Al Analyzer for Revealing Insights of Traffic Crashes
The project will develop an artificial intelligence (AI)-based software program to analyze traffic crash narratives in free text form and identify factors that increase the severity of the crashes. Stage 1 work will focus on developing the proposed software solution. A literature review of the state-of-the-art techniques in NLP for sentence classification and explainable AI will be performed to ensure application of the most up-to-date and sophisticated NLP techniques for best results and insights extracted from crash narratives. The techniques identified from literature review will then be implemented. Classification algorithms will be trained using data provided by partner agencies, and explainable AI techniques will be developed and applied to identify contributing factors and evaluate the consistency of obtained insights across different data sources. These efforts will result in a raw AI analyzer and also in the development of a web-based software for a broader application by the industry and other users. The web-based software will use, as default settings, techniques and parameters that are found to offer the best overall performance. Additional configuration parameters for the AI techniques will be provided so that the analysts can test custom analysis approaches. Work in Stage 2 will focus on system validation and refinement for a broad adoption by the industry. Results from the proposed system will be compared with those from the classic analysis of quantitative data using statistical techniques, such as data count and discrete outcome models. The comparison is expected to generate insights not only in terms of similarity but also about their statistical significance and correct association with different crash severities. The results will be evaluated for quality and suitability, and collaborative sessions will be held with traffic safety engineers from partner agencies to corroborate their validity. A website will be created to share information about the system, library, and licenses. A webinar directed at transportation agencies will also be conducted. The final report will provide the research results and details about the developed system, such as the architecture and functionalities, along with recommendations. Limitations and expansion opportunities will also be discussed
Language
- English
Project
- Status: Active
- Funding: $82899
-
Contract Numbers:
20-30/IDEA 231
-
Sponsor Organizations:
Safety Innovations Deserving Exploratory Analysis (IDEA)
Transportation Research Board
500 Fifth Street, NW
Washington, DC United States 20001National Cooperative Highway Research Program
Transportation Research Board
500 Fifth Street, NW
Washington, DC United States 20001American Association of State Highway and Transportation Officials (AASHTO)
444 North Capitol Street, NW
Washington, DC United States 20001Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC United States 20590 -
Project Managers:
Jawed, Inam
-
Performing Organizations:
University of Nevada at Las Vegas
, -
Principal Investigators:
Park, Jee
- Start Date: 20210714
- Expected Completion Date: 0
- Actual Completion Date: 0
Subject/Index Terms
- TRT Terms: Artificial intelligence; Crash analysis; Crash causes; Crash severity; Literature reviews; Software; State of the art; Traffic crashes; Validation
- Subject Areas: Data and Information Technology; Highways; Safety and Human Factors;
Filing Info
- Accession Number: 01776582
- Record Type: Research project
- Source Agency: Transportation Research Board
- Contract Numbers: 20-30/IDEA 231
- Files: TRB, RIP
- Created Date: Jul 14 2021 5:38PM