Pith. sign in

REVIEW 4 major objections 5 minor 29 references

Utilizing AI for Aviation Post-Accident Analysis Classification

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read NLP and deep learning can classify aviation safety narratives with up to 90% accuracy.

desk verdict A self-review of a useful line of work, undermined by internal numeric inconsistencies and an unproven comparability assumption. read the letter →

arxiv 2506.00169 v1 pith:ZRLOAAI4 submitted 2025-05-30 cs.AI

Aziida Nanyonga , Graham Wild This is my paper
classification cs.AI
keywords aviationsafetynaturallanguageprocessingdeeplearningtopicmodelingaccidentreportclassificationNTSBATSBASN
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that AI techniques, specifically natural language processing (NLP), deep learning, and topic modeling, can automate the classification of aviation safety reports well enough to make post-accident analysis faster and more accurate. The authors review their own sequence of studies applying recurrent neural network models (sRNN, LSTM, BLSTM, GRU) and a simplified ResNet to classify aircraft damage level and phase of flight from narrative text in NTSB, ATSB, and ASN reports, along with topic models (LDA, NMF, LSA, pLSA) that draw out recurring themes. They report damage-level accuracies of about 89–90% for the NTSB and ATSB datasets, flight-phase accuracies of about 83–87% for those larger datasets but only 64% for the much smaller ASN dataset, and LDA as the strongest topic model with coherence scores around 0.58–0.60. The paper's working conclusion is that these methods are ready to support safety management by triaging reports and surfacing latent risk themes, while it flags dataset size and narrative length as the main variables still needing separate investigation.

What carries the argument

The load-bearing mechanism is the pairing of text preprocessing (tokenization, encoding, stop-word removal, lemmatization, TF-IDF and Word2Vec features) with two families of models: sequential deep networks (sRNN, LSTM, BLSTM, GRU, and combinations, plus a simplified ResNet) that read narratives word by word for classification, and topic models (LDA, NMF, LSA, pLSA with K-means in one study) that assign each report to latent themes evaluated by coherence scores. The same basic pipeline was applied to three publicly available safety datasets — NTSB, ATSB, and ASN — and the paper uses the shared pipeline to make cross-dataset performance comparisons.

What would settle it

Train the same classifiers on one large corpus while holding labels and preprocessing fixed, varying only the number of narratives through random subsets or only narrative length through truncation, and see whether accuracy tracks report count, word count, or both; if accuracy tracks word count rather than report count, the paper's conclusion that the number of narratives matters more than narrative length would be overturned. An equally decisive check is running identical models on identical label sets across all three data sources, since the ASN uses different damage categories (A1/A2) and includes only accidents.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that sequential deep-learning models trained on raw narrative text can classify two standard aviation-safety categories with near-practical accuracy: level of aircraft damage (sRNN reaching 90% accuracy on NTSB and 89% on ATSB reports) and phase of flight (LSTM reaching 87.4% on ATSB and sRNN 83.2% on NTSB, versus 64% for BLSTM on the smaller ASN set). The paper also claims that topic modeling, particularly LDA, reliably uncovers coherent latent themes such as mechanical failures and fuel-system issues across both NTSB and ATSB corpora, with coherence scores of 0.597 and 0.58 respectively. The central interpretive claim is that larger datasets improve accuracy, with report count mattering more than total narrative length, although the paper explicitly concedes that report count and narrative length were varied simultaneously in the underlying studies.

Load-bearing premise

The comparison rests on treating the NTSB, ATSB, and ASN studies as one controlled experiment with the same recipe, even though the number of reports, the length of the reports, and the damage categories all differed between sources.

