A Text Classification Model via Multi-Level Semantic Features

Round 1

Reviewer 1 Report

The article from a methodological point of view meets the expectations of a scientific article. The authors establish clear categories of analysis and comprehensively explain the procedure used machine learning and deep learning models. The methodology is also very clear and well worked out from a practical point of view.

Minor.

Lines 65-79 are for the contribution of this paper.

I had a little bit of uncertainties as follows:

In this paper, the authors proposed the structure made with the methods by upgrading and improving the various methods previously presented. I wonder if this structure, which has many existing methods, is improved only for Chinese classification. This is because the data-set used in the experiment are from THUCNews, LTNews and MCNews. And, if so, how did the authors get the insight that the proposed structure would be more suitable than any other method for classifying Chinese?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The contribution of the paper is low since the methods employed are well known and results are not interesting from a practical standpoint. Literal presentation is poor, written style has room for improvement. Several recent approaches were omitted in literature review. Performance evaluation of the methods do not include significant figures of merit for the proposed application. The discussions have not been done in depth.

- A rationale of the selection of the methods and tuning of all the parameters an explanation of the selected features (from the user standpoint) should be added. In addition, bibliographic review is not comprehensive. Several recent approaches should be reviewed.

-         The introductory part of this paper did not systematically address the problems of existing methods, and the highlights of this work were not highlighted in the paper.  It is recommended that the authors further summarize the contributions of this work from the common problems of existing methods.

-    The motivation is not clear. Please specify the importance of the proposed solution

-  More mathematical analysis and related equations should be given.

-    The "Conclusions and Future Work" section appears excessively concise, then the authors should expand it by recapping all the steps to the proposed work, as to offer a brief but complete summary of it to the readers.

-  The overall presentation is not bad. Though, there are lots of clarification the authors need to make.

-   Discuss the reasons why the proposed work is offering better results in comparative analysis with the other similar methods.

- it is not clear if the authors have adopted a cross-validation criterion or something similar in order to limit the impact of the data dependency in the experimental results because the adoption of a simple data-split criterion does not produce reliable/valuable experimental evidence: the author should clarify this aspect;

- The obtained experimental results must be more deeply discussed and justified. The Results and Discussion section must be enriched.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

1.     he authors have correctly addressed the issues I made w.r.t. to the previous version of the manuscript. I thank the authors for their effort in improving the manuscript. The only pending issue is a thorough revision of language. The writing still shows quite a few deficiencies and I strongly recommend a professional proofread before publication.