Intelligent Recognition of Key Earthquake Emergency Chinese Information Based on the Optimized BERT-BiLSTM-CRF Algorithm

Round 1

Reviewer 1 Report

This manuscript presents an intelligent recognition algorithm based on a BERT pre-trained language model for earthquake emergency information. However, I want to give some comments to improve this draff as follows:

The authors should change some Chinese words to English in Figures 2, 3, table 2, 6.

The authors should indicate the four layers in figure 2.

It is better to explain the components of equations (1), (2), (3), and (5).

The difference between the proposed method and others is adding an attention layer between the CRF layer and the BiLSTM layer. However, the authors must explain it in more detail.

The authors should revise the whole manuscript to improve English grammar and typos.

The authors should discuss, based on the experimental results, table 5.

 

The authors must present the parameter, including accuracy (Acc), accuracy rate (P), and recall (R), in the experimental results, which indicate that the performance of the proposed method will be better.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

1-     I the abstract section, what do the authors means using " the text of earthquake emergency information"? did you mean the corresponding news about recent earthquake on the media? Please clear it up and maybe needs to use a more appropriate alternative word.   

2-     Some sentences need to support with appropriate references, such as line 21, line 25 or 29 and several other statement all over the manuscript. Please check all the manuscript in regarding this issue.  

3-     The author should have mentioned their innovation and its behind idea in the abstract and introduction section in more detail. I cannot find this research novelty in the manuscript. Also, the introduction needs some improvement to provide more information on similar works, the identified gaps, and the need for this study.

4-     Since important information related to earthquakes is usually announced online by official authorities, what is the necessity of Internet monitoring for this issue?

5-     Line 55(line 78), review the past literature a bout convolutional Neural Network (Recurrent Neural Network), but I think the description and base of this method should be introduced before.

6-     Line 136, it would be very appropriate if a flowchart regarding the research process is presented in this section.

7-     Line 183, please show the proposed layer in figure 2

8-     Line 204, did your proposed model can also consider different language? please clarify this in the text. If there are any restrictions in this regard, please state them clearly

9-     In some figure or table, the Chinese language is inserted. I cannot follow them. If it is due to the limitations of the model in identifying the Chinese language, it is suggested to write its Latin equivalent in parentheses

10- I think if the y datum in figure 6 set to 65, the figure presentation improves.

11- How can the prepared model be developed and used in other engineering sciences? Can you give examples for research or application of this model?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

1 - Symbol from china language must be changed in all the text, tables, and Figures;

2 - Models must be compared to experimental data;

3 - Results must be improved based on experimental data.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

Major remarks:

- I have some doubts about the test set construction. In lines 291-296, the Authors stated that 3 additional earthquake datasets are used in the training process. However, in lines 338-342, they wrote that the same data are used in the test part of the process. If it is true, we have serious leakage of data problems.

Minor remarks:
- In my opinion, Figure 2 is too big.
- Convenient standard in mathematical notation is to bold vectors and matrices.
- At the end of most formulas should be a comma or full stop.
- We enumerate only equations that we cite inside the text.
- Finally, you use only the F1 measure. In such a situation, I propose to reduce (remove) the description of other, unused, quality measures.
- Figure 6 is not necessary. In my opinion, Table 6 is sufficient.
- I propose changing Figure 7 to be more similar to other figures in the paper (eg. colors).

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 5 Report

In paper entitled ‘’ Intelligent Recognition of Key Seismic Chinese Information,Based on BERT-BiLSTM Algorithm’’ authors propose intelligent recognition algorithm of earthquake emergency information based on BERT- 4 BiLSTM. There are few issues that need to be addressed by authors before this manuscript could be reconsidered for publication:

1.       Title of the manuscript should be without comma

2.       Title of the manuscript should be changed in a way that no acronyms are used. BERT-BiLSTM algorithm does mean much unless is widely accepted in the scientific community (like, for instance ANN) or it is a newly developed algorithm in the paper (which is not).

3.       Please explain the phrase ‘’mature Internet’’? If possible, It would be good to exclude it from the paper.

4.       I am not sure about the novelty of the paper. Bert model was proposed in 2018, and BiLSTM model has been used from 2015. So, what is the novelty of the paper?

5.       It seems that the term ‘’seismic information’’ is not correct. Maybe the use of term ‘’key information of the recorded earthquakes’’ or even ‘’earthquake news’’ is more justified.

6.       Manuscript has limited potential audience, since the algorithm is developed for Chinese language. Hence, the further potential use of algorithm in some other parts of the world, using English for instance is impossible.

Once these suggestions are taken into account, I will be pleased to review the revised manuscript and reconsider my recommendation.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors make a response to comments. However, there were no changes in the manuscript compared to the old version.

 

Author Response

I'm sorry, but I didn't use "Track changes" to modify the manuscript, which may cause you trouble reading. Now I have displayed the changes in template.tex. At the same time, for the modification of the image, I added the suffix "_old" to the image name before the modification and "_new" to the newly added image. I hope you can further review the draft, put forward your comments

Reviewer 3 Report

Paper can be accepted in the present form.

Author Response

Thank you for your suggestion, which is very important for my manuscript.

Reviewer 4 Report

The authors have responded to all my comments. Overall I am satisfied.

Author Response

Thank you for your suggestion, which is very important for my manuscript.

Reviewer 5 Report

I cannot see any changes in the revised paper. Authors are strongly encouraged to mark all the corresponding changes compared to the version 1 of the manuscript.

 

Author Response

I'm sorry, but I didn't use "Track changes" to modify the manuscript, which may cause you trouble reading. Now I have displayed the changes in template.tex. At the same time, for the modification of the image, I added the suffix "_old" to the image name before the modification and "_new" to the newly added image. I hope you can further review the draft, put forward your comments

Round 3

Reviewer 1 Report

The authors responded to all comments.

However, the references and citing them in the text were not presented. The authors must cite and present all the references in the Reference section.

 

 

Author Response

I am glad to receive your reply. As for the reference, I have included it as an external reference in the zip file submitted as “cites.bib” and quoted it in template.tex. I hope you can confirm it further.

Reviewer 5 Report

Authors provided responses to the recommendations, and made some changes to the manuscript. However there are sill several issues that need to be addressed:

(1) the term ''key seismic information'' is not a proper one and it is not widely used. As suggested previously, authors should use another term instead. This is also valid for the title of the manuscript.

(2) Novelty of the paper is under question. News on the recorded earthquakes are commonly short, containing EQ magnitude, intensity, number of casualties, epicenter region and similar. Hence, one may wonder about the necessity of the suggested algorithm.

(3) About the limitations on the use of the suggested algorithm. If the suggested algorithm can be used for English language (as the mostly used one), authors should show the potential use of the suggested algorithm also for English language.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx