Exploring the Tourist Experience of the Majorelle Garden Using VADER-Based Sentiment Analysis and the Latent Dirichlet Allocation Algorithm: The Case of TripAdvisor Reviews
, Marc Bidan 3, Dalia Perkumienė 4,*, Marius Aleinikovas 4, Benas Šilinskas 4 and Aidanas Perkumas 4Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsI am presented with a sentiment analysis conducted with the typical method of VADER on a corpus of 18,865 reviews scraped from TripAdvisor of a public garden in Marocco (let's call M this garden). The sentiment analysis is followed by a topic modelling conducted with a LDA.
The results of the sentiment analysis describe a situation where almost uniformely, territorial consumers of the garden report a positive impression of it on Tripadvisor.
The manuscript is presented for the Special Issue Sustainable Tourism and Hospitality: Destination Decision Making and Digitization. M garden has an historical value as national heritage for Morocco and I believe that this study potentially fits well the special issue. At the same time it does not provide a sufficient scientific value to be considered for a Sustainability, so I suggest to reject it. I suggest authors to either submit this manuscript to a journal of cultural or philosophical studies open to mixed and empirical methods, or to rework on their research design to properly fit a journal oriented on scientific conclusions, going beyond descriptive work.
I motivate my rejection by observing how authors characterised their theoretical understanding of the "garden" as an architectural concept, with only weak takes on its function. e.g. I suggest an accurate reading of lines 104-139. Here authors adopt a writing style that clearly fits more the philosophical discussion more then the typical categories of social sciences. While their approach is definitely valid and useful, I believe that the typical reader of the section "Tourism, Culture, and Heritage" of Sustainability would be more comfortable with a characterisation of the function of urban gardens closer to middle-range theories of economic geography or quantitative urban sociology. In general I would suggest to pair the subsequent quantitative method with a prior analytical theoretical framework. For example, are there differences between generalist tourists who wants to just roam around Marrakesh, and tourists who are explicitly interested in visiting M Garden? Try to apply a rationalist framework behind the behaviour of the consumer of M Garden. This will be useful later to check some forms of interpretative hypothesis for the topic model. The review of the literature is impressive but rather disconnected with the conclusions, and it only highlights that other works already adopted similar methodology. Yes, but I believe that some of their proposed inspiration went over purely descriptive analysis and, nevertheless, I expected at least a longer discussion on the implications for policy-making of such analysis for Sustainability.
I suggest to authors to reflect on the fact that, while well-executed, a sentiment analysis on texts that ignored the information of the rating score (what they call, erroneously, "bubble score") is a bit a missed opportunity. I would be much more interested in the distribution of ratings among the three clusters (instead of the VADER statistics, which is redundant if not to certificate that they followed standard parameters). Ratings are more immediate information for the management of gardens, since it would tell managers how many consumers are satisfied looking just at summaries of score. Linear aggregation of these which can be misleading as it usually follow an inverted Beta-Binomial shape, i.e. a U or a J, not a bell-curve, so the correlation with the VADER statistics may tell something on the robustness of, e.g. averaging the scores within timespans, etc.
Even as a purely descriptive method, I am quite disappointed from the results, which need extra step of pruning the set of tokens in the pre-process. I believe that the tokens "garden" and "marrakesh" are uniformative (since we are talking about...a garden in Marrakesh!), and I am undecided about "museum". Even if the algorithm converged into an optimal statistic, the inclusion of these tokens clearly confuses the identificability of the topics. For example, just based on the information provided, I would attempt a radically different interpretation of the topics. Authors says that Positive Topic 1 has to do with visits in the morning, but the word 'morning' does not appear among the top words of it. I find Positive Topic 3 interesting because it is the one touch the concept of heritage ('berbers' and 'yves', referring the Fondation), and yet authors misses this aspect focussing on photography, nevertheless they mention this for Topic 5.
Now, I make this remark to showcase my point of view: I don't think this method at this level of pre-processing revelead something inherently useful for the management of this garden. Authors should also set a sort of condition to evaluate if topics are just collections of semantically uninterpretable words, because in its absence they will always force their biases onto the method. Maybe authors could try to validate their interpretation of topics by presenting the token-weight dataset to a Large Language Model. My impression is that the manuscript fails at its explicit goal of "have a direct impact on the management and marketing strategy of Majorelle Garden" or even to produce qualitative insights on managements of urban gardens in Morocco, North Africa or in general.