Editorial extensions

If this is right

  • Post-accident investigators could use these trained classifiers to pre-label incoming reports, cutting the time spent on manual categorization.
  • Topic models can draw out recurring themes like mechanical failures and fuel problems directly from report text, giving safety managers a candidate list of risk areas to focus on.
  • Aviation authorities collecting safety narratives should prioritize gathering many short reports over a few long ones, since the paper finds report count is the stronger driver of accuracy.
  • Combined recurrent architectures, for example LSTM+BLSTM, can improve flight-phase classification on small datasets where single models lag.
  • Future work named in the paper, such as incorporating flight data recorder information, could push accuracy higher still.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial extension: the paper's cross-source accuracy gaps could be partly an artifact of label schemes, since the ASN reports use a two-level A1/A2 damage scale while NTSB/ATSB use four levels, so an accuracy gap of 20 points may reflect task difficulty rather than data scarcity.
  • Editorial extension: a practical deployment not tested in the paper would use these classifiers as triage tools that flag low-confidence narratives for human review, which would lower the cost of manual analysis even if raw accuracy is below 90%.
  • Testable extension: comparing models on reports fixed to the same length, such as 50, 100, or 200 words, with identical label sets would separate the effect of narrative length from the effect of dataset size, a separation the paper identifies as needed for future work.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This paper is a synthesis and comparative review of the authors' own prior work on applying NLP, deep learning, and topic modeling to aviation safety report narratives from the NTSB, ATSB, and ASN. The two classification tasks are aircraft damage level identification and phase-of-flight identification, with topic modeling added to uncover latent themes. The central claim, stated in the abstract and conclusion, is that NLP, deep learning, and topic modeling can significantly improve the efficiency and accuracy of aviation safety analysis. The paper presents comparative accuracy tables, discusses dataset size and narrative length as explanatory factors, and identifies future directions for separating these factors.

Significance. If the comparative claims were fully supported, the paper would provide a useful consolidation of a research program on NLP for aviation safety, with practical guidance on dataset characteristics and model choice. Its strengths include the use of real, publicly available safety report corpora and the explicit recognition that dataset size and narrative length were confounded in earlier comparisons. However, the paper's independent value is limited by two factors: the evidence base is almost entirely the authors' own prior publications (eight of the primary comparisons cite references [8, 15, 19-25]), and the cross-dataset conclusions depend on an assumption of methodological comparability that the manuscript itself does not establish and that is contradicted by internal numerical inconsistencies. No code, data, or machine-checked artifacts accompany the paper, and no new experiments are reported. The significance is therefore conditional on the authors reconciling the contradictions and providing external grounding for the comparability assumption.

major comments (4)
  1. [Section III and Section IV.B/Table 2] The NTSB phase-of-flight accuracy is reported inconsistently: Section III states that both models achieved accuracy exceeding 68%, with sRNN outperforming ResNet, while Section IV.B and Table 2 report the NTSB sRNN accuracy as 83% and 83.2%, respectively. This is not a rounding difference; it is a 15-percentage-point gap. Because the paper's comparative conclusion that ATSB (87.4%) clearly outperformed NTSB (83.2%) depends directly on this number, the authors must reconcile the two values or explicitly identify which value is correct and why the other appears. As written, the aggregated evidence is internally unstable.
  2. [Section IV.A and Table 1] The damage-level comparison contains two related inconsistencies. First, Section IV.A states that both studies finding accuracies of 88% for the NTSB and ATSB, but Table 1 lists NTSB accuracy as 90% and ATSB accuracy as 89%. Second, the comparability claim that both studies utilize the same methodology and enable a direct side by side comparison is undermined by the manuscript's own caveat that two variables were changed simultaneously: the number of narratives and the narrative length. Section V.A further reports expected narrative lengths of 36.6 words for NTSB and 23.3 words for ATSB, whereas Section IV.A gives average lengths of 107.5 and 20.4 words. These discrepancies mean the reader cannot tell what evidence supports the parity conclusion, and the assertion that dataset size does not affect damage-level accuracy is not supported by the presented numbers.
  3. [Section IV.B and Table 2] The flight-phase comparison across the three studies assumes that all three studies share a similar methodology, but the models tested differ (sRNN and ResNet for [20]; LSTM, CNN, BLSTM, and sRNN for [21]; LSTM, GRU, BLSTM, and combinations for [22]), and Section V.B notes that the ASN dataset contains only accidents while NTSB and ATSB contain both accidents and incidents. These are not minor details: if label distributions, class definitions, or the accident/incident mix differ, the reported accuracy differences (87.4 vs 83.2 vs 64) cannot be attributed to dataset size or source. The paper also gives 67% for the ASN best combined model in Section IV.B.3 but 64% for the ASN BLSTM in Table 2; the relationship between these numbers needs to be clarified. The comparability assumption is load-bearing for the central claim and is not demonstrated.
  4. [Section VI and Conclusions] The conclusion states that larger datasets, like those available from the NTSB, generally lead to higher accuracy in both damage level and flight phase classification tasks. This statement conflicts with the damage-level result in Table 1, where the larger ATSB corpus (50,778 narratives) and the NTSB corpus (27,000 narratives) yield comparable accuracy (89-90%), and with Section IV.A's explicit observation that the larger ATSB dataset did not improve performance. The paper even concedes that the two variables of narrative count and narrative length were changed together, making any inference about dataset size alone impossible. The conclusion overstates what the evidence can support.
minor comments (5)
  1. [Section V.A] The phrase 'narratives expected to have lengths of 36.6 words' for NTSB is unclear; Table 1 does not report narrative lengths, and Section IV.A gives a different average (107.5 words). Please add a consistent definition, e.g., median vs mean, and ensure the numbers match across sections.
  2. [Section V.B] Typographical error: 'sever phases of flight' should be 'seven phases of flight.' Also, the list of phases (approach, enroute, landing, standing, takeoff, taxi, unknown) should be checked against the label sets used in the original studies to ensure they are identical across [20]-[22].
  3. [Figures 2 and 3] The captions for Figures 2 and 3 are nearly identical and both reference 'approach' and 'runway' word sizes; Figure 3 should be described in terms of the LDA topic distributions, not repeated from Figure 2. Please differentiate the captions.
  4. [References] Several reference entries omit the page range or article number (e.g., [15], [19], [24]) or use inconsistent formatting for conference proceedings. Please standardize the bibliography.
  5. [Section IV.C] The sentence 'This includes tokenization, encoding categorical data, and addressing special characters and stop words' is repeated from Section IV.A; the TM section's preprocessing description should be specific to topic modeling (e.g., lemmatization, TF-IDF/Word2Vec) rather than duplicated text.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the comparative findings are summaries of prior primary experiments, with external corroboration.

