The authors develop an NLP pipeline that extracts structured entities (substances, equipment, events, causes, consequences) from chemical accident databases (e.g., eMARS) and automatically populates a predefined process‑safety ontology. The semantically enriched accident knowledge base supports queries beyond keyword search and helps reveal cause–effect relationships and similarities between cases. For PSM incident investigation, the approach enables faster retrieval of precedent events, more consistent classification of findings, and better reuse of lessons learned across investigations and sites.
