Multi-Task Deep Learning Model with an Attention Mechanism for Ship Accident Sentence Prediction
Round 1
Reviewer 1 Report
This paper proposes a multi-task deep learning model with an attention mechanism for predicting the sentencing of ship accidents. This is an interesting work, and the Korea Maritime Safety Tribunal (KMST) dataset was used. My minor issues were listed below. The presentation needs to be improved. Some equations and figures are difficult read due to low resolutions. Comparison with other machine learning or SOTA language models to confirm the advantages of the proposed approach. It would be better to provide more deep analysis rather than accuracy only for the experiments results.
Author Response
Thanks for your appropriated review.
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
It is a very interesting paper and definitely paves the way for further research in this field. However, this research is still at early stages of maturity.
In particular, the authors should explain the methodology and the research approach. It is not clear to the reader, why an entropy loss function was preferred vis-a-vis MAE or MSE (L1 and L2 respectively).Having said that, it is important for the authors to clearly present the methodology as well as results that verify or challenge methodological options.
Last but not least: it is clear that Deep Learning Models could enhance safety as they can consider various criteria. The authors could consider linking their research with works in multi-criteria decision-making in the field of safety, such as of https://scholar.google.com/scholar?cluster=10722828639252927928&hl=en&as_sdt=2005 . Such a link could enhance the visibility of this work and relate it better to the mainstream research.
Author Response
Thank you for your appropriated review
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors proposed a multi-task deep learning model with an attention mechanism for predicting the sentencing of a regional maritime safety tribunal on ship accidents. As in aviation accidents, the verdict on the ship accident event is delivered through a very complex decision process by many professional figures. The commissions analyze a series of events, predictable or unpredictable, and the behaviour of insiders and passengers involved. Consequently, I am not convinced of the practical implications of applying Deep Learning approach in predicting the sentencing of ship accidents.
Author Response
Thank you for your appropriated review
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors have addressed my concersn.
Reviewer 3 Report
The authors better specified the objectives of the paper, giving the real application to their research. In my opinion the paper, in this form, can be accepted as is it.