for Highway Construction Workers), serving as the structured semantic carrier model for the STC-HCW corpus. Three key aspects are considered in STO-HCW modeling: first, guided by practical application needs, focusing on the relationship between the knowledge requirements of construction workers during the learning process and the risk factors in the construction phase; second, ensuring the ontology structure is flexible and scalable, supporting content updates and expansions; and third, establishing a scientifically sound knowledge system, fully inheriting and reusing the existing ontology's knowledge system, and promoting knowledge integration and sharing among different construction safety knowledge ontology systems. By introducing a five-tuple structure modeling method, the domain ontology structure is abstracted into a five-tuple of concepts, relations, attributes, rules, and instances, whose formal representation is shown in equation (1).

[0025] Optionally, the ontology framework includes ontology concept classes with the same characteristics. Defined as a set of three concepts: operation control, risk prevention, and emergency response, expressed in the following relational formula: (2) In the formula, For work control, For risk prevention, For emergency response; Concept set A set of top-level concept classes used to construct the operation control dimension; concept set The top-level concept set used to construct the risk prevention dimension; concept set The three top-level concept sets used to construct the emergency response dimension satisfy the following relationships: (3) (4) (5) In the formula, Construction location, For construction activities, As required by the assignment, As a source of danger, As a potential risk, As a precaution, Accident type As a measure to deal with the situation; The three top-level concept sets above each contain j second-level subclasses, and the second-level subclasses satisfy the following relation: (6) (7) (8) In the formula, This is a second-level concept class in the task control dimension. 'j' represents the top-level class number, and 'j' represents the second-level class number.

[0026] In the above embodiments, to account for the complexity of the knowledge system in the field of safety training for highway construction workers, the keyword system was collected according to the three dimensions of the keyword system during the STC-HCW corpus collection. The ontology concept is defined as three sets of concepts: operation control, risk prevention, and emergency response. The three sets of concepts are shown in equation (2).

[0027] , , Further subdividing, we define the top-level concept set for the dimension of task control. , , See formulas (3)-(5) for details.

[0028] As the top-level conceptual class for the task control dimension, it contains j second-level subclasses. Similarly, a subset of class concepts based on the dimensions of risk prevention and emergency response can be constructed. and See formulas (6)-(8) for details.

[0029] In this implementation, the source, meaning, and function of the top-level concept class are shown in Table 1:

[0030] An ontology class represents a set of concepts with the same characteristics, and the class hierarchy is defined by the relationships between classes. In the Protégé ontology modeling tool, a class is represented by "Class", the class hierarchy is expressed by "Classhierachy", and the inheritance relationship is defined by "Subclass of", which represents the logical hierarchy between hierarchical concepts. The formal expression is shown in formula (9).

[0031] 1,8); (9) In the formula: and It refers to the hierarchical relationship between the same element in the ontological concept set; This indicates the inheritance relationship between hierarchical concepts, such as Construction location and Roadbed engineering Concepts such as road engineering have a relationship of inheritance and exhibit the same attributes.

[0032] Set "Owl:Thing" as the parent class of all classes. Based on the top-level concept ontology determined in the previous step, the top-level concept class includes eight items: construction location, construction activity, and operation requirements; hazard source, potential risk, and preventive measures; accident type and response measures. Each top-level concept class has corresponding subclasses. Subclasses can be further subdivided to form a multi-level class hierarchy. In Protégé, any class (except the top-level class) is a subclass of a certain parent class and can also be a parent class of other classes. This recursive hierarchy can be extended indefinitely, limited only by practical application needs and modeling complexity.

[0033] The ontology knowledge system for construction worker safety training constructed in this invention is divided into three dimensions: operation control, risk management, and emergency response. Each dimension has multiple top-level concept classes, as detailed in the appendix. Figure 2 .

[0034] In this embodiment, the operation control dimension includes three top-level concepts: construction location, construction activities, and operation requirements. .

[0035] The construction location category refers to both where construction activities take place and where construction risks occur, and it changes as construction activities evolve. This top-level concept defines "where" construction workers perform their work. For construction location categories, the highway construction project can be subdivided using a Work Breakdown Structure (WBS). First, the project can be subdivided into unit project subcategories, such as... Roadbed engineering Road surface engineering Bridge engineering Tunnel engineering...etc., received Subclass concept set. Then, using the unit project as the parent class, further subdivide it into subclasses of sub-items of engineering and subclasses of measures / projects. For example, Roadbed engineering can be broken down into Cutaway Embankment, Subcategories such as slopes, The parent class of bridge engineering can be decomposed into Pile foundation, cofferdam, platform, pier,..., Subcategories, etc., can be further divided in detail with reference to relevant engineering construction specifications.