full rationale

This manuscript is a narrative review and cross-study comparison of the authors' earlier NLP/deep-learning/topic-modeling experiments; it does not fit parameters, derive equations, or generate new predictions, so no step reduces by construction to its own input. The central claim that NLP, deep learning, and topic modeling can improve aviation safety analysis is supported not only by the self-cited studies ([8], [15], [19-25]) but also by independent external sources ([9-11], [27], [28]), so the self-citations are not the sole load-bearing justification. The prior studies are primary experimental reports and are externally falsifiable, which makes them legitimate evidence rather than circular premises. Two genuine concerns appear in the manuscript, but they are correctness or validity issues rather than circularity: (1) Section IV.A explicitly concedes that 'two variables were changed simultaneously, the number of text narratives used to train, and the size of the text narratives,' so the claimed dataset-size comparison cannot isolate its cause; and (2) Section III reports the NTSB phase-of-flight accuracy as 'exceeding 68%,' while Section IV.B and Table 2 report 83.2% for the same sRNN result. These internal inconsistencies weaken the comparative conclusions but do not constitute a self-referential derivation. Accordingly, no circular step is identified and the score is 0.

Assumptions & free parameters 0 free parameters · 2 assumptions · 0 invented entities

No new free parameters or invented entities appear because the paper is a review. The central claims rest on the comparability of the underlying datasets and studies, which is assumed rather than verified.

assumptions (2)
  • domain assumption The NTSB, ATSB, and ASN reports are representative of aviation safety occurrences, and their damage/phase labels are internally consistent across datasets.
    The comparative analysis in Sections IV and V assumes that the same semantic categories (e.g., four damage levels, seven flight phases) can be compared across datasets, though Section V.A itself notes ASN uses only A1/A2 accident classes.
  • domain assumption The reported accuracy numbers from the different studies are directly comparable because the studies used the same preprocessing, model training, and evaluation protocols.
    Section IV.A states 'Both studies utilize the same methodology... enabling a direct side by side comparison' and Section IV.B claims 'All three studies share a similar methodology'. This comparability is asserted, not demonstrated, and Section IV.A later acknowledges that dataset size and narrative length were changed simultaneously.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Utilizing AI for Aviation Post-Accident Analysis Classification." pith.science (2026). https://pith.science/paper/ZRLOAAI4

@misc{pith2026250600169,
  author       = {Pith},
  title        = {Pith review of: Utilizing AI for Aviation Post-Accident Analysis Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZRLOAAI4}},
  note         = {Machine review of arXiv:2506.00169}
}
read the original abstract

