BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Phoenix
X-LIC-LOCATION:America/Phoenix
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19700101T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20241014T203101Z
LOCATION:FLW Salon B
DTSTART;TZID=America/Phoenix:20240911T150000
DTEND;TZID=America/Phoenix:20240911T151000
UID:HFESAM_ASPIRE - Presented by HFES_sess116_LBR139@linklings.com
SUMMARY:A Multidimensional Accident Analysis Framework Based on Hazard Tri
 angle and Large Language Models
DESCRIPTION:Late Breaking Results\n\nSouvik Das (Purdue University); Sinja
 na Choudhuri (Indian Institute of Technology Kharagpur, Purdue University)
 ; and Gaurav Nanda (Purdue University)\n\nFor effective injury prevention,
  accurate and detailed understanding of accident causation factors is esse
 ntial. Raw injury data is typically collected in the form of incident narr
 atives at hospitals and accident reports. Due to the unstructured and nois
 y nature of the narratives with misspellings and improper grammar, it is c
 hallenging to automatically extract specific information about the acciden
 t causation factors. Manual analysis of narratives is time and resource in
 tensive. Recent advances in large language models (LLMs) have shown promis
 ing performances for various natural language processing tasks involving a
 nalysis of noisy textual data. We propose a novel framework to extract the
  key accident causation factors based on Hazard Triangle from the textual 
 incident narratives using LLMs. To evaluate the performance of this framew
 ork, we examined its accuracy on a sample of 50 construction related injur
 y reports from January 2015 to February 2021 obtained from the occupationa
 l safety and health administration (OSHA’s) website. We used ChatGPT-3.5 L
 LM and developed prompts based on the concepts of persona-assignment and c
 hain-of-thought to extract the following three Hazard Triangle elements fr
 om the narratives: hazardous elements, initiating mechanisms, and threats.
  The hazard elements for these cases were also manually identified by rese
 archers and compared against the output from LLM framework. The analysis i
 ndicated that the LLM framework observed the following Recall values: 70% 
 for hazardous elements, 82% for initiating mechanisms, and 54% for targets
 /threats. This indicated that the framework can efficiently and reliably e
 xtract structured accident causation information from injury narratives th
 us enabling quicker and effective prevention measures.\n\nSession Chair: M
 ustafa Demir (Texas A&M University)
END:VEVENT
END:VCALENDAR