Relational Graph Convolutional Network for Text-Mining-Based Accident Causal Classification
Round 1
Reviewer 1 Report
- Does the author believe this method will provide a real case scenario used in the future? Will the algorithm be sufficient enough to be used as a worthy tool to ease the reviewing causes of aviation accidents in terms of simplifying the case reports?
- What were the results in cases when the algorithm produced a "wrong" result? Could these false-positive results cause misinterpretation or misleading information for the user of the algorithm?
- The conclusion missing the future direction
Author Response
Dear Editors and Reviewers:
Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “Relational graph convolutional network for text mining-based accident causal classification”. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our research. We have studied the comments carefully and have made correction which we hope to meet with approval. The main corrections in the paper and the responds to the reviewers’ comments are as following:
Reviewer #1:
Comment 1: Does the author believe this method will provide a real case scenario used in the future? Will the algorithm be sufficient enough to be used as a worthy tool to ease the reviewing causes of aviation accidents in terms of simplifying the case reports?
Response: Thanks for your question. The method proposed in this study has strong practical value and significance. Many similar studies have been proved to be feasible and meaningful in practical scenarios, such as References [1,2]. However, it should be noted that the proposed method can only play an auxiliary decision-making role to improve the processing speed of accident investigation reports, but the proposed method cannot completely replace the role of accident experts.
[1] Cheng, M. Y., Kusoemo, D., & Gosno, R. A. (2020). Text mining-based construction site accident classification using hybrid supervised machine learning. Automation in Construction, 118, 103265.
[2] Zhang, F., Fleyeh, H., Wang, X., & Lu, M. (2019). Construction site accident analysis using text mining and natural language processing techniques. Automation in Construction, 99, 238-248.
Comment 2: What were the results in cases when the algorithm produced a "wrong" result? Could these false-positive results cause misinterpretation or misleading information for the user of the algorithm?
Response: Thanks for your question. As mentioned earlier, the field of accident causation classification is still developing, and there is still a certain distance from fully replacing experienced experts. The method proposed at this stage can only play an auxiliary decision-making role, which is essentially equivalent to reviewing all accident investigation reports in the training database and giving the most likely accident causes for the analysis report. But I believe that as the technology continues to mature and advance, the methods used may one day enable the full interpretation and analysis of accident investigation reports.
Comment 3: The conclusion missing the future direction.
Response: Thanks very much for your suggestions. We have added discussions on future work to the revised manuscript.
Once again, thank you very much for your constructive comments and suggestions which would help us both in English and in depth to improve the quality of the manuscript.
Best regards,
Zaili Chen
20/02/2022
Author Response File: Author Response.docx
Reviewer 2 Report
General comment
This study compares the newly developed method combining R-GCN and pre-trained BERT and the conventional methods in terms of the quality of analysis in natural language processing for analyzing the causes of accidents. The results show that the quality of the results is better with the new method. This study provides useful insights for further research using natural language processing in accident analysis.
Specific comments
line 82
There is a discrepancy between the text and the reference list in the use of capital letters in an author's name (ZhOng vs Zhong). Both of them may be correct...
line 196
Schlichtkrull (2018) proposed a R-GCN structure ...
This paper is one of the most important papers for this study, so that the authors may wish to make it explicit as a reference.
Schlichtkrull M., Kipf T.N., Bloem P., van den Berg R., Titov I., Welling M. (2018) Modeling Relational Data with Graph Convolutional Networks. In: Gangemi A. et al. (eds) The Semantic Web. ESWC 2018. Lecture Notes in Computer Science, vol 10843. Springer, Cham. https://doi.org/10.1007/978-3-319-93417-4_38
line230
The text input to BERT will add a ‘CLS’ mark
The term ‘CLS’ seems self-explanatory to those involved, but the authors may wish to indicate for the benefit of the reader that it is derived from “classification”.
Line 451
The author of Ref. 46 is an organization, not individuals, so that the organization should be indicated.
If the values presented in Table 2 and Table 5 have uncertainties, the authors may wish to display that information as well.
Author Response
Dear Editors and Reviewers:
Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “Relational graph convolutional network for text mining-based accident causal classification”. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our research. We have studied the comments carefully and have made correction which we hope to meet with approval. The main corrections in the paper and the responds to the reviewers’ comments are as following:
Reviewer #2:
Comment 1: line 82: There is a discrepancy between the text and the reference list in the use of capital letters in an author's name (ZhOng vs Zhong). Both of them may be correct.
Response: Thanks for your suggestion. ZhOng in the original manuscript was a typo, which we have corrected in the revised manuscript.
Comment 2: line 196: Schlichtkrull (2018) proposed a R-GCN structure. This paper is one of the most important papers for this study, so that the authors may wish to make it explicit as a reference.
Schlichtkrull M., Kipf T.N., Bloem P., van den Berg R., Titov I., Welling M. (2018) Modeling Relational Data with Graph Convolutional Networks. In: Gangemi A. et al. (eds) The Semantic Web. ESWC 2018. Lecture Notes in Computer Science, vol 10843. Springer, Cham. https://doi.org/10.1007/978-3-319-93417-4_38
Response: Thanks for your suggestion. We made a reference error here, because we wrote the first draft in LaTex and forgot to add a reference to the manuscript. I have added them to the revised manuscript.
Comment 3: line230: The text input to BERT will add a ‘CLS’ mark. The term ‘CLS’ seems self-explanatory to those involved, but the authors may wish to indicate for the benefit of the reader that it is derived from “classification”.
Response: Thanks for your valuable suggestions. It is true that the lack of relevant descriptions in the original manuscript may have caused readers confusion, and we have added relevant introductions in the revised manuscript as you suggested.
Comment 4: Line 451: The author of Ref. 46 is an organization, not individuals, so that the organization should be indicated.
Response: Thanks for your valuable suggestions. We have modified the format of the references to specify the full name of the institution publishing the relevant literature. See Ref. 47 of the revised manuscript for details.
Comment 5: If the values presented in Table 2 and Table 5 have uncertainties, the authors may wish to display that information as well.
Response: Thanks for your valuable suggestions. Since 200 pieces of data were used at test time, both precision and recall were calculated with two decimal places after them. The data in Tables 2 and 5 are also rounded to two decimal places, so it is not necessary to show the uncertainties of the relevant data.
Once again, thank you very much for your constructive comments and suggestions which would help us both in English and in depth to improve the quality of the manuscript.
Best regards,
Zaili Chen
20/02/2022
Author Response File: Author Response.docx