The volume of textual data available in aviation safety reports presents a challenge for timely and accurate analysis. This paper examines how Artificial Intelligence (AI) and, specifically, Natural Language Processing (NLP) can automate the process of extracting valuable insights from this data, ultimately enhancing aviation safety. The paper reviews ongoing efforts focused on the application of NLP and deep learning to aviation safety reports, with the goal of classifying the level of damage to an aircraft and identifying the phase of flight during which safety occurrences happen. Additionally, the paper explores the use of Topic Modeling (TM) to uncover latent thematic structures within aviation incident reports, aiming to identify recurring patterns and potential areas for safety improvement. The paper compares and contrasts the performance of various deep learning models and TM techniques applied to datasets from the National Transportation Safety Board (NTSB) and the Australian Transport Safety Bureau (ATSB), as well as the Aviation Safety Network (ASN), discussing the impact of dataset size and source on the accuracy of the analysis. The findings demonstrate that both NLP and deep learning, as well as TM, can significantly improve the efficiency and accuracy of aviation safety analysis, paving the way for more proactive safety management and risk mitigation strategies.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

29 extracted references · 24 canonical work pages

  1. [20]

    Natural Language Processing and Deep Learning Models to Classify Phase of Flight in Aviation Safety Occurrences,

    Nanyonga, A., Wasswa, H., Molloy, O., Turhan, U., and Wild, G., "Natural Language Processing and Deep Learning Models to Classify Phase of Flight in Aviation Safety Occurrences," Proceedings of the 2023 IEEE Region 10 Symposium (TENSYMP), 2023, pp. 1-6

    work page 2023

  2. [21]

    Aviation Safety Enhancement via NLP & Deep Learning: Classifying Flight Phases in ATSB Safety Reports,

    Nanyonga, A., Wasswa, H., and Wild, G., "Aviation Safety Enhancement via NLP & Deep Learning: Classifying Flight Phases in ATSB Safety Reports," Proceedings of the 2023 Global Conference on Information Technologies and Communications (GCITC), 2023, pp. 1-5

    work page 2023

  3. [22]

    Phase of Flight Classification in Aviation Safety Using LSTM, GRU, and BiLSTM: A Case Study with ASN Dataset,

    Nanyonga, A., Wasswa, H., and Wild, G., "Phase of Flight Classification in Aviation Safety Using LSTM, GRU, and BiLSTM: A Case Study with ASN Dataset," Proceedings of the 2023 International Conference on High Performance Big Data and Intelligent Systems (HDIS), 2023, pp. 24-28

    work page 2023

  4. [1]

    Urban Aviation: The Future Aerospace Transportation System for Intercity and Intracity Mobility,

    Wild, G., "Urban Aviation: The Future Aerospace Transportation System for Intercity and Intracity Mobility," Urban Science; Vol. 8, No. 4, Date 2024. doi: 10.3390/urbansci8040218

    doi:10.3390/urbansci8040218 2024

  5. [2]

    Urban Air Mobility Airspace Integration Concepts and Considerations,

    Thipphavong, D. P., Apaza, R., Barmore, B., Battiste, V., Burian, B., Dao, Q., Feary, M., Go, S., Goodrich, K. H., Homola, J., Idris, H. R., Kopardekar, P. H., Lachter, J. B., Neogi, N. A., Ng, H. K., Oseguera -Lohr, R. M., Patterson, M. D., and Verma, S. A., "Urban Air Mobility Airspace Integration Concepts and Considerations," 2018 Aviation Technology, ...

    work page 2018

  6. [3]

    A Concept of Operations (ConOps) of an In-time Aviation Safety Management System (IASMS) for Advanced Air Mobility (AAM),

    Ellis, K. K., Krois, P., Koelling, J., Prinzel, L. J., Davies, M., and Mah, R., "A Concept of Operations (ConOps) of an In-time Aviation Safety Management System (IASMS) for Advanced Air Mobility (AAM)," AIAA Scitech 2021 Forum. American Institute of Aeronautics and Astronautics, 2021

    work page 2021

  7. [4]

    The Regulatory Framework for Safety Management Systems in Airworthiness Organisations,

    Batuwangala, E., Silva, J., and Wild, G., "The Regulatory Framework for Safety Management Systems in Airworthiness Organisations," Aerospace; Vol. 5, No. 4, Date 2018. doi: 10.3390/aerospace5040117

    work page doi:10.3390/aerospace5040117 2018

  8. [5]

    An artificial intelligence approach to onboard fault monitoring and diagnosis for aircraft applications,