So, what to do to improve the research design behind this manuscript? The key factor is that it lacks of contrast. Contrasts are the core of modern science for decision making. For example, authors can try to match the reviews collected for M Garden to reviews collected for another Garden in North Africa, or Europe. This would provide them a 'true' IDF, since in reviews there is no 'document' (yes, there is the review, which is usually quite short). Ideally they could sample a representative dataset of around 20k reviews from e.g. 20 other gardens and then retrive what makes M Garden truly special. Alternatively they could try to isolate time trends.
Finally, keeping within the sphere of descriptive analysis, they could easily investigate the demographics and also a bit of the 'reviewing past' of the users of Tripadvisor, given the information at their disposal. Authors fails to mention that the problem with online reviews is that they are not truly representative of the typical consumer, and they suffer from selection bias. Topic modelling on online reviews is very useful when it allows to say something about the customers, not about the place. Authors are too much focused on M Garden, while they should try to understand, for example if the consumer interested in berber heritage showed consistent interests (e.g. in other national heritage, or in berber moden culture) across their past reviews.
Minor issues:
Equations are not always formatted consistently. In the manuscript I am presented with, sometimes the scalar multiplication is represented through a dot in the center of the line and other times with a dot down. I see this as a minor issue, but it confused me. In Eq. 1 they forgot to mention that inverse frequency along the documents is commonly logged, and if they did not, they overestimated TF-IDF.
Figure 2 and Table 2 are confusing in their descriptions, which seem a bit poor, especially once their information is crossed. I suggest to report in the description of Figure 2 that it refers to the dataset. Probably authors can stress more the n of the groups in the figure, instead of their inner variation (the boxplots).
Author Response
Dear reviewer,
We are honored that you have read and evaluated our work. Your valuable comments have shown us the importance of a number of changes to get the job done right. We support your decision to reject this article, as you had several academic expectations. We would like to thank you sincerely for the time you have devoted to reading our document and for the pertinent observations you have made.
Response from the article's authors:
1
We had already submitted our work to a number of journals specializing in cultural studies, but their response was to submit the article to a journal specializing in tourism and urban studies. Finally, we found the opportunity to submit our work to Sustainability in the special issue.
2
2.1
We accept that our theoretical framework is primarily descriptive and does not necessarily follow the intermediate theories of disciplines such as economic geography or quantitative urban sociology. However, it is relevant that we have established this framework with solid hypotheses, using a methodological tool such as the questionnaire, while adopting a hypothetico-deductive approach to testing these hypotheses. In our article, our approach was purely inductive, where we let our empirical data dictate the reality we sought to represent in our theoretical framework.
2.2
You can see the abstract, keywords and conclusion, from which we have added arguments relating to sustainability and how our work fits into this scope.
2.3
We are limited to what our database provides in the way of previously written tourist reviews. We'd be grateful if you could suggest a way of answering your question "Are there differences between general tourists who simply want to walk around Marrakech and tourists who are explicitly interested in visiting Jardin M?" as we can't see how to do it! We think an online questionnaire might answer your question !!!
3
We have reproduced some passages from our literature review. I also recommend that you consult the section on the work carried out, which we have discussed and explained, as well as the discussion section, which is in-depth, and the conclusion, which contains the implications you are expecting, particularly in terms of sustainability.
4
4.1
We would have liked to explore the perspectives according to which analyzing reviews based on ratings is not part of the methodological objective of our study. Had we followed your suggestion, we would not have needed to subsequently use a sentiment analysis tool such as VADER. Indeed, the platform already categorizes ratings as follows: 1 (very negative), 2 (negative), 3 (neutral), 4 (positive), 5 (very positive). This categorization could be complemented by an analysis of emotions, such as using the NRC lexicon, to detect emotions such as joy, sadness, disgust, fear, anger, and surprise in each rating. These ratings are important indicators in reputation studies, as visitors consulting TripAdvisor view the rating stars as a marker of the quality of the location. It is through these ratings that attractions benefit from ranking. Our main objective in this study is to segment sentiments by category and apply a topic modeling approach to better understand the visitor experience during their visit to the garden. While this article cannot cover all the details of this experience, we have avoided dwelling too much on details since we already have informative outputs.
4.2
For an analysis of the evolution of sentiment, see section 3.1.3 (the newly added figure 5) shows the variation in average scores over time.
