Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Hierarchical Multi-label Classification for Fine-level Event Extraction from Aviation Accident Reports

Xinyu Zhao, Hao Yan, Yongming Liu

TL;DR

This article argues that it can identify the events more accurately by leveraging the event taxonomy by incorporating a novel hierarchical attention module into the bidirectional encoder representations from transformers model and regularizing the proposed model according to the relationship and distribution among labels.

Abstract

A large volume of accident reports is recorded in the aviation domain, which greatly values improving aviation safety. To better use those reports, we need to understand the most important events or impact factors according to the accident reports. However, the increasing number of accident reports requires large efforts from domain experts to label those reports. In order to make the labeling process more efficient, many researchers have started developing algorithms to identify the underlying events from accident reports automatically. This article argues that we can identify the events more accurately by leveraging the event taxonomy. More specifically, we consider the problem a hierarchical classification task where we first identify the coarse-level information and then predict the fine-level information. We achieve this hierarchical classification process by incorporating a novel hierarchical attention module into BERT. To further utilize the information from event taxonomy, we regularize the proposed model according to the relationship and distribution among labels. The effectiveness of our framework is evaluated with the data collected by National Transportation Safety Board (NTSB). It has been shown that fine-level prediction accuracy is highly improved, and the regularization term can be beneficial to the rare event identification problem.

Hierarchical Multi-label Classification for Fine-level Event Extraction from Aviation Accident Reports

TL;DR

This article argues that it can identify the events more accurately by leveraging the event taxonomy by incorporating a novel hierarchical attention module into the bidirectional encoder representations from transformers model and regularizing the proposed model according to the relationship and distribution among labels.

Abstract

A large volume of accident reports is recorded in the aviation domain, which greatly values improving aviation safety. To better use those reports, we need to understand the most important events or impact factors according to the accident reports. However, the increasing number of accident reports requires large efforts from domain experts to label those reports. In order to make the labeling process more efficient, many researchers have started developing algorithms to identify the underlying events from accident reports automatically. This article argues that we can identify the events more accurately by leveraging the event taxonomy. More specifically, we consider the problem a hierarchical classification task where we first identify the coarse-level information and then predict the fine-level information. We achieve this hierarchical classification process by incorporating a novel hierarchical attention module into BERT. To further utilize the information from event taxonomy, we regularize the proposed model according to the relationship and distribution among labels. The effectiveness of our framework is evaluated with the data collected by National Transportation Safety Board (NTSB). It has been shown that fine-level prediction accuracy is highly improved, and the regularization term can be beneficial to the rare event identification problem.
Paper Structure (26 sections, 28 equations, 8 figures, 5 tables)

This paper contains 26 sections, 28 equations, 8 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Information retrieval for narrative reports
  4. Hierarchical classification
  5. Methodology
  6. Review of related methodology
  7. Review of Hierarchical Attention
  8. Review of Recursive Regularization
  9. Review of Label Distribution Learning
  10. Proposed Model for Hierarchical Multi-label classification
  11. Problem definition
  12. Overview of the model architecture
  13. BERT-based feature extractor
  14. Coarse-level model
  15. Hierarchical attention
  16. ...and 11 more sections

Figures (8)

  • Figure 1: Label statistics for 100 most frequent fine-level labels in NTSB
  • Figure 2: An example of label taxonomy defined by NTSB
  • Figure 3: Aviation accident report to event labels: Report 20001208X07734 from NTSB. The left side is the raw accident report. On the right side is the event sequence labeled by NTSB. We highlight the keywords in the narrative reports and plot their relationship with the corresponding event labels
  • Figure 4: Overview of the two-stage hierarchical attention.
  • Figure 5: Coarse level model architecture
  • ...and 3 more figures