[0036] The construction activity category defines the specific "what" a construction operation entails. It consists of "construction of sub-projects" and "construction of temporary works," encompassing both construction activities related to the physical structure of a highway (e.g., "pier cap construction") and preliminary construction activities not related to the physical structure (e.g., "scaffolding construction"). It can be further subdivided based on construction sequence. A set of subclass concepts. For example... Foundation construction activities include Break the pile head, Reinforcing bar installation Template installation Construction activities such as concrete pouring. Specific details should be determined in accordance with relevant specifications.

[0037] The "Work Requirements" class defines "how" construction workers should perform their tasks, i.e., the standards, specifications, and operating procedures that must be followed during implementation. It can be divided into subclasses such as construction preparation, technical points, material requirements, and quality requirements. Subclass concept set.

[0038] In this embodiment, the risk prevention dimension includes three top-level concepts: hazard source, potential risk, and preventive measures. .

[0039] The top-level concept of the hazard source class defines the potential factors that may lead to accidents. This study inherits the existing ontology subclassing method [33,35, 41] and divides it into three subclasses: unsafe worker behaviors, unsafe conditions of objects, and unsafe environmental factors. (1,3) subset.

[0040] The potential risk category defines risk events that may be triggered by risk factors (hazard sources) but have not yet occurred. When potential risks occur, they are closely related to the "where, what, what, and how" of the construction process. This study fully considers the characteristics of highway engineering, including roads, bridges, and tunnels, and consults with experts in the field. Based on the "Safety Standard for Prevention and Control of Common Accidents in Construction" (JGJ / T 429), it identifies 14 common risks during highway construction. Potential risk subclass, obtained (1,14) Subclass concept set, including Collapse and overturning risks Crane injury risk Risk of being struck by an object Risk of falling from height mechanical injury risk Risk of electric shock Vehicle injury risk Risk of poisoning and suffocation Fire risk Explosion risk Risks of blasting Drowning risk Burn risk Risk of collapsing rooftops. Each Construction activities may involve one or more Potential risks, such as The construction activities of the foundation cap have the following problems: Collapse and overturning risks Risk of being struck by an object Potential risks such as the risk of falling from heights.

[0041] The preventative measures category defines the necessary measures to ensure construction safety and control the hazards of safety accidents. Referring to the requirements of construction site safety and civilized construction specifications and consulting with experts in the field, these measures are divided into subcategories such as safety briefings, personal protective equipment, workspace protection, warning signs, protective sheds, patrols and inspections, and other protective measures. (1,n).

[0042] In this embodiment, the emergency response dimension includes two top-level concepts: accident type and response measures. .

[0043] The Accident Type class defines injury accidents that have occurred during construction and caused personal injury or economic loss. Its subclass names are similar to... The corresponding subclasses of the potential risk class include... Collapse and overturning accidents Crane injury accidents Object strike accident 14 types of accidents, including falls from heights, were identified. (1,14) Subclass concept set. Same type Accident types may occur in different During construction activities, such as Falls from heights can occur during various work activities, such as foundation pit construction, scaffolding construction, formwork construction, and hoisting construction.

[0044] The "Response Measures" category defines the emergency measures taken to mitigate harm and conduct rescue operations after an accident occurs. These measures are further subdivided based on the type of personal injury. (1,n) subclass concept set, including Call emergency services Report accident information CPR, Hemostasis and bandaging Drowning first aid, etc.

[0045] Optionally, in step S3, the data attributes include: name attributes, coding attributes, description attributes, risk level, and legal basis; the object attributes include: location attribute words, behavioral attribute words, and result-oriented words. Relationships between object attributes of classes The semantic relations in the equation are abstracted into triples, which satisfy the following relation: (10) In the formula, Let r be the starting class of the relation, and r be the predicate attribute. For the target class, A collection of domain concept classes; And based on semantic function, classify the object attribute relationships between classes. The object attribute relationships are classified into three categories: spatial relationships, normative constraint relationships, and causal relationships. Based on the above classification of object attribute relationships, the triplet forms satisfy the following relationships: Formal expression of triples in spatial relations: (11) Formal representation of triples under normalized constraints: (12) Formal representation of triples under causal relationships: ； (13) Based on the above object attribute relationships, a semantic relationship network of the ontology is formed.