    Schutte, P., and Abbott, K., "An artificial intelligence approach to onboard fault monitoring and diagnosis for aircraft applications," Astrodynamics Conference. American Institute of Aeronautics and Astronautics, 1986

    work page 1986

Show all 29 references
  1. [6]

    Intelligent Decision Aiding in the Flight Deck: An Integration of Artificial Intelligence and Human Factors Research,

    Abbott, K., "Intelligent Decision Aiding in the Flight Deck: An Integration of Artificial Intelligence and Human Factors Research," Aircraft Design and Operations Meeting. American Institute of Aeronautics and Astronautics, 1991

    1991

  2. [7]

    Machine Learning for Air Transport Planning and Management,

    Wild, G., Baxter, G., Srisaeng, P., and Richardson, S., "Machine Learning for Air Transport Planning and Management," AIAA AVIATION 2022 Forum. American Institute of Aeronautics and Astronautics, 2022. 9

    2022

  3. [8]

    Sequential Classification of Aviation Safety Occurrences with Natural Language Processing,

    Nanyonga, A., Wasswa, H., Turhan, U., Molloy, O., and Wild, G., "Sequential Classification of Aviation Safety Occurrences with Natural Language Processing," AIAA AVIATION 2023 Forum . American Institute of Aeronautics and Astronautics, 2023

    2023

  4. [9]

    Aviation - BERT: A Preliminary Aviation-Specific Natural Language Model,

    Chandra, C., Jing, X., Bendarkar, M. V., Sawant, K., Elias, L., Kirby, M., and Mavris, D. N., "Aviation - BERT: A Preliminary Aviation-Specific Natural Language Model," AIAA AVIATION 2023 Forum. American Institute of Aeronautics and Astronautics, 2023

    2023

  5. [10]

    SafeAeroBERT: Towards a Safety -Informed Aerospace -Specific Language Model,

    Andrade, S. R., and Walsh, H. S., "SafeAeroBERT: Towards a Safety -Informed Aerospace -Specific Language Model," AIAA AVIATION 2023 Forum. American Institute of Aeronautics and Astronautics, 2023

    2023

  6. [11]

    BERT for Aviation Text Classification,

    Jing, X., Chennakesavan, A., Chandra, C., Bendarkar, M. V., Kirby, M., and Mavris, D. N., "BERT for Aviation Text Classification," AIAA AVIATION 2023 Forum . American Institute of Aeronautics and Astronautics, 2023

    2023

  7. [12]

    Airbus A32x Versus Boeing 737 Safety Occurrences,

    Wild, G., "Airbus A32x Versus Boeing 737 Safety Occurrences," IEEE Aerospace and Electronic Systems Magazine, Vol. 38, No. 8, 2023, pp. 4-12. doi: 10.1109/MAES.2023.3276347

    2023

  8. [13]

    Visual Flight into Instrument Meteorological Condition: A Post Accident Analysis,

    Ayiei, A., Murray, J., and Wild, G., "Visual Flight into Instrument Meteorological Condition: A Post Accident Analysis," Safety; Vol. 6, No. 2, Date 2020. doi: 10.3390/safety6020019

    2020 doi

  9. [14]

    Classification and analysis of errors reported in aircraft maintenance manuals,

    Rogers, B. L., Hamblin, C. J., and Chaparro, A., "Classification and analysis of errors reported in aircraft maintenance manuals," International Journal of Applied Aviation Studies, Vol. 8, No. 2, 2008, p. 295. doi: https://commons.erau.edu/publication/1168

    2008

  10. [15]

    Topic Modeling Analysis of Aviation Accident Reports: A Comparative Study between LDA and NMF Models,

    Nanyonga, A., Wasswa, H., and Wild, G., "Topic Modeling Analysis of Aviation Accident Reports: A Comparative Study between LDA and NMF Models," Proceedings of the 2023 3rd International Conference on Smart Generation Computing, Communication and Networking (SMART GENCON) , 202...

