Large Language Models for topic analysis: Insights from robotic accident narratives

The integration of robotics in manufacturing has significantly enhanced productivity and safety by performing hazardous tasks. However, it also introduces new risks and accident profiles that require thorough analysis. This study explores the application of Large Language Models (LLMs) for topic modeling on accident narratives from the Occupational Safety and Health Administration (OSHA) database, focusing on incidents involving robots. Advanced Natural Language Processing (NLP) techniques, specifically LLMs, were employed to uncover prevalent themes, common risk factors, and patterns in these narratives. The methodology includes data collection, text processing, statistical analysis, and various topic modeling techniques. The analysis identifies six topics, and key themes including maintenance-related accidents, types of machinery involved, and specific body parts frequently affected. The findings highlight critical areas for improving safety measures, such as rigorous lockout/tagout procedures and enhanced training for maintenance personnel. This study provides a novel approach to analyze accident narratives.