Enhancing Explainable Recommendations: Integrating Reason Generation and Rating Prediction through Multi-Task Learning
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe authors present an interesting proposal for enhancing explainable recommendations. The paper can be improved as follows:
- Please improve Eq. (18). It must be divided in two equations for the sake of clarity.
- Regarding results presented in Tables 1,2,3, it should be noted if these are training, validation or test results.
- Please justify the reason why you conducted ablation experiments only using one dataset.
- Please change figure 4(b), there is no continuity on the independent variable.
- The authors do not mention if they perform independent runs to obtain reported results. It should be justified given the presence of randomness in their proposal.
Author Response
Dear Reviewer
Hope you are doing very well.
Thank you very much, your revised suggestions have helped me a lot. I have made and highlighted the corresponding modifications. Please refer to the attachment or the following picture.
If you have any questions, please do not hesitate to contact me.
Best,
Xiaona
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsMy suggestions follow on the attached text - the paper
Comments for author File: Comments.pdf
I consider the english is correct.
Author Response
Dear Reviewer
Hope you are doing very well.
Thank you very much, your revised suggestions have helped me a lot. I have made and highlighted the corresponding modifications. Please refer to the attachment or the following picture.
If you have any questions, please do not hesitate to contact me.
Best,
Xiaona
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThis paper proposes our paper proposes to apply a recommendation model integrating Graph Contrastive Learning with Transformers within an Encoder-decoder framework. According to the authors it can perform simultaneously rating prediction and reasoning. The performance of the algorithm is verified using three real-world datasets, which is compared with other models found in the literature.
This work falls in a very active research area, although some aspects can be improved as detailed below.
1.- Abstract and Section 1. The authors talk about “recommender systems”, but the paper lacks of a definition of this term. It is also needed to further explain the contexts in which “recommender systems” are used and add some examples.
2.- Line 29: InfoNCE, Line 72: please define GCN. The same in line 79 for SGL,line 83 for Xsimgcl and in other lines of the manuscript.
3.- I suggest to add an Acronyms Section.
4.- Figure 1. Please explain better, and why “bos” instead of “dog”.
5.- Tables 4-6.Please detail the meaning of the scores/numbers shown in these tables regarding BLEU and rogue, since they are cited in the text but not explained.
6.- Please explain in detail the results presented in Table 7, and in particular “Improve”.
7.- Section 6.4. The authors discuss the effect of hyperparameters lambda, but I guess thet the algorithms used require a larger number of hyperparameters. Please develop this part and explain how the values of the other hyperparameters have been selected/optimized.
8.- A section detailing the computational load of the proposed methods is required.
The Reviewer encourages the authors to revise the work based on the suggestions above in order to improve its quality.
Author Response
Dear Reviewer
Hope you are doing very well.
Thank you very much, your revised suggestions have helped me a lot. I have made and highlighted the corresponding modifications. Please refer to the attachment or the following picture.
If you have any questions, please do not hesitate to contact me.
Best,
Xiaona
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for Authors"Enhancing Explainable Recommendations" provides a model to explain recommendations that improves on the accuracy of existing approaches. While this interpretable recommendation model is analysed in the context of e-commerce and digital content recommendations, the model is useful more broadly in other applications. I was recently searching to see if a similar study had been conducted utilising a transformer model, which only goes to highlight the importance of the research. I don't have any substantive criticisms of the article. I would, however, suggest adding a few keywords to increase the exposure of the article. In addition to "recommendation systems", the terms "recommendation algorithms" or "recommendation platforms" are commonly used. Adding these as keywords may result in wider readership.
Author Response
Dear Reviewer
Hope you are doing very well.
Thank you very much, your revised suggestions have helped me a lot. I have made and highlighted the corresponding modifications. Please refer to the attachment or the following picture.
If you have any questions, please do not hesitate to contact me.
Best,
Xiaona
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
Comments and Suggestions for AuthorsThank you for your reply. I still have some minor remarks:
Lines 23-25 contain repeated sentences.
Line 327 and equation (9): ROUGE instead of rouge
Author Response
Dear Reviewer
Hope you are doing very well.
Thank you very much, your revised suggestions have helped me a lot. For the revision of Round 2, I have made and highlighted the corresponding modifications, meanwhile, the revision contents of round 1 been converted into black font. Please refer to the attachment or the following picture.
If you have any questions, please do not hesitate to contact me.
Best,
Xiaona
Author Response File: Author Response.pdf
