What is it about?
We want to design an algorithm to identify the fatal events from Aviation Accident Reports automatically. In other words, the algorithm can generate a brief summary of each accident report using keywords defined by domain experts. And the generated summary can be used for downstream tasks such as information retrieval, risk analysis, and predictive maintenance.
Featured Image
Photo by Sai Kiran Anagani on Unsplash
Why is it important?
The events are traditionally identified by people which requires them read through the accident reports using a lot of time. The algorithm will save a lot of human labor. On the other hand, the human labeling process can be subjective which leads to inconsistency. After generating accurate and consistent labels for the accident reports, we can achieve several downstream tasks such as information retrieval, risk analysis, and predictive maintenance.
Perspectives
There are many previous works focusing on this problem. In our work, we present a strong baseline with a deep learning approach. And we are the first to formulate the multi-label classification task as a sequence generation task. In this way, the temporal dependency of each event can be considered.
Xinyu Zhao
Arizona State University
Read the Original
This page is a summary of: Event Extraction for aviation accident reports through attention-based multi-label classification, June 2022, American Institute of Aeronautics and Astronautics (AIAA),
DOI: 10.2514/6.2022-3831.
You can read the full text:
Contributors
The following have contributed to this page