5
5.1
Most previous research, including that mentioned in the section on mapping studies of gardens and green spaces, tends not to include examples of visitor comments in their articles. Instead, they prefer to explore general themes through the use of topic modeling. As experts in data science, we would like to emphasize that the results of topic modeling are robust and accurate. A crucial additional step was to extract the comments associated with each topic, enabling us to select at least one example comment for each topic covered. Some terms, such as those you have highlighted, are often not present in every thematic model. This is because thematic models use scores to assess the probability of occurrence of each term in relation to the other terms present in the model. This means that terms that are not strongly related to the main themes, or that occur less frequently in the data, may not be included in the final thematic model. This approach ensures that the terms that are most relevant and representative of the themes identified by the model are the ones that are prioritized, offering a more accurate and meaningful representation of the data analyzed.
5.2
We understand your ambition to further explore several dimensions, in particular by deepening the section on topics within Large Language Models. However, we have deliberately chosen our methodology to orient the results in line with our research objectives. For example, by opting for the application of BERTopic, we could have generated several different topics, explored the data in depth and potentially created clusters to identify themes linked to positive or negative feelings. We would also have considered positioning these clusters on a two-dimensional map crossing density and centrality to assess their importance and impact… Nevertheless, we recommend that you read the discussion for a better understanding of the practical impact of our study. We have been rigorous in our methodological choices, remaining true to our research objectives.
6
6.1
We would like to point out that the database collected on the platform has significant gaps in terms of tourists' geographical and demographic data. Age and socio-professional categories are not available on TripAdvisor. It might be interesting to conduct interviews with the managers of Majorelle Garden, as they have valuable statistics that could complement our analysis in terms of visitor profiles.
6.2
Thank you for consulting the section on the limitations of our study. We have included an important consideration here: online reviews do not always represent the typical consumer and may suffer from selection bias.
6.3
The assertion that topic modeling of online reviews is mainly useful for analyzing customers rather than places is limiting. In reality, this technique provides valuable insights into both customer behaviors and the qualities or shortcomings of locations. Overlooking information about the places would mean neglecting essential data needed to improve the offered experience and optimize services. Analyses from online reviews reveal trends and specific needs crucial for strategic development and continuous improvement of locations. Examples include:
Review analysis: Reviews left by visitors on platforms like TripAdvisor or Google Reviews are analyzed through topic modeling to identify recurring themes. These themes may include aspects such as the beauty of the garden, the organization of visits, the cleanliness of facilities, the quality of café service, or the souvenir shops.
Insights on customers: Topic modeling may reveal that visitors particularly appreciate the tranquility and beauty of the landscapes at Majorelle Garden, with frequent mentions of the variety of plants and the vibrant color of the blue walls. These themes highlight what visitors value, offering perspectives on their preferences and expectations.
Insights on the location: Emerging themes can also indicate areas for improvement. For example, if many reviews mention crowding or long queues, this suggests that the garden could benefit from better management of visitor flow or expanded opening hours.
Strategic action: Based on these analyses, the management of Jardin Majorelle might consider increasing staff during peak visit times or improving signage for better circulation. Moreover, by detecting frequent praise for certain elements like the peace and beauty of the place, management could emphasize these aspects in their marketing to attract even more visitors.
6.4
The main objective of our study is to explore the tourist experience of visitors to the Majorelle Garden. We welcome your suggestion to broaden our perspective to include aspects such as consumer interest in Berber heritage. However, we prefer not to stray into the specific details of each garden facility or architecture, in order to maintain our focus on the overall analysis of the visitor experience.
7
7.1
Error corrected: Yes, in the equation (1), the dot (⋅) represents multiplication. This formula calculates the TF-IDF value of term ii in document j, combining the term frequency (tf(i,j)) with the inverse document frequency (idf(i)) through multiplication. The result, j, reflects the importance of term i in document j relative to its occurrence in all documents. Multiplying these two values gives the TF-IDF score, which is a numerical statistic designed to reflect the importance of a word for a document in a collection or corpus.
7.2
We have adjusted the interpretation, ensuring that the table 2 completes the box diagram (Figure4).
Reviewer 2 Report
Comments and Suggestions for AuthorsRemove references from the abstract.
