A Study on the Calibrated Confidence of Text Classification Using a Variational Bayes

Round 1

Reviewer 1 Report

The goal of this paper, as exposed by the authors, is to propose a method that uses Variational Bayes to reduce the difference between accuracy and likelihood in text classification.

The authors present the personal contributions for THIS paper in section 1. Please quantify the main results in the abstract and conclusion sections.

Figure 1 needs further discussion regarding the encoder presented (the encoder shown in Figure A1 is introduced perhaps too late in the article). Please add additional information regarding the scaling and one-hot encoding operations. Describe the blocks for extracts various predictors in the context of text classification and VB-based generative model.

A comparison of the results obtained by the authors with those presented in related work may be included (can be added in the continuation of the discussion section). The authors must consider additional experimental results situations based on various predictors and different VB-based generative models.

The reference section is good, citing new and relevant articles in the research area.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Respected editor

In this paper the authors proposed predictions based on big data are becoming more successful. In fact, research using images or text can make a long-imagined future come true. However, often the data contains a lot of noise, or the model does not account for the data, which increases uncertainty. Moreover, in modern predictive models, the gap between accuracy and likelihood is widening. The gap may in crease the uncertainty of the prediction. In particular, applications such as self-driving cars and healthcare have problems that can be directly threatened by these uncertainties. Previous studies have proposed a method for reducing uncertainty in applications using images or signals. However, although studies that use natural language processing are being actively conducted, there is insufficient discussion about uncertainty in text classification. Therefore, we propose a method that uses Variational Bayes to reduce the difference between accuracy and likelihood in text classification. In this paper, an experiment was conducted using patent data in the field of technology management to confirm the practical applicability of the proposed method. The results showed that the output of the method was stable even when the text label imbalance was severe

The paper can be accepted if the authors response these comments carefully

1) Improve the abstract, introduction and other parts of the paper grammatically.

2) Highlight the novelty that have been presented in the paper

3) The organization of the paper is missing in the paper

4) The methodology of the paper is not clear. Provide more details of the Hypothesis

5) There is a quality of work that has been done by Mohmed R. Ali (https://scholar.google.com/citations?hl=en&user=bjNjsmoAAAAJ&view_op=list_works&sortby=pubdate), You can use this work in introduction and cite also

6) Discussion is so short. Provide more discussion of the results along with more detail of the figures

7) Provide the conclusion with the obtained results

8) Update the references with latest studies 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The paper has been improved sufficiently to be accepted for publication.

Reviewer 2 Report

The paper can be accepted in the present form