The Analysis of Controlled Flight Into Terrain Incidents From Flight Crew Perspective Using Named Entity Recognition and Bayesian Networks
Bibliographic record
Abstract
Controlled flight into terrain (CFIT) can result in significant aircraft damage and human casualties. Analyzing incident factors and their evolutionary relationships in aviation safety reports helps explore the inherent mechanisms of CFIT, thereby potentially reducing their occurrence. This study proposes a methodology combining named entity recognition (NER) and Bayesian network (BN) to address the challenges of efficiently extracting incident factors from textual reports from the crew’s perspective and analyzing the overall evolution process of CFIT incidents to better prevent accidents. First, this study collected 354 CFIT incident reports in the Aviation Safety Reporting System (ASRS) for the period November 2021 to August 2023. Second, important concepts from Threat and Error Management (TEM) were referenced to determine principles for extracting factor types and their evolutionary relationships. Third, NER was applied using the BERT–BiLSTM–MHA–CRF model to extract incident factors, followed by model comparison. Experimental results demonstrated good performance with precision, recall, and F 1 score of 0.97, 0.90, and 0.90, respectively. Last, BN was then employed to analyze the CFIT evolution process. Results indicate that if factors such as Terrain (0.04) and Unfamiliarity/Inexperience (0.036) are present, CFIT risk will increase. Conversely, if protective factors such as Perfect Weather/Great Visibility (0.397) and Perform the Escape Maneuver (0.341) are present, CFIT risk will decrease. The analysis reveals that Airline Operational Pressure, Fatigue (57%), Lack of Situational Awareness (21%), Automation Errors (45%), Aircraft Handling Deviations (34%), Aviation System–Based Countermeasures (72%), Perform the Escape Maneuver (75%), and Make a Stabilized Approach (89%) form the highest probability evolution pathway for CFIT incidents. This study concludes that reducing these identified risk factors and increasing protective factors can contribute to reducing CFIT accidents.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.009 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.006 | 0.004 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".