OurSCARA: Awareness-Based Recommendation Services for Sustainable Tourism

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The idea of integrating sustainability information into tourism recommendation systems is certainly interesting and useful.
However, the article suffers from some problems that need to be solved before it can be published:
1) The definition of OurScara (called OurARS in line 61) is based on three elements P, E, and C, which are weighted. Nowhere is it explained how these elements (and their weights) are calculated.
2) The data: Table 1 lists statistics from the data set. Which of these elements are used as the data set to evaluate the different systems? 254 instances per training set and 64 instances per test set are mentioned. What data, what length, what language, ...?
How many of these datasets are related to sustainability? The text states: This includes carbon footprint, temperature, energy consumption, waste production, and eco-certifications (line 199). Examples of the data used would be helpful for the reader's understanding, and information about number, quality, length,...
3) What is the ground truth (line 247)?
4) Formulas 2, 3, and 4 talk about movies  ....
5) Results: the values given in Table 2 are not consistent with your analysis (lines 256-262)
6) What are the real-time data, and how are they collected? To be useful to the reader, information about collection, quality, etc. should be added.
7) The captions in Tables 2 and 3 mention AI-based recommender systems, while the text mentions CF-based. No definitions of what is meant by CF.

In order to be published, the paper must address the aspects of method definition, choice of weights, data set used, evaluation of results, and comparison with other methods.

Author Response

Please check the attachments. Thank you so much for your kind comments

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The main issue concerning this paper is methodological and clarity/reproducibility of the experiments. As the problems appear to be quite relevant, I require a deep rewriting of the paper.

1. The formulae for precision and recall are completely wrong. In fact, this should have been t_p/(t_p+f_p) and t_p/(t_p+f_n) respectively, while it seems that the author has flipped the numerator and the denominator. Still, it is not clear what it means for "No. of recommended movies " and with "No. of similarity movies"

2. As a consequence of 1), the author should also re-compute the scores from tables 2 and 3 by using the right formula. Usually, the higher the precision/recall, the better. Here, the author seems to he used the wrong formula, for which the values are flipped, thus providing misleading information (that the proposed methodology has a lower score).

3. The results are not fully reproducible, as the author does not clarify which are precisely the data processing steps being used to process the data.

4. The given algorithm does not clarify the exact formula by which each component is computed; for example, how can the author represent the cultural relevance as one single number: is this provided by TripAdvisor, or is this something estimated by sentiment analysis within the text and, if so, which is the exact approach being used?. Furthermore, the ranking function seems just a weighted score, for which the author does not reveal how the parametric weights are derived. So, we cannot claim that the proposed algorithm is a methodology surpassing the state of the art in terms of innovation. Furthermore, the tuning algorithm for choosing the parameters is not provided (model training algorithm), as the author only provides an algorithm to basically compute the scoring function, a weighted sum, and to rank the obtained results according to the score (top-k/N filtering).

Given the coarse metric formulation errors, the non-full-disclosure of the training algorithm as well as non-disclosing of the resulting parameters and clarifying how the E/C/R functions are derived and inferred from the data (is the author just obtaining the scores available from TripAdvisor, or do they derive such information from the text?), I regret to reject this paper despite the appreciable intent of providing a societal impact on Vietnam.

Comments on the Quality of English Language

The paper contains numerous sentences that require a deep revision of the text prior to the next submission.

"the Python 3.9 environment" (L. 228): Python is not an environment (Conda might be an environment) but rather a programming language.

"Evaluate algorithms on the Test-set using various evaluation metrics." (L. 245) Just to say one thing, the sentence's subject is missing.

"The formulation of these metrics is as follows." (L. 247) -> The metrics are defined as follows.

"We collect tourist attractions data from TripAdvisor regarding Vietnam country" -> We crawl TripAdvisor for scraping rankings and reviews of Vietnamese tourist attractions.

"The Top-N recommendation is
fixed as ten attractions to calculate the accuracy of the prediction tasks." (L.231) -> I am not entirely sure what the author really means, maybe "among the collected datasets, we focus or analysis on the top-10 tourist attractions (according to what, higher score?), for which we train our model".

