Leveraging Natural Language Processing to Identify Health and Safety Challenges during Post-Disaster Reconstruction

Post-disaster reconstruction efforts are becoming increasingly more important with the increase in natural disasters in recent years. These efforts support rapid recovery and are fundamental to the resilience of the impacted communities. However, such efforts that involve complex activities in difficult and uncertain environments impose considerable health and safety challenges. The health and safety challenges are particularly heightened in the context of developing nations that lack necessary resources as compared to their more developed counterparts. An understanding of the health and safety challenges in the context of these developing nations is fundamental to the development of health and safety best practices that support effective and rapid recovery. This study examines the health and safety challenges as reported by stakeholders impacted in the 2015 Gorkha earthquake in Nepal. The data were captured using a set of semi-structured interviews with several stakeholders that include general contractors, owners, government officials, engineers, supervisors, and workers. After transcribing the interviews as textual data, a popular and frequently used topic modeling algorithm, Latent Dirichlet Allocation (LDA), was adopted to unveil health and safety related themes as reported by the stakeholders. The findings of the study identify the major health and safety related challenges in the context of a post-disaster reconstruction effort and will serve as the foundation for future research efforts that focus on health and safety management solutions. The findings of the study can also be used by the governments and stakeholders with supervisory role to create a safer work-environment for the reconstruction workers.