The facts in the introduction should be confirmed by references.
e.g. This is the Majorelle Garden, created in 1931 by the 49 French Orientalist artist Jacques Majorelle, which covers 8,000 square meters and reflects 50 the traditional Moroccan style, with its use of bright colors, geometric patterns and a mix 51 of 300 plant species from five continents.
74 - The impressive length of the observed period.
97, 148, 200, 205, 313, 331, 335, 336, 345, 356, 363, 382, 388, 468 etc. - We will - use passive voice
250-251 - The facts in the case study should be confirmed by scientific references.
e.g. "Marrakech's Majorelle Garden ranks as the second most beautiful garden in the 250 world"
Table 4 - YSL reference - What is that ? (write Yves Saint Laurent at least once)
I agree with the authors that it would be nice to perform a geographic and demographic analysis, based on the cultural origin of tourists, as well as their age groups and professional social categories.
It's nice that the authors used different methods to get the results and numerical values, but I'm sure that the same conclusions would be reached by simply reading the comments on TripAdvisor.
Every third reference was written in the past 5 years. Try adding more relevant references.
Author Response
Dear reviewer,
We are honored to have our work read and evaluated by you. Your valuable comments and suggestions have greatly helped us to improve our work. We would like to thank you sincerely for the time you have devoted to reading our document and for the pertinent observations you have raised.
The recommended modifications in the article are highlighted in red!
Pending a favorable response, we are open to new possibilities for improvement if necessary.
Response from the article's authors:
1
We have removed the references that appear in the abstract.
2
We have included references to all information relating to the Majorelle Garden in the introduction and in the section (3.1. Case study).
3
For the duration of the data, we used in our case study, it extends from 2006 to 2023. To our knowledge, no previous research has been carried out on the case of Jardin Majorelle. This garden stands out as the most attractive tourist attraction in Marrakech, Morocco and Africa, making it unique in terms of available data.
4
We have rendered the requested texts in the passive voice
5
We have replaced YSL with Yves Saint Laurent.
6
We would like to point out that the database collected on the platform has significant gaps in terms of tourists' geographical and demographic data. Age and socio-professional categories are not available on TripAdvisor. It might be interesting to conduct interviews with the managers of Majorelle Garden, as they have valuable statistics that could complement our analysis in terms of visitor profiles.
7
Yes, that's right! There are several methods of sentiment analysis and subject modeling, but the difference generally lies in the degree of performance in terms of accuracy of results. The VADER and LDA tools are frequently used and recognized for their robustness, particularly in the study of customer feedback.
8
We have added several references as you requested.
Reviewer 3 Report
Comments and Suggestions for AuthorsPlease justify the limited generalizability of your case study. it may not be possible to apply the findings to a broader population or context.
justify as well the subjectivity that my be arised and the introduction of bias into the study, which might affect the validity and reliability of the findings.
minor
Author Response
Dear reviewer,
We are honored to have our work read and evaluated by you. Your valuable comments and suggestions have greatly helped us to improve our work. We would like to thank you sincerely for the time you have devoted to reading our document and for the pertinent observations you have raised.
The recommended modifications in the article are highlighted in red!
Pending a favorable response, we are open to new possibilities for improvement if necessary.
Response from the article's authors:
We invite you to consult the research limitations section where you will find a section added in response to your comments. This section includes a complete response to the limit of generalizability of the results, as well as to the validity and relevance biases of the results
Reviewer 4 Report
Comments and Suggestions for Authors1. Lines 205-228, the review of visitors' perceptions of parks needs to be revised, as it is currently a simple list of studies with lack logical relationship to each other. More details of each case need to be reflected and summarized according to themes, key results, etc.
2. Line 238, please add the title of Table 1. And the second column of Table 1 is unnecessary, which could be replaced with more useful information such as the name, category, and scale of the case study.
3. Section 3.1, it is suggested that supplementary figures represent the following information, on the one hand, to add the map of the geographical location of the park, on the other hand, to select a few typical views of the park (echoing the authors’ description of it as "the second most beautiful garden in the world" in the text).
4. Equation (1), are the two variables multiplying each other? Looks like a full stop.
5. Section 5.1 to 5.3, this part of the content is a description of the results, not a discussion. It is recommended to move the relevant text to section 4.2.