Author Response

Please check the attachments. Thank you so much for your kind comments

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

1. It is recommended to add a description of the data in 4.1, preferably to give some visual representation

2.   In Chapter 3, it is recommended to refer to the referenced method.

3. In 4.1,“The statistics of the 219 collected data are described in Table 3”should be“The statistics of the 219 collected data are described in Table 1”.

4. In 4.3, KNN and SVM are typical classification methods.  Why do you use them as the baseline in the recommendation task?

5. Chapter 3 is too vague.  Please add more descriptions.

6. The indicators are used in classification task.  I am confused whether the proposed method is classification method or recommendation method.

Author Response

Please check the attachments. Thank you so much for your kind comments

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

This article proposes an innovative recommendation system called OurSCARA that aims to promote sustainable tourism by incorporating environmental and socio-cultural awareness into attraction recommendations. The system leverages data analytics techniques like sentiment analysis, user profiling, and collaborative filtering to provide personalized recommendations based on user preferences and sustainability criteria. A prototype is implemented and evaluated using TripAdvisor datasets to demonstrate OurSCARA's potential to influence travelers towards more sustainable choices.

 

Strengths:

 

1. The article presents a unique and timely concept of integrating sustainability considerations into tourism recommendation systems. This is an important step towards promoting responsible tourism practices.

 

2. The proposed OurSCARA framework combines various data analytics techniques to provide contextually relevant and personalized recommendations. The inclusion of real-time data sources like weather and local events further enhances the system's capabilities.

 

3. The authors implement a prototype of OurSCARA and evaluate it using real-world TripAdvisor datasets. This lends credibility to their findings and demonstrates the system's practical applicability.

 

4.  The research highlights the intersection of computer science and sustainable tourism, paving the way for further exploration in this emerging field.

 

Weaknesses:

 

1. While the article mentions using techniques like sentiment analysis and collaborative filtering, it lacks in-depth explanations of the specific algorithms employed. More technical details would allow for better assessment and reproducibility.

 

2.  The evaluation is conducted using datasets from a single platform, TripAdvisor. Testing OurSCARA on a wider variety of data sources would strengthen the generalizability of the findings.

 

3. The article does not include a user study to gauge travelers' receptiveness to sustainability-based recommendations. Incorporating user feedback would provide valuable insights into the system's real-world impact.

 

4. The performance of OurSCARA is not compared against existing state-of-the-art recommendation systems. Such a comparison would better highlight the novelty and effectiveness of the proposed approach.

 

Overall, this research article presents a promising direction for promoting sustainable tourism through awareness-based recommendation systems. The proposed OurSCARA framework demonstrates the potential to positively influence traveler behavior by incorporating sustainability considerations. However, the article would benefit from more technical details, diverse datasets, user studies, and comparative analyses to strengthen its contributions. Nonetheless, it lays a solid foundation for future research at the intersection of recommendation systems and sustainable tourism.

Author Response

Please check the attachments. Thank you so much for your kind comments

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The review of the article was timely, and many of the requested revisions were addressed. Unfortunately, not all changes to the article were highlighted, making it difficult to assess the overall changes. While the new version of the article appears to be improved, there are still some areas that need further improvement:

Formulas 1-4 Weights: Each of these formulas has weights. Could you clarify how these weights are determined and specify the values used in the experimental results?

Ground truth: Could you add a sentence explaining how the ground truth was obtained?

Table 2 results: The results in Table 2 have changed significantly from the previous version of the article. What are the reasons for these changes? Also, the description of the results still does not seem to match the values presented.

Definition of real-time data: The definition of real-time data was included in the letter to the reviewers, but not in the text of the article.

Feature extraction method: The feature extraction method described in lines 246-255 needs further clarification. Could you specify the feature extraction method more clearly, e.g. by using keywords (which ones?) or more sophisticated procedures? It seems that in the second step, the scores are assigned manually - could you confirm this?

Addressing these issues will make the article ready for publication.

Author Response

Thank you for your comments. Please check the attactments

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

After revising the manuscript's language, the way weights were assigned to the sentences seemed arbitrary, and no automated way to infer the weights from the text was given. This also motivates why the algorithm comes with no training parameters, as the weights are manually given for each possible component. Still, the author must provide the methodology required for deriving such weights from the text.

Author Response

Thank you so much for your kind comments. Please check the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

Thanks for your contribution.

Author Response

Thank you for your kind comments. It helps us much to complete the manuscripts.