[0046] In the above embodiments, location attribute terms include "located," behavioral attribute terms include "follow," "avoid," "control," and "prevent," and result-oriented terms include "generate" and "trigger." The class attributes mainly include two types: data properties and object properties. Data properties are used to define the relationship between a class and specific data values. For example, the name attribute (hasName) is used to identify the name of a class instance, such as the construction location name or activity name; the code attribute (hasCode) is a unique identifier for a class instance, mainly based on... The encoding rules are clearly defined in the formal formulas of the ontology concept set, such as... For roadbed engineering, the unique identifier for each instance can be set to M.1.1. Furthermore, based on safety training management needs, data attributes such as description (hasDescription), risk level (hasRiskLevel), and regulatory basis (hasRegulationBasis) can be defined to enrich the ontology's knowledge representation. Object attributes are used to define relationships between classes. Top-level conceptual classes have diverse semantic relationships that are precisely defined through different object attributes.

[0047] Object attributes construct a semantic network between ontology classes, enabling organic connections between knowledge. It should be noted that there is no perfect definition for ontology classes and attributes. This study, based on the analysis of security knowledge text content and combined with application requirements, extends the definition of object attributes for top-level concept classes, as shown in Table 2.

[0048]

[0049] Based on the above object attribute definitions, this embodiment further constructs a semantic relationship network structure diagram, demonstrating the relationship connections and distribution characteristics between concept classes. See Appendix for details. Figure 3 .

[0050] In the domain ontology quintuple structure Representing the semantic relationships of object attributes between classes, these relationships are important structural connection units in ontology. To further clarify the structural forms and types of these relationships, this paper will... The semantic relations in the equation are abstracted into triples, as detailed in equation (10).

[0051] At the structural level, It is a connection Semantic bridges between various internal categories. To enhance semantic clarity, this embodiment classifies them according to semantic function. The object attribute relationships in the equation are classified into three categories: spatial relationships, normative constraint relationships, and causal relationships. Spatial relationships are detailed in equation (11), normative constraint relationships in equation (12), and causal relationships in equation (13). Among them, spatial relationships (R... spatial ): such as locate, used to indicate the spatial location or structural affiliation of objects such as construction activities and risk sources; normative constraint relationships (R regulatory ): such as require, avoid, express the construction activities' adherence to work specifications and risk avoidance; causal relationship (R causal): such as trigger, prevent, generate, and respond, are used to characterize the causal, prevention, and response relationships between risks, accidents, and measures; these object attribute relationships are formally expressed in ontology construction through domain and range, and can be configured and annotated in Protégé 5.6.4.

[0052] Optionally, in step S4, the semantic rules are constructed using SWRL rules, wherein the general form of SWRL rules satisfies the following relation: (14) In the formula, The object attribute relationships in the premise of the rule are derived from the semantic relationship set. , This indicates the conclusion of the rule.

[0053] In the above embodiments, the knowledge in the field of highway construction safety training has more complex logical relationships, such as generating new semantic conclusions when multiple conditions are met simultaneously. To model these complex semantic relationships, this invention employs semantic rules. See equation (14) for details. Semantic rule construction is a five-tuple structure in the ontology model. The core formalization methods of components, and Semantic relation definition The instance triples together form a complete reasoning and expression loop, and in the example, the SWRL rule is as follows: "Construction activity (?a)^fall from height risk (?b)^trigger (?a,?b)^preventNot (?m,?b)→high-risk operation (?a)" The above rules derive implicit new semantic conclusions by logically combining multiple defined semantic relationships. The implementation example can design multiple SWRL-based semantic rules based on high-frequency semantic relationships identified in actual safety training corpora. These rules provide logical modeling and automatic reasoning support for typical scenarios such as operational behavior, equipment usage, risk assessment, and emergency response, effectively improving the intelligent representation capabilities of the domain knowledge graph. Some rule examples are shown in Table 3.

[0054]

[0055] Optionally, in step S5, generating instance triples includes: Specific instances of classes in the ontology An instance triple is represented as a triple structure consisting of a subject entity, a semantic relation predicate, and an object entity. The formal definition of an instance triple satisfies the following relation: (15) In the formula, The semantic entities in the instance layer are derived from the ontology concept class. The instantiation result, Represents a semantic object attribute, used to define the semantic relationship between two entities.

