Intelligent Multi-Lingual Cyber-Hate Detection in Online Social Networks: Taxonomy, Approaches, Datasets, and Open Challenges

Round 1

Reviewer 1 Report

In the article entitled "Intelligent Multi-Lingual Cyber-hate Detection in Online Social Networks: Taxonomy, Approaches, Datasets, and Open Challenges", the authors present an extensive bibliographical review on toxic online textual content in social networks, as well as their different classifications

I find the article very interesting, but I think it should include some study of what they establish as future work, since the work seems to be an extensive bibligraphic review.

Finally, as a summary of the above, I believe that the article is acceptable for publication if it improves the following aspects:

1.- The authors include some study of what they establish as future work.

2.- Page 20 shows the following error "as shown in Table 1Error! Reference source not found."

Author Response

Thank you for your time, consideration and valuable comments.


Point 1: The authors include some study of what they establish as future work.
 Done, future work has been expanded. 

Point 2: Page 20 shows the following error "as shown in Table 1Error! Reference source not found."

Done, we have fixed the error. 

 

Reviewer 2 Report

Please add more new 2022 and 2023 related works.

Author Response

Thank you very much for your consideration, and we really appreciate the comments.

Point 1: Please add more new 2022 and 2023 related works.

Done, the dataset and system descriptions sections have been updated with 2022 and 2023 works. 

 

Reviewer 3 Report

My comments are listed below:

1. Please explain how the taxonomy proposed in section 2 would be useful. Please also explain how this classification method can be applied to the existing methods described in section 4.

2. In addition to listing the references, please explain the connections and differences between the references described in section 2 (This comment also applies to the references described in section 3).

3. Please explain the methods described in section 4, in what situations and for what purpose they are used, and what are the advantages and disadvantages of each method?

4. Is there any literature related to the hybrid approach?

Author Response

Thank you very much for your consideration, and we really appreciate the comments. 

Point 1: Please explain how the taxonomy proposed in section 2 would be useful. Please also explain how this classification method can be applied to the existing methods described in section 4. 

The taxonomy proposed in section 2 shows the difficulty of classifying cyber-hate, since there are many ways of doing so. This taxonomy can be useful for classifying datasets and detection approaches so that researchers can focus on those categories that are of interest to them, easily identifying which datasets and algorithms have been developed up to date. Table 1 has been updated with a new column called "Category" in which the taxonomy category to which the described datasets belong to has been included. Moreover, Table 2 has been updated with a new column called "Approach" in which the detection approach has been included. 

Point 2: In addition to listing the references, please explain the connections and differences between the references described in section 2 (This comment also applies to the references described in section 3). 

The references in section 2 are connected in the sense that they are part of the different categories that make up the cyber-hate taxonomy. The first paragraph of subsection 2.2 shows this connection: "There is much more to cyber-hate than meets the eye. For instance, many people once believed that cyber-hate only consisted of physical bullying and name-calling. However, there are ten types of cyber-hate, which range from excluding and gossiping about people to making fun of their race or religion, as shown in Figure 3."
The difference between the references in section 2 is that each of them shows an example that fits into each of the defined categories. This difference is shown by including each reference in the relevant category (exclusion, denigration, flooding, etc.). 

As mentioned in point 1, Table 1 has been updated with the column category to clarify the relation between the datasets and the taxonomy category that they belong to. 

Point 3: Please explain the methods described in section 4, in what situations and for what purpose they are used, and what are the advantages and disadvantages of each method? 

As mentioned in paragraph starting from 474 to 477 "A potential limitation of this approach regarding its classification efficacy, is it's dependency on a domain-specific words presented in a dictionary, also, it needs an automatic methodology for classification and scoring of words to reduce the amount of manpower required for manual scoring of domain-specific words", if you don't have training data and you do have a lexicon it may be a good idea to use the lexicon-based approach. If you have training data, you can use machine learning. 

Point 4: Is there any literature related to the hybrid approach? 
Done, we have updated related works with hybrid approaches also lexicon-based approaches. 

Round 2

Reviewer 1 Report

The authors have made the requested changes

Reviewer 3 Report

The manuscript has improved a lot after revision.