Prediction of Opinion Keywords and Their Sentiment Strength Score Using Latent Space Learning Methods
, Daniel Gómez-Vergel 1,2, Luis Gracia-Expósito 1,2, Jose M. López-López 1,2 and María Vela-Pérez 3Round 1
Reviewer 1 Report
The paper is interesting and well written, and of potential interest for readers of Applied Sciences.
I have just some comments: in Fig. 4 the maximas scores are for K=100 or 200, but authors then prefer K=2-0 (line 251): is it a good compromise between score and speed? Why not K=10?
I would welcome a mor detailed explanation of the meaning and implications of what the performances indicators (MSE, accuracy, F1-score, precision, recall and AUC) aqctually measure.
I would also enlarge the distinction between rational sentiments and emotional ones, as for instance the reported sentence "This phone is awesome" in my opinion carries emotions, it is not like "this phone is cheap/expensive".... A reader should be informed about the criteria used to distinguis between these two classes.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Overall merit: Accept
Title: Prediction of Opinion Keywords and their Sentiment Strength Score using Latent Space Learning Methods
Paper summary:
This work aims to propose an adapted recommendation system based on the prediction of opinion keywords assigned to different item’s characteristics and their sentiment strength scores. This approach has two main advantages, which are prediction of interpretable textual keywords and its associated sentiment (positive/negative) which will help to elaborate a more precise recommendation and justify it, and allows the use of different dictionary sizes to balance performance and users’ opinion interpretability.
Comments to Author:
I found that the paper well organised and the objectives, methodology and results are well written. Therefore, it could be useful for the field. On the other hand, I do not feel confident about related and future works that the authors can rewrite these two sections to be more understandable. Review rating prediction is an important research topic. The problem was approached from either the perspective of recommender systems (RS) or that of sentiment analysis (SA). It world be good if you could discuss recommendation vs SA and explain more about your idea for continuing this method and its further development. According to figures, the diagram captions of figure 6 and 7 need your consideration. Therefore, I think that the paper could be beneficial for the field and I recommend the manuscript for publication after the modification.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
