Using Text Mining to Analyze Key Topics in Civil Petitions Related to Hazardous Chemicals

Abstract

Various accident prevention studies and related policies have been developed to ensure public safety when handling and transporting harmful chemical substances. However, these policies primarily focus on improving government-level laws and policies, often overlooking the immediate needs of consumers. Therefore, this study proposed the extraction of meaningful topics and keywords from questions and answers pertaining to the safe handling of harmful chemicals using civil petitions data related to these substances, as posted on the Anti-corruption and Civil Rights Commission platform. The idea is to prioritize these topics in policy formulation. In addition, topic modeling techniques, namely Latent Semantic Analysis and Latent Dirichlet Allocation algorithms, were employed, and the results and implications of each algorithm were compared and analyzed. The main topics identified through the Latent Dirichlet Allocation algorithm were “piping and valve management and inspection,” “manufacturing and storage facility safety,” “outdoor impact assessment,” “damage reduction and accident prevention facility,” and “education and related law.” The results derived from this study are expected to contribute to the development of accident prevention measures by directly addressing the requirements of consumers when it comes to harmful chemical substances.