Text mining of accident reports using semi-supervised keyword extraction and topic modeling

Proposes a semi-supervised text-mining approach (keyword extraction and topic modeling) to organize and classify large volumes of accident reports with limited labeled data. For PSM and MOC, this helps teams quickly identify recurring contributors tied to changes—such as procedure gaps, training deficiencies, or inadequate change controls—and to retrieve similar cases when evaluating proposed modifications. Embedding such analytics into MOC workflows can improve the quality of pre-change hazard reviews and the effectiveness of post-change learning and corrective actions.