6. Accordingly, the discussion part is not deep enough and needs to be modified and supplemented. It is suggested to carry out the discussion from the following aspects: First, compare the results of this study with other existing studies to analyze the similarities or differences. Second, on the basis of section 5.4, further in-depth analysis of the policy recommendations. Third, after reading the manuscript, I am very interested in (I believe other readers also want to know) the changes of tourists' evaluation of the park over time. I suggest adding an analysis of the changes of three sentiment categories (positive, negative and neutral) at different time stages.
Author Response
Dear reviewer,
We are honored to have our work read and evaluated by you. Your valuable comments and suggestions have greatly helped us to improve our work. We would like to thank you sincerely for the time you have devoted to reading our document and for the pertinent observations you have raised.
The recommended modifications in the article are highlighted in red!
Pending a favorable response, we are open to new possibilities for improvement if necessary.
Response from the article's authors:
1
We have explained each study and presented its main results, in order to make this framework for analyzing previous studies more comprehensible.
2
We have replaced the column of research titles with a column presenting the objective of each research. We have also added a column indicating the practical implications of each research listed in the table.
3
In this section (3.1. Case study) of the presentation of our case study of the Majorelle Garden, we have added two figures representing actual brochures produced by the garden. The first brochure presents the Berber Museum (Figure 1), while the second (Figure 2) is a reduced version of a panel located inside the garden, helping to understand the internal structure of the garden and its various installations.
Following your recommendation, we also tried to produce a geospatial map for the garden. However, we encountered technical problems when converting the map into English, as the map configurations in Morocco are all in French by the Google company. It seems to us that figure 2 is sufficient, what do you think?
4
Yes, in the equation (1), the dot (⋅) represents multiplication. This formula calculates the TF-IDF value of term ii in document j, combining the term frequency (tf(i,j)) with the inverse document frequency (idf(i)) through multiplication. The result, j, reflects the importance of term i in document j relative to its occurrence in all documents. Multiplying these two values gives the TF-IDF score, which is a numerical statistic designed to reflect the importance of a word for a document in a collection or corpus.
5
We have moved these elements to the results section and developed discussions.
6
We have redone the discussion section, which contains all your recommendations. For an analysis of the evolution of sentiment, see section 3.1.3 (the newly added figure 5) shows the variation in average scores over time.
Author Response File: 
Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsDear authors,
Thank You for Your kind words.
I will still suggest to reject your manuscript. I do this with a sentiment of empathy towards your efforts in satisfying reviewers with different backgrounds and making different requests.
Unfortunately, I see two clearly distinct problems with this manuscript.
The first problem concerns the hybrid nature of its content. As you wrote back to me, you first considered to submit it to journals in Humanities, which rejected it for its high specificity. Unfortunately, no matter the extension of quantitative analysis presented, it is still a full descriptive analysis of a specific place. I mentioned to you already that, for my own understanding, a descriptive analysis is not useful and interesting for the audience of Sustainability.
I suggested to you to consider your current analysis a pilot for a grounded theorisation on the tourism of gardens in North Africa/Mediterranean but then verify these contrasting your findings with a larger sample including other public gardens, e.g. in North Africa. I do not value positively circumstantial information provided in new Fig. 1 and Fig. 2 because they do not show "social action", which is the object of social sciences, but "historical facts". I understand that the reasons to include these new figures may be related to requests from another reviewers...
Again, I am just trying to convey with honesty my understanding of the scientific role of the type of article your manuscript aims to be, and such hybridisation between history, management and computational social science is relatively hard to evaluate objectively.
Nevertheless, this is just my own understanding of the current circumstances. It is possible that this specific Special Issue could actually be welcoming of your contribution. If not, I can suggest you to propose a revision of your manuscript to journals dedicated to the History of North Africa. Alternatively, you can try on journals specialised in Digital Humanities, too.
Personally, I enjoyed most of your rework of the introduction and conclusion, but I deem important to explain a key point that seems still missing in your understanding of the limitations of the manuscript. I wrote that the scientific insight of online reviews lies in understanding the agency of people, not in understanding (latent) features of the object under review. You rebutted against this argument, but I am not sure if you understood the roots of my argument in the self-selective nature of all online reviews.
For example, in marketing generally a random sample of people is gathered to test a new product. It is extremely important to specify before the experimental phase if these will be allowed to influence each other through communication, or if they must rate the item independently. The classic approach is the second.
