Cross-sector accident-labeling transfers with results depending on report style
Moving a narrative tagging system between industries can work, but results depend on how reports are written.
- Publication
- arXiv
- Stage
- Preprint
- What we read
- Summary of the paper
- Authors
- Aho Yapi, Pierre Latouche, Arnaud Guillin, Yan Bailly
- Universities and research institutions
- Not yet supplied in verified metadata; the Brief does not guess.
What they did and found
Researchers trained models to label small facts in accident stories from construction, metalwork, chemistry-plastics, and a private firm. When tested on other sectors, task-tuned models did better across fields, but results varied with incident descriptions.
Why it matters
If you reuse a model at different sites, review how reports are written; otherwise results may drop and need expert review.
Breaking accident stories into labeled facts helps compare cases across groups. The biggest gains came when the model was tuned for the task, not kept as-is.
In practice, results shift when a company writes incidents differently. Leave expert review for unclear cases and unusual report styles.
Original sources · 1
- Cross-sector generalization of accident-process role classification in occupational accident narratives ↗arXiv · 2026-09-18
Check the original paper for its authors, methods, version and access terms.
What remains uncertain
Findings depend on the exact setting and report style; transferability is not guaranteed across all sites or languages either.
What is your take?
Ask a question, add useful context or share a different perspective. Keep the conversation respectful and grounded.