Emotion Estimation Method Based on Emoticon Image Features and Distributed Representations of Sentences

Round 1

Reviewer 1 Report

The overall quality of the pursued study is very sound and solid, and the manuscript is well presented. However, there are some crucial points need to be elaborated in more details. For instance, it is quite clear how many data sets were used for training and how many for validation purposes (Table 2 - does not say explicitly). Moreover, Fig 1, 2, 3, 4 show general components of the employed models, but that would be necessary to describe a general model structure of the study using a flow-chart. 

One spelling err "we uses" on page 7 should be fixed.   

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper focuses on emotion recognition based on tweets that contain emotional images and language features. The combination of texts and images are considered for making a decision. However, the novelty of the paper is limited. It is not clear how the proposal would improve the existing works. Furthermore, the experimental study/simulation does not add much value to the overall context. The organization can be improved, and a use-case scenario discussing the benefits in a real-world setting would be useful. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The Authors presented research relevant to proposition of  an emotion recognition method for tweets containing emoticons using their emoticon image and language features.

 

 

Scope of the research is up-to-date and vital in wider international discussions. The paper generally was elaborated clearly and comprehensively introduces the theme, however needs some major improvements.

 

Below a few major comments to the specific assumptions and results:

-           The authors didn’t constitute the scientific thesis or hypothesis;

-           The designed network has been not presented sufficiently. There is lack of detail parameters for hidden layers such as: Max Pool and Dropout .

- There is lack of loss and accuracy metrics of the model during training epochs (function draft)  which is crucial for the NN efficiency analyses.

 

- The  very doubtful issue is that the emoticons are analyzed on the basis of generalized images from 46 pixels to 64-dimensional feature vectors. While you can treat emoticons as binary ASCII codes, without any generalization or loss of information.

 

- The bibliography in the text needs improvements for example Katarznya et al. [7] this is the first name of author nor mentioned in the references

- The abbreviations such as: ecv, ev etc. need to be clearly explained under the figures.

 

The paper needs significant improvement.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Thanks for the revision

 

 

 

Reviewer 3 Report

The paper can be publish.