Assessing the Use of UAS-Related Terms in ASRS Using Seed Topic ModelingContext: The NASA Aviation Safety Reporting System (ASRS) is a voluntary confidential system that disseminates reports received from personnel involved in aviation operations after de-identifying them. These reports are used by the community to improve overall aviation system safety.
Aim: We propose and execute an experiment to assess the use of seed term topic modeling over the database narratives to identify Unmanned Aircraft System (UAS) reports. The use of seed term topic modeling enables users to identify groups of conceptually similar narratives associated to a topic of their interest.
Method: We use a collection of narratives, expert-selected words, and report metadata that separates UAS from non-UAS reports to assess if seed topic modeling can be used to improve ASRS searches.
Results: For simpler queries, seed topic search observes a higher recall and lower precision than the existing DBOL (DataBase OnLine) search in operation. However, the best results are obtained when seed topic search is used as a search suggestion system to be executed on the DBOL.
Conclusion: Utilizing a combination of both the existing method and the proposed method, users can expand their search vocabulary about subjects of interest while improving the quality of results.
Carlos Paradis (KBR (United States) Houston, Texas, United States)
Misty D. Davies (Ames Research Center Mountain View, California, United States)
Becky L. Hooey (Ames Research Center Mountain View, California, United States)
November 23, 2022
Systems Analysis And Operations Research
Meeting: AIAA Science and Technology Forum and Exposition (2023 AIAA SciTech Forum)
Location: National Harbor, MD
Start Date: January 23, 2023
End Date: January 27, 2023
Sponsors: American Institute of Aeronautics and Astronautics