Summary of Hierarchical Multi-label Classification For Fine-level Event Extraction From Aviation Accident Reports, by Xinyu Zhao et al.
Hierarchical Multi-label Classification for Fine-level Event Extraction from Aviation Accident Reports
by Xinyu Zhao, Hao Yan, Yongming Liu
First submitted to arxiv on: 26 Mar 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel approach to automatically identifying underlying events from accident reports in the aviation domain is proposed. Leveraging an event taxonomy, researchers have developed algorithms to classify these events more accurately. The paper presents a hierarchical classification task, first predicting coarse-level information and then fine-level details using a novel hierarchical attention module incorporated into BERT. To further utilize the event taxonomy, a regularization term is introduced to account for relationships and distributions among labels. Evaluation on data collected by the National Transportation Safety Board (NTSB) shows significant improvements in fine-level prediction accuracy and rare event identification. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Accident reports in aviation are crucial for improving safety. The problem is that labeling these reports requires a lot of work from experts. To make this process more efficient, researchers have developed algorithms to automatically identify the most important events or factors. This paper suggests using an “event taxonomy” to help with this task. It’s like a big dictionary that explains what different events are. The algorithm works by first identifying general information and then getting more detailed information. A special attention module is used to make sure it gets the right information. To make the system even better, a special term is added to account for how labels relate to each other. This was tested on data from the National Transportation Safety Board (NTSB) and showed big improvements. |
Keywords
» Artificial intelligence » Attention » Bert » Classification » Regularization
