SIG DSA - Data Science and Analytics for Decision Support
Loading...
Paper Type
ERF
Paper Number
1739
Description
Occupational health and safety are significant issues for many organizations. Workplace accidents have significant consequences for employees, and they also cause significant disruptions to business processes. In this paper, we analyze accident narrative data collected by the Occupational Safety and Health Administration (OSHA) to classify events leading to injuries and body parts injured. These datasets are collected in textual form, which can create difficulties for nuanced statistical analyses. By creating high-performing text classifiers, we can automatically code these records and provide decision-makers with useful information. We use BERT word embeddings and several variants thereof to perform these classifications, finding accuracy of approximately 89 percent and 79 percent for event types and body parts respectively. These contributions will improve organizations’ capacities to understand and act upon textual accident narratives.
Recommended Citation
Goldberg, David M. and Zaman, Nohel, "Text mining for classifying workplace severe injury events" (2022). AMCIS 2022 Proceedings. 14.
https://aisel.aisnet.org/amcis2022/sig_dsa/sig_dsa/14
Text mining for classifying workplace severe injury events
Occupational health and safety are significant issues for many organizations. Workplace accidents have significant consequences for employees, and they also cause significant disruptions to business processes. In this paper, we analyze accident narrative data collected by the Occupational Safety and Health Administration (OSHA) to classify events leading to injuries and body parts injured. These datasets are collected in textual form, which can create difficulties for nuanced statistical analyses. By creating high-performing text classifiers, we can automatically code these records and provide decision-makers with useful information. We use BERT word embeddings and several variants thereof to perform these classifications, finding accuracy of approximately 89 percent and 79 percent for event types and body parts respectively. These contributions will improve organizations’ capacities to understand and act upon textual accident narratives.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
SIG DSA