    2023

  11. [16]

    Artificial Intelligence in Aviation Safety: Systematic Review and Biometric Analysis,

    Demir, G., Moslem, S., and Duleba, S., "Artificial Intelligence in Aviation Safety: Systematic Review and Biometric Analysis," International Journal of Computational Intelligence Systems , Vol. 17, No. 1, 2024/11/12 2024, p. 279. doi: 10.1007/s44196-024-00671-w

    2024 doi

  12. [17]

    Natural Language Processing (NLP) in Aviation Safety: Systematic Review of Research and Outlook into the Future,

    Yang, C., and Huang, C., "Natural Language Processing (NLP) in Aviation Safety: Systematic Review of Research and Outlook into the Future," Aerospace; Vol. 10, No. 7, Date 2023. doi: 10.3390/aerospace10070600

    2023 doi

  13. [18]

    Examining the Potential of Generative Language Models for Aviation Safety Analysis: Case Study and Insights Using the Aviation Safety Reporting System (ASRS),

    Tikayat Ray, A., Bhat, A. P., White, R. T., Nguyen, V. M., Pinon Fischer, O. J., and Mavris, D. N., "Examining the Potential of Generative Language Models for Aviation Safety Analysis: Case Study and Insights Using the Aviation Safety Reporting System (ASRS)," Aerospace; Vol. ...

    2023 doi

  14. [19]

    Comparative Study of Deep Learning Architectures for Textual Damage Level Classification,

    Nanyonga, A., Wasswa, H., and Wild, G., "Comparative Study of Deep Learning Architectures for Textual Damage Level Classification," Proceedings of the 2024 11th International Conference on Signal Processing and Integrated Networks (SPIN), 2024, pp. 421-426

    2024

  15. [23]

    Impact of Dataset Size & Data Source on Aviation Safety Incident Prediction Models with Natural Language Processing,

    Nanyonga, A., and Wild, G., "Impact of Dataset Size & Data Source on Aviation Safety Incident Prediction Models with Natural Language Processing," Proceedings of the 2023 Global Conference on Information Technologies and Communications (GCITC), 2023, pp. 1-7

    2023

  16. [24]

    Comparative Analysis of Topic Modeling Techniques on ATSB Text Narratives Using Natural Language Processing,

    Nanyonga, A., Wasswa, H., Turhan, U., Joiner, K., and Wild, G., "Comparative Analysis of Topic Modeling Techniques on ATSB Text Narratives Using Natural Language Processing," Proceedings of the 2024 3rd International Conference for Innovation in Technology (INOCON), 2024, pp. 1-7

    2024

  17. [25]

    Exploring Aviation Incident Narratives Using Topic Modeling and Clustering Techniques,

    Nanyonga, A., Wasswa, H., Turhan, U., Joiner, K., and Wild, G., "Exploring Aviation Incident Narratives Using Topic Modeling and Clustering Techniques," Proceedings of the 2024 IEEE Region 10 Symposium (TENSYMP), 2024, pp. 1-6. 10

    2024

  18. [26]

    Applications of natural language processing in aviation safety: A review and qualitative analysis,

    Nanyonga, A., Joiner, K., Turhan, U., and Wild, G., "Applications of natural language processing in aviation safety: A review and qualitative analysis," AIAA SciTech 2025 Forum. American Institute of Aeronautics and Astronautics, 2025

    2025

  19. [27]

    A Survey Protocol to Assess Meaningfulness and Usefulness of Automated Topic Finding in the NASA Aviation Safety Reporting System,

    Paradis, C. V., Kazman, R., Davies, M. D., and Hooey, B. L., "A Survey Protocol to Assess Meaningfulness and Usefulness of Automated Topic Finding in the NASA Aviation Safety Reporting System," AIAA AVIATION 2021 FORUM. American Institute of Aeronautics and Astronautics, 2021

    2021

  20. [28]

    Sequential deep learning from NTSB reports for aviation safety prognosis,

    Zhang, X., Srinivasan, P., and Mahadevan, S., "Sequential deep learning from NTSB reports for aviation safety prognosis," Safety Science, Vol. 142, 2021/10/01/ 2021, p. 105390. doi: https://doi.org/10.1016/j.ssci.2021.105390

    2021

  21. [29]

    A survey on deep learning tools dealing with data scarcity: definitions, challenges, solutions, tips, and applications,

    Alzubaidi, L., Bai, J., Al -Sabaawi, A., Santamaría, J., Albahri, A. S., Al -dabbagh, B. S. N., Fadhel, M. A., Manoufali, M., Zhang, J., Al -Timemy, A. H., Duan, Y., Abdullah, A., Farhan, L., Lu, Y., Gupta, A., Albu, F., Abbosh, A., and Gu, Y., "A survey on deep learning tools...

    2023 doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.