Natural language processing and social network analysis based framework for falling accidents at construction sites

As urban development accelerates, the need for construction projects has significantly increased. Despite this growth, the construction industry remains the most dangerous sector, with falls from heights being the leading cause of fatal accidents, ranking first among the "five major injuries". Construction safety demands strong attention, as understanding and analyzing the causes of accidents is crucial for reducing their frequency and enhancing overall safety levels. However, traditional accident analysis methods often rely on expert experience and limited data, which restricts a comprehensive and systematic understanding of the complex factors contributing to these incidents. This study utilizes advanced techniques, including Natural Language Processing (NLP) and Social Network Analysis (SNA), to systematically investigate the causes and characteristics of falling accidents, analyze their interrelationships, and propose effective improvement measures. Through NLP topic analysis, five direct and eight indirect cause topics were identified, offering a comprehensive classification that enhances understanding of the specific contexts and backgrounds of accidents. The results of SNA provide frameworks for identifying key nodes and relationships among these causes by assessing degree centrality, closeness centrality, betweenness centrality, and eigenvector centrality. These frameworks offer essential insights for preventing falling accidents, assisting relevant departments and companies in reducing the likelihood of incidents, and enhancing safety management standards within the construction industry.