Automatic classification of accident roles works across job sectors
Cross-sector generalization of accident-process role classification in occupational accident narratives
Computation and Language
Summary
Work accident reports contain useful details but are written differently depending on the industry, which makes automatic understanding hard. The authors tested a computer method trained on construction accident reports to see if it could understand reports from metallurgy and chemical sectors without extra training. They found that adapting their models specifically to the task helped the computer classify parts of reports accurately across these sectors. This shows it’s possible to build tools that help experts analyze safety incidents from various industries without needing to retrain for each sector.
What this means in practice
- •For workplace safety teams: Automatically structure accident reports from different sectors to speed up expert risk analysis and improve workplace prevention strategies.
- •For industrial compliance officers: Use generalized role classifiers to review accident narratives from multiple industries without retraining models for each company or sector.
Authors
Aho Yapi, Pierre Latouche, Arnaud Guillin, Yan Bailly
Abstract
Occupational accident narratives contain valuable information about work situations, unfavourable conditions, accident events, and their consequences. Automatically structuring these narratives can facilitate large-scale accident analysis and support occupational risk prevention. However, the terminology and writing styles used to describe accidents vary considerably across sectors and organisations, raising questions about the ability of automated coding systems to generalize beyond their training domain. In this paper, we evaluate the cross-sector generalization of accident-process role classification in French occupational accident narratives. We construct an expert-annotated corpus in which factual units are classified into four roles: work situation (A0), explicitly reported unfavourable condition (A1), accident event or deviation (B), and reported consequence (C). The role classifiers are developed and selected exclusively on 42,244 factual units extracted from 6,040 construction-sector narratives and are then evaluated on unseen corpora from the metallurgy and chemistry--plastics sectors, as well as on an independently collected company corpus, without retraining or target-domain tuning of the role classifier. We compare frozen pretrained representations with task-specific fine-tuning and supervised representation-learning strategies. The results show that task-specific adaptation consistently improves cross-domain transfer over frozen representations. Across repeated training runs, the three leading task-adapted strategies achieved average balanced accuracies between 85.6% and 85.8% across the three target corpora. These findings support the development of transferable assisted-coding systems capable of consistently structuring heterogeneous occupational accident narratives for expert review and cross-sector prevention analysis.