Research on the Classification Methods of Social Bots
Round 1
Reviewer 1 Report
The article: "Research on the classification methods of Social Bots" i'ts isteresting study of classification method of social bots. The introduction is well prepared, but needs to be expanded. The introduction needs to be extended with other publications and articles. The description of the test method is correct and understandable. Point: "Social bot Classification Based on Transfer Learning" is also prepared correctly. It discusses the problem extensively. The research results are presented in a sufficient way. Drawings should be enlarged. They must be uniform and consistent. Need improvement. The conclusions do not refer to all the topics discussed in the article. Also need improvement.
Author Response
Thank you very much for the feedback from the reviewers. We have made the following modifications.
Revision 1: The introduction section was not expanded due to limited research data on multi classification of social robots, and no more reference materials can be found.
Revision 2: The drawing size has been adjusted.
Revision 3: The conclusion section has been modified. Add the following content:
The topic expansion module can more richly and effectively express event content, improving the effectiveness of topic blog relevance judgment. The viewpoint sentence recognition method combining generation rules and deep learning can fully leverage the advantages of both methods. It can accurately identify some viewpoint sentences that comply with the social robot blog post generation rules and may not be smooth through keywords, and further identify as many remaining viewpoint sentences as possible through the TextCNN model, greatly improving the accuracy of social robot viewpoint sentence recognition. With the help of transfer learning technology, the training results of blog data of a large number of ordinary accounts can be transferred to the classification model of social robots, so as to obtain more accurate classification results of social robots.
Reviewer 2 Report
Research on the classification methods of Social Bots
===============
The paper proposed an approach for social bot classification using well-established tools/techniques.
The paper is interesting to read.
However, the paper has the following concerns to be fixed for consideration.
1. The open question in the abstract/introduction is unclear. Please rewrite it.
2. The result summary is absent in the abstract.
3. What is the significance of the work? Please highlight it.
4. Please discuss some recent text classfication approaches as follow:
https://arxiv.org/abs/2203.10286
https://link.springer.com/article/10.1007/s10462-021-10093-1
These papers explain some recent methods used for text classification.
5. Existing methods comparison is weak. Please use more existing algorithms and compare with them
6. Please do a class-wise performance analysis. This helps to understand how the model reacts against each class.
7. The paper needs statistical tests eg Wilcoxon or t-test against each other.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 3 Report
CONCLUSION CAN BE MORE ELABORATE
Authors must add more recent and relevant references
Authors must discuss possible future development based on these limitations.
Author Response
Thank you very much for the feedback from the reviewers. We have made the following modifications.
Revision 1: The conclusion section has been expanded. Add the following content:
The topic expansion module can more richly and effectively express event content, improving the effectiveness of topic blog relevance judgment. The viewpoint sentence recognition method combining generation rules and deep learning can fully leverage the advantages of both methods. It can accurately identify some viewpoint sentences that comply with the social robot blog post generation rules and may not be smooth through keywords, and further identify as many remaining viewpoint sentences as possible through the TextCNN model, greatly improving the accuracy of social robot viewpoint sentence recognition. With the help of transfer learning technology, the training results of blog data of a large number of ordinary accounts can be transferred to the classification model of social robots, so as to obtain more accurate classification results of social robots.
Revision 2: The introduction section was not expanded due to limited research data on multi classification of social robots, and no more reference materials can be found.
Revision 3: The conclusion section has been expanded. Add the following content:
In the future, research on emotional polarity discrimination of news commentators' opinions will be increased, which can further identify the public opinion tendencies of social robots. By determining whether they hold a positive or negative attitude to-wards a specific topic, further exploration and research on the purpose of social robots will be achieved.
Round 2
Reviewer 2 Report
Thanks for the revision; however, it is incomplete. The paper has still not addressed the comments (#4 and #7). Also, it is not good writing technique to put 'reference' explicitly in the paper.
Author Response
Thank you for your rigorous comment and we are sorry for our previous negligence. We have made the following modifications.
Revision 4: The content of relevant papers has been added to the introduction. We have identified it in blue font in the paper.
Revision 7: We have expanded the statistical tests in section 3.4.5 and also identified it in blue font.
Round 3
Reviewer 2 Report
Thanks to authors for their hard work to improve the manuscript. Here are minor comments:
1. Please replace the 'Reference[x]' with Author's last name et al. for the paper(e.g., Zhang et al. ) .
2. In the Wilcoxon ranked test, it is unclear which pairs are used to calculate the p-value. Please clarify them in the description. E.g., MCNN vs proposed, MCNN vs VAE.
Author Response
Thank you very much for your recognition of our modification. We have revised the manuscript according to your suggestions.
Revision 1: We have replaced the 'Reference[x]' with Author's last name et al. for the paper and highlighted with yellow background markers.
Revision 2: In the Wilcoxon ranked test, the samples used to calculate the p-value has been further described in section 3.4.5 and also marked with yellow background.
