ABMM: Arabic BERT-Mini Model for Hate-Speech Detection on Social Media

Round 1

Reviewer 1 Report

1). It seems to me that the paper was written in a rush. The analysis of the proposed method takes only 2.5-3 pages (including large figures). I couldn’t find any novelty in this contribution. It is just a simple applications over a big data set. Standard tools, like LSTM and BERT algorithm were used.

2). The authors split the data into a training set (70%) and a testing set (30%). It is better to use a cross-validation analysis.

3). Although section 3.2 is the most important, it is spanned over only 1 page (including the Figure 2). No analytical description of the individual parts of the method is provided. I would expect to see a rigorous analysis related to the (mathematical) structure of the method's parts (i.e., encoders, size of masks, optimization approach, etc.). This section could easily fit in the introduction section (i.e., as a general description of the method) but not in the main analysis of the paper. The main analysis of the paper must include in a very detailed manner all the steps that were taken place during the development and implementation of the method.

4). In the simulations, the authors use standard indices like F-measure, recall, etc. However, I would expect to see a rigorous comparative statistical analysis with other similar methods.

 

5). Are the data available and free downloadable from a link or there are privacy issues related to the data?

 

Author Response

Once again, we want to thank all the reviewers who have taken their precious time and read our paper carefully and provide us with such relevant comments that can help us in improving the quality of the research paper. We are so much grateful to you all for pointing out that our work could be understood in this way and can be considered for publication in this highly reputed Journal. At this stage, we have addressed the comments given and highlighted them in the revised manuscript.

Author Response File: Author Response.pdf

Reviewer 2 Report

In this paper, the authors propose ABMM, an Arabic BERT-Mini Model for Hate Speech Detection on Social Media.

The topic considered by the authors is extremely interesting. The setting of the paper is very good. The authors clearly describe the proposed approach. They, also, present an experimental campaign that aims to demonstrate the accuracy of their framework.

The paper is pleasant to read and interesting.

I have two suggestions for the authors to further improve their work:

- The authors speak highly of the strengths of their approach. However, to make the presentation more objective, it would be interesting to also point out what the weaknesses and limitations of the approach are.

- In my opinion, the proposed approach could be extended to other related research. For example, it would be interesting to evaluate the texts of groups that debate in a polarized and opposing way on the same topic (e.g., pro-vax and no-vax in COVID-19; see the paper "New approaches to extract information from posts on COVID-19 published in Reddit).

The authors could add a "Discussion" section before the conclusions in which they could present their thoughts on these two issues, citing the case of COVID-19 mentioned above and other similar cases they can identify in the literature.

 

 

Author Response

Once again, we want to thank all the reviewers who have taken their precious time and read our paper carefully and provide us with such relevant comments that can help us in improving the quality of the research paper. We are so much grateful to you all for pointing out that our work could be understood in this way and can be considered for publication in this highly reputed Journal. At this stage, we have addressed the comments given and highlighted them in the revised manuscript.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The paper has been improved. However, I think that the statistical comparison between LSTM and CNN-LSTM should be taken into account.

The English must be improved.

 

Author Response

Dear Prof.

Thank you and all the reviewers who took the time to read our paper and provide such relevant comments that can help us improve the quality of the research paper. In response to your comments and the reviewers' comments, we have revised our manuscript. In the new version, the following work has been done:

• All typos and grammar errors have been corrected.
• We have checked the affiliation information including authors name, authors' order, corresponding author, the addresses, and author's
  e-mail in the paper, to make sure that these are correct and consistent with
  your system.
• In the latest version, all changes have been made.
• In response to the referees' comments, we have corrected all typos and grammar errors in the paper. As a result of reviewer 1's comment, we conducted a statistical comparison between LSTM and CNN-LSTM.
• References were checked to make sure they related to the content of the paper.
• All changes in the new version have been highlighted.

Lastly, the paper has been significantly revised and reorganized considering the reviewers’ comments. The grammatical and spelling errors have been checked carefully.

Thank you for your consideration of this manuscript.

Sincerely,

Dr. Elsayed Atlam

satlam@taibahu.edu.sa

Taibah University, College of Computer Science and Engineering, Yanbu, 42353, Saudi Arabia 

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors have made an effort to meet my requirements. The new version of the paper is better than the previous one. Therefore, in my opinion, the paper can be accepted.

Author Response

Dear Prof.

Thank you and all the reviewers who took the time to read our paper and provide such relevant comments that can help us improve the quality of the research paper. In response to your comments and the reviewers' comments, we have revised our manuscript. In the new version, the following work has been done:

• All typos and grammar errors have been corrected.
• In the latest version, all changes have been made.
• In response to the referees' comments, we have corrected all typos and grammar errors in the paper.
• All changes in the new version have been highlighted.

Lastly, the paper has been significantly revised and reorganized considering the reviewers’ comments. The grammatical and spelling errors have been checked carefully.

Thank you for your consideration of this manuscript.

Sincerely,

Dr. Elsayed Atlam

satlam@taibahu.edu.sa

Taibah University, College of Computer Science and Engineering, Yanbu, 42353, Saudi Arabia 

Author Response File: Author Response.pdf