Research On Establishment Of Knowledge Graph For Building Construction Safety Based On Natural Language Processing

The construction project uses large scale,complex technology and changeable environment,so it is easy to happen the construction safety accident.The construction safety accident is the result of multiple risk factors,so the construction safety management needs the comprehensive support of all aspects of safety knowledge.Although the construction industry has accumulated a wealth of construction safety data,these data cannot be fully utilized in traditional construction safety management due to the lack of means to transform these data into reusable knowledge.The knowledge graph describes the concepts,entities and their relationships in the objective world in a structured form,stores and uses knowledge in a way close to the human cognitive world,and provides a tool for organizing and managing massive amounts of knowledge.Therefore,this article uses textual materials such as construction safety accident reports and construction safety regulations to systematically carry out research on the establishment of knowledge graphs in the field of construction safety based on natural language processing.The rapid retrieval of construction safety knowledge and statistical analysis of construction safety accidents are used to validate the proposed framework.The main work of this thesis is as follows:(1)The research background of the thesis is introduced in detail from three aspects: 1)construction safety of construction projects,2)construction safety knowledge management and 3)policy for informatization development of construction industry.The research purpose and significance of this thesis is also explained.This thesis comprehensively expounds the research status of natural language processing and knowledge graph,and analyzes in detail the application status of natural language processing and knowledge graph in the field of construction engineering.(2)The named entity recognition of construction safety accident text is proposed.The modified Bidirectional Encoder Representations from Transformers(BERT)pre-trained language model is used to obtain dynamic word vectors,and the Bidirectional Long Short Term Memory-Conditional Random Field(Bi LSTM-CRF)model is used to obtain the optimal tag sequence of entities,with which a named entity recognition model suitable for the field of construction safety is proposed.1,000 construction safety accident reports were used as experimental data to carry out the named entity recognition experiment.The overall harmonic mean of the proposed model reaches 95.18%,which indicates that the proposed model has better entity recognition effect in construction safety accident texts.(3)The entity relationship identification framework of construction safety accident text is developed.The Roberta pre-trained language model is used to obtain dynamic word vector and entity position vector,and Convolutional Neural Network(CNN)model is used to classify the relationship between entities,with which an entity relationship recognition model suitable for the field of building construction safety is proposed.1,000 construction safety accident reports were used as experimental data to carry out the named entity recognition experiment.The overall harmonic mean of the proposed model reaches 96.27%,which indicates that the proposed model has better entity relationship recognition effect in construction safety accident texts.(4)The method of constructing the knowledge base of construction safety code is proposed Following the analyzing the text characteristics of the construction safety code,summarizing the expression characteristics of the general code clauses and the knowledge of the clauses,a method for constructing a normative knowledge base with mixed granularity based on attribute association is proposed.On the basis of "Technical Code for Safety in Elevated Operation of Building Construction",the knowledge base of building construction safety code is constructed by the ontology modeling software Protege,which provides rich structured data support for the formation of the knowledge graph of construction safety.(5)The application of knowledge graph in the field of construction safety is carried out.Based on the named entity,entity relationship and knowledge base data,the Neo4 j graph database is used to construct the knowledge graph of construction safety field,and the constructed knowledge graph is used to carry out the application research of construction safety accident portrait,accident information search,accident statistical analysis and normative knowledge search.The above applications illustrate that the proposed knowledge graph of construction safety domain can maximize the role of construction safety knowledge,thus providing necessary decision support for construction safety management.In this thesis,natural language processing and ontology technology are integrated to construct the knowledge graph of construction safety,which transforms the construction safety text data into reusable structured knowledge,provides decision support for construction safety management,and effectively improves the level of construction safety information management.