In online reviews, the sample of review is biased, but not because it is drawn only from TripAdvisor (and not, e.g. Google), but because people who like to online review are quite different from independent, unengaged, samples. This is extremely well studied, see for example a classic:
https://dl.acm.org/doi/10.1145/1562764.1562800
or two modern approached which I enjoyed much:
https://doi.org/10.1080/13683500.2019.1599828
https://doi.org/10.1177/0022243720941832
Practically it is rather obvious that customer who rate it on TripAdvisor will have positive sentiments about the Garden; for example because they already read about the garden on it, they enjoyed what they've seen, and they just reified the general consensus they were reading about. You don't really catch a counterfactual here.
I think that if you understand this problem (something called "collider bias of self-selection) you could understand why I am skeptical about the possibility to quantify objective features of Garden M with your LDA, and why I suggested that if you wanted to keep this a descriptive analysis, then you should focus more on the features of the customers, over than features of the place, to provide useful insights to managerial science. How to study the customer behaviour? I would start by scraping their past reviews... where have they been? Are they always positive? Is there someone who is usually grumpy and hypercritical, but was not in reviewing M Garden???
This introduces the second core problem: I still think your technical analysis is poor and potentially misspecified.
You have not addressed properly the TF-IDF formula. In the mainstream specification of IDF from K. Jones, the IDF ratio is logged:
https://www.emerald.com/insight/content/doi/10.1108/eb026526/full/html
You do not specify this detail, and if you did not log the number, your results are questionable, also considering the stretched definition of "document" (these are collections of reviews, not of large documents).
Just to be short on this technical evaluation, your Topic Model just keeps repeating the same tokens along the topics you present; then you improvise a posteriori interpretations. I understand that you just followed an automated procedure to establish the number of topics, but I cannot just trust the results. You should seriously consider the possibility that there are no topics at all in your corpus, and people just repeat more or less the same things. This suspect on the technical validity of your results haunts me. My suggestion is to really prune out a lot of trivial words ("museum", "garden", "place", "marrakesh") and keep checking the robustness of your number of topic.
Author Response
Dear reviewer, thank you for the extra effort you put into reading our replies
1
We would like to respond to your comments by underlining the legitimacy of our work and demonstrating its relevance to the themes of the special issue of Sustainability magazine entitled "Sustainable Tourism and Hospitality: Destination Decision Making and Digitization".
Our research responds to a request for publication in this special issue by directly addressing several of its priority axes, notably "Tourist Experience", "Big Data and Data Mining" and "E-Tourism". Our study focuses on the tourist experience at Majorelle Garden, a case that has not been addressed by researchers before. This garden is the top-rated tourist attraction on TripAdvisor in Marrakech, Morocco and Africa, as the links and images show. The Majorelle Garden deserves to be studied as a unique case, given the data between 2006 and 2023, which is an interesting timeframe.
Don't you accept unique cases? Is this illegal in scientific research???
https://www.tripadvisor.com/Attractions-g293734-Activities-Marrakech_Marrakech_Safi.html
https://www.tripadvisor.com/Attractions-g293730-Activities-Morocco.html
https://www.tripadvisor.com/Attractions-g6-Activities-Africa.html
2
You mentioned that our work doesn't fit in with the audience of the "Sustainability" journal. However, I think if you look at this link and the journal's editorial objectives, you'll see that it's a generalist journal covering many areas of scientific research.
https://www.scimagojr.com/journalsearch.php?q=21100240100&tip=sid&clean=0
3
Figures 1 and 2 have been added at the request of other reviewers. However, it's important to note that sometimes we play the role of an archaeologist to dig up information and enrich the single case. The inclusion of these figures in the case description section serves several purposes:
- Historical and Cultural Context: They provide a detailed overview of the historical and cultural significance of the Berber Museum and Majorelle Garden
- Visual Guidance: They provide visual cues to help researchers and readers better understand the layout and key elements of the garden.
- Documentation: They serve as visual documentation, enriching the analysis and description of the case by providing tangible evidence of the elements discussed.
4
Is it illegal to conduct a single case study using a multidisciplinary approach? we note that Sustainability is also a generalist journal that encourages research within this framework.
How can we talk about an interesting garden without mentioning its history, particularly that of the culture highlighted in its museum, such as Berber culture…? The study focuses on the tourist experience, combining the concept of tourism with that of experience, a term frequently supported in marketing research. The study also aims to help the management of garden-type attractions, which falls within the field of management. Finally, it tackles the data science aspect by combining statistics, natural language and machine learning…
we think the multidisciplinary nature of our study adds value to our case study by analyzing real feedback from garden visitors.