[0056] In this embodiment, the example is expressed. As a component of the quintuple structure, it is used to define the basic unit of expression for domain knowledge at the instance layer. It is represented as a triplet structure consisting of a subject entity, a semantic relation predicate, and an object entity. For a formal definition, see Equation (15).

[0057] This application also provides a semantic modeling system for highway construction safety training, used to implement all examples of the semantic modeling method for highway construction safety training. The system includes: The knowledge acquisition and preprocessing module is used to acquire and process construction safety templates from template sources, and to perform optical character recognition, regular expression cleaning and clause segmentation on the templates to form structured knowledge clauses. The semantic ontology construction module is used to abstract multiple core concept classes based on the three-dimensional logical framework of operation control-risk prevention-emergency response, and to construct an ontology framework containing the core concept classes and their hierarchical relationships. A semantic rule base is used to connect object attributes of different concept classes and data attributes used to describe the characteristics of concept classes to form a semantic relation network of the ontology, and to construct semantic rules for logical reasoning based on the semantic rule language in the semantic relation network. The reasoning engine is used to load and execute rules in the semantic rule base to perform logical reasoning. The instantiation module is used to batch map structured knowledge terms to instances and generate instance triples to populate the ontology framework. The training recommendation engine is used to dynamically identify risks in input construction tasks based on the constructed ontology, semantic rules, and instance triples, and output corresponding safety training recommendations.

[0058] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A semantic modeling method for highway construction safety training, characterized in that, include: S1. Obtain templates related to construction safety, and perform optical character recognition, regular expression cleaning, and clause segmentation on the templates to form structured knowledge clauses; S2. Based on the three-dimensional logical framework of operation control, risk prevention and emergency response, multiple core concept classes are abstracted from structured knowledge clauses, and an ontology framework containing the core concept classes and their hierarchical relationships is constructed. S3. Define object attributes used to connect different concept classes and data attributes used to describe the characteristics of concept classes to form a semantic relationship network of the ontology; S4. Construct semantic rules for logical reasoning based on semantic rule language in semantic relation network; S5. Using semantic rules, the structured knowledge terms are mapped to instances in batches, and instance triples are generated and populated into the ontology framework. S6. Based on the filled ontology framework, semantic rules and instance triples, perform dynamic risk identification on the input construction task and output corresponding safety training recommendations. In step S2, the core concept categories include: construction location, construction activities, work requirements, hazard sources, potential risks, preventive measures, accident types, and handling measures. The ontology framework, which includes the core concept classes and their hierarchical relationships, satisfies the following relation: ； In the formula, This refers to the core of safety training for construction workers in highway construction. This represents ontology concept classes that have the same characteristics. This represents the relationship between object attributes of different classes. Data attributes that represent the relationship between classes and data values. Represent semantic constraint rules in domain knowledge. Represents a specific instance of a class in the ontology; Ontology concept classes with the same characteristics in the ontology framework Defined as three concept sets: operation control, risk prevention, and emergency response, the ontology concept class is expressed according to the following relation: ； In the formula, For work control, For risk prevention, For emergency response; Concept set A set of top-level concept classes used to construct the operation control dimension; concept set The top-level concept set used to construct the risk prevention dimension; concept set A set of top-level concept classes used to construct the emergency response dimension; concept set , , They respectively satisfy the following relations: ； ； ； In the formula, Construction location, For construction activities, As required by the assignment, As a source of danger, As a potential risk, As a precaution, Accident type As a measure to address the issue; Concept set , , Each class contains j second-level subclasses, and the second-level subclasses satisfy the following relationship: ； ； ； In the formula, This is a second-level concept class in the task control dimension. 'j' represents the top-level class number, and 'j' represents the second-level class number.

2. The semantic modeling method for highway construction safety training according to claim 1, characterized in that, In step S1, optical character recognition includes: performing optical character recognition on scanned documents in templates related to construction safety, converting the scanned documents into processable text; regular expression cleaning includes: removing symbols unrelated to the text content from the processable text; and clause segmentation includes: constructing clause feature words, and performing semantic segmentation on the regular expression-cleaned processable text based on the clause feature words.