5
Logically, do you believe that the profile of this article would be suitable for a journal on the history of North Africa? And still far from the logic of digital humanities journals!!! a highly subjective prescription that does not reflect the article's research purpose!!!!
6
You claim that the scientific value of online reviews lies in understanding people's agency (i.e. their actions, motivations, and behaviors), rather than in understanding the (latent) characteristics of the object being evaluated (the garden). ààààà Our study focuses on exploring the experience of visitors to Jardin Majorelle by analyzing their reviews on TripAdvisor. These reviews help us to understand how visitors perceive and experience their visit. In this sense, the scientific interest of online reviews may lie both in understanding visitors' motivations and behaviors (the agency of people) and in evaluating the garden's features (the object being evaluated).
In the limits section, we have added the text for the auto-selection limit. (In blue color)
7
Please check this point, as we have no problems with TF-IDF. The process shown is the same as the one adopted. We have included the final formula directly in our paper
8
Regarding your comments on the results and the technical process adopted:
https://www.tripadvisor.com/Attraction_Review-g293734-d477277-Reviews-Jardin_Majorelle-Marrakech_Marrakech_Safi.html
you can retrieve the data yourself and carry out the analysis, in accordance with the methodology, to dispel any doubts you may have about the problems you suspect with our results.
Thank you once again!!!
With kind regards ?
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsThe authors have made acceptable revisions on their manuscript and addressed most of my comments and concerns. I have no further comments.
Author Response
Dear Reviewer,
Thanks for a positive review.
Our kind regards
Authors
Round 3
Reviewer 1 Report
Comments and Suggestions for AuthorsI will shortly explain why I keep rejecting this manuscript.
This manuscript is very well suited for a venue focused on the history and urban geography of gardens or/et N. Africa. It is suited for a venue in digital humanities too.
It has almost nothing of interest for a venue of interdisciplinary or social sciences. With my comment I do not intend to dismiss Humanities, but just to state that their methods have a different scope from what I would associate with the journal Sustainability. Within the theoretical framework you propose, your method falls into digital methods for History.
I already provided examples of what differentiate social sciences from history, and I will reiterate it again.
History provide us facts. Social sciences provide us models to relate these facts and understand their mutation over time and space.
I suggested to you that if you want to publish in a journal of interdisciplianry / social sciences you must provide a control study.
I will be very explicit about what you should do to achieve a result sufficient for this study to be considered useful for managerial tasks.
It seems to me that your results suggest that M Garden:
- Is liked because it aesthetically pleasing
- It is disliked because (sometimes) there is mismanagement of queues, maybe related with linguistic barriers
Notice that you can get to this conclusion even without a LDA, just by looking at tf-idf of terms in the two groups (large group of positive reviews vs small group of negative reviews)
So there is a clear research question for management. Is queues + linguistic barriers really problematic?
In my opinion, you should setup a synthetic control downloading reviews from comparable gardens (possibly in N. Africa) and see if the issue of queues + linguistic barriers emerge in the control.
If yes, you can state that these are commonal problems; if not the result is more interesting because it could act as a signal of specific mismanagement.
I just provided you an example of a research question proper for social sciences, not humanities. It would still be a "case study", but the method would be in line with a scientific approach.
Finally, no matter the diagnostics, if you want to stick into your LDA being informative you should run a robustness analysis pruning more words. I don't understand why you refuse to do that.
Finally, I encourage you to upload your dataset and your code on online repositories so everybody can review them if necessary. As a review, I will commit to trying to reproduce your analysis, even if it is beyond my duties to produce robustness analysis for it.
Author Response
Dear reviewer,
We thank you once again for your openness to further discussions in evaluating our work. We regret that, despite our continuous exchanges, this does not lead to the acceptance of our article on your end, as you have already insisted on its rejection for reasons that seem reasonable to you. From our side, we have responded positively to all other reviewers, and even for you, we have taken several of your remarks into consideration to improve the paper. We believe we have completed this work to a scientifically acceptable level. We are convinced that our article makes a significant and well-structured contribution to the existing literature on online reviews. Furthermore, this contribution includes the unique case of the Majorelle Garden, which we are the first to explore in this context.
Yours sincerely
Author Response File: Author Response.pdf