3. The semantic modeling method for highway construction safety training according to claim 1, characterized in that, In step S3, the data attributes include: name attribute, coding attribute, description attribute, risk level and legal basis, and the object attributes are location attribute words, behavior attribute words and result-oriented words; Relationships between object attributes of classes The semantic relations in the equation are abstracted into triples, which satisfy the following relation: ； In the formula, Let r be the starting class of the relation, and r be the predicate attribute. For the target class, A collection of domain concept classes. It is in triplet form; And based on semantic function, classify the object attribute relationships between classes. The object attribute relationships are classified into three categories: spatial relationships, normative constraint relationships, and causal relationships. Based on the above classification of object attribute relationships, the triplet forms satisfy the following relationships: Formal expression of triples in spatial relations: ； In the formula, It is a triplet form under spatial relations. It is a spatial relation predicate. Words with location attributes; Formal representation of triples under normalized constraints: ； In the formula, To standardize the form of triples under constraints, To standardize the predicates of constraint relations, All are behavioral attribute terms; Formal representation of triples under causal relationships: ； In the formula, In the form of a triple under causal relationship, It is a predicate of cause and effect. All are result-oriented terms; Based on the above object attribute relationships, a semantic relationship network of the ontology is formed.

4. The semantic modeling method for highway construction safety training according to claim 1, characterized in that, In step S4, the semantic rules are constructed using the Semantic Web Rule Language, wherein the general form of the Semantic Web Rule Language rules satisfies the following relation: ； In the formula, This indicates the relationship between object attributes in the premise of the rule. Indicates the conclusion of the rule. For Semantic Web rule language rules, It represents the relationship between object attributes between classes.

5. The semantic modeling method for highway construction safety training according to claim 1, characterized in that, In step S5, generating instance triples includes: Specific instances of classes in the ontology An instance triple is represented as a triple structure consisting of a subject entity, a semantic relation predicate, and an object entity. The formal definition of an instance triple satisfies the following relation: ； In the formula, The semantic entities in the instance layer are derived from the ontology concept class. The instantiation result, Represents a semantic object attribute, used to define the semantic relationship between two entities.

6. A semantic modeling system for highway construction safety training, used to implement the semantic modeling method for highway construction safety training as described in any one of claims 1-5, characterized in that, The system includes: The knowledge acquisition and preprocessing module is used to acquire and process construction safety templates from template sources, and to perform optical character recognition, regular expression cleaning and clause segmentation on the templates to form structured knowledge clauses. The semantic ontology construction module is used to abstract multiple core concept classes based on the three-dimensional logical framework of operation control-risk prevention-emergency response, and to construct an ontology framework containing the core concept classes and their hierarchical relationships. A semantic rule base is used to connect object attributes of different concept classes and data attributes used to describe the characteristics of concept classes to form a semantic relation network of the ontology, and to construct semantic rules for logical reasoning based on the semantic rule language in the semantic relation network. The reasoning engine is used to load and execute rules in the semantic rule base to perform logical reasoning. The instantiation module is used to batch map structured knowledge terms to instances and generate instance triples to populate the ontology framework. The training recommendation engine is used to dynamically identify risks in the input construction tasks based on the constructed ontology, semantic rules, and instance triples, and output corresponding safety training recommendations. Core concepts include: construction location, construction activities, work requirements, hazards, potential risks, preventive measures, accident types, and response measures; The ontology framework, which includes the core concept classes and their hierarchical relationships, satisfies the following relation: ； In the formula, This refers to the core of safety training for construction workers in highway construction. This represents ontology concept classes that have the same characteristics. This represents the relationship between object attributes of different classes. Data attributes that represent the relationship between classes and data values. Represent semantic constraint rules in domain knowledge. Represents a specific instance of a class in the ontology; Ontology concept classes with the same characteristics in the ontology framework Defined as three concept sets: operation control, risk prevention, and emergency response, the ontology concept class is expressed according to the following relation: ； In the formula, For work control, For risk prevention, For emergency response; Concept set A set of top-level concept classes used to construct the operation control dimension; concept set The top-level concept set used to construct the risk prevention dimension; concept set A set of top-level concept classes used to construct the emergency response dimension; concept set , , They respectively satisfy the following relations: ； ； ； In the formula, Construction location, For construction activities, As required by the assignment, As a source of danger, As a potential risk, As a precaution, Accident type As a measure to address the issue; Concept set , , Each class contains j second-level subclasses, and the second-level subclasses satisfy the following relationship: ； ； ； In the formula, This is a second-level concept class in the task control dimension. 'j' represents the top-level class number, and 'j' represents the second-level class number.

Images