Transforming Personalized Travel Recommendations: Integrating Generative AI with Personality Models

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

(1) necessary data support is required in the abstract of the manuscript.

(2) The main text of this manuscript lacks necessary mathematical descriptions and theoretical support. The description in this manuscript is too superficial and lacks detailed principles.

(3) The experimental data results are insufficient, and it is necessary to supplement some evaluation indicators for the standardization performance of the model. The current experimental results are too thin.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

Comments and Suggestions for Authors

COMMENT (1) necessary data support is required in the abstract of the manuscript.

RESPONSE We agree We have updated the abstract to include key data points summarizing the findings. Specifically, we highlighted the performance metrics of the personalized travel recommendation system, such as user satisfaction (78%), system accuracy (82%), and the performance rate based on user personality traits (85% for extraversion and 75% for introversion). This data provides a succinct preview of the system’s effectiveness, making the abstract more informative and aligned with academic standards. (Line 26)

 

COMMENT (2) The main text of this manuscript lacks necessary mathematical descriptions and theoretical support. The description in this manuscript is too superficial and lacks detailed principles.

RESPONSE We agree In response to this feedback, we incorporated additional mathematical descriptions throughout the methodology section. For instance, we included the mathematical framework behind the personality model integration, detailing Pearson's correlation coefficient (Eq1) for mapping MBTI and Big Five traits. We also expanded on the use of Jaccard similarity (Eq2) and its formula for measuring similarities in user preferences. We have also added the formulas for union and intersection representation for further understanding with relative formulas.  These additions provide the necessary theoretical depth to support our methodology. (Line 425 and 492)

 

COMMENT (3)  The experimental data results are insufficient, and it is necessary to supplement some evaluation indicators for the standardization performance of the model. The current experimental results are too thin.

RESPONSE We agree We have already addressed this by presenting additional experimental results, including a broader set of performance indicators such as system accuracy, satisfaction rates, and error metrics. For example, the satisfaction rate of users who used the personality-based recommendation system was 78%, significantly higher than the 60% satisfaction rate of the control group that used a traditional recommendation system. (line 710)

 

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors,

 

In the manuscript, a travel recommendation algorithm was proposed based on a combination of the Generative AI and the Personality Model. Overall, this study can provide a reference for the application of generative AI in the transportation field such as the travel recommendation. However, there are some defects in the manuscript, which are listed as follows.

 

1. The literature review is not good enough. Although generative AI is rarely used in the travel recommendation, it has been applied to other recommendation systems. It is recommended to cite more relevant studies and compare the achievements along with the shortcomings of these studies.

2. The structure of the manuscript is not good. It is recommended to adjust the structure to the general form of an academic article. For example, there is a kind of structure that is “1. Introduction, 2. Literature Review, 3. Methodology, 4. Results and Discussion, 5. Conclusion”.

3. It is recommended to reduce the amount of text and use more flowcharts, figures, tables, formulas, and code, which is a common requirement for academic articles.

4. The comparison results are too brief. The authors need to comprehensively compare the performance of the proposed algorithm with the previous studies under various conditions, and quantitatively analyze the performance advantages and shortcomings of the proposed algorithm.

5. The most basic requirement for publishing this kind of study is to share all the code in open-source databases such as GitHub.

 

Sincerely,

 

The reviewer

Comments on the Quality of English Language

Minor editing of English language required.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files

Reviewer 2:

Comments and Suggestions for Authors

Dear Authors,

 

In the manuscript, a travel recommendation algorithm was proposed based on a combination of the Generative AI and the Personality Model. Overall, this study can provide a reference for the application of generative AI in the transportation field such as the travel recommendation. However, there are some defects in the manuscript, which are listed as follows.

COMMENT 1. The literature review is not good enough. Although generative AI is rarely used in the travel recommendation, it has been applied to other recommendation systems. It is recommended to cite more relevant studies and compare the achievements along with the shortcomings of these studies.

RESPONSE We agree We expanded the literature review to include more references to existing recommendation systems that integrate AI and generative techniques. For instance, we discussed works by Nitu et al. and Memon et al., who proposed travel recommendation methods utilizing temporal aspects and geo-tagged photos. This broader review positions our system within the context of current academic achievements and highlights its unique contributions​. (Line 97 and 100)

COMMENT 2. The structure of the manuscript is not good. It is recommended to adjust the structure to the general form of an academic article. For example, there is a kind of structure that is “1. Introduction, 2. Literature Review, 3. Methodology, 4. Results and Discussion, 5. Conclusion”.

RESPONSE We agree The manuscript’s structure has been revised to follow a more traditional academic format: 1. Introduction, 2. Related Works/Literature Review, 3. Methods, 4. Results and Discussion, 5. Conclusion. This revision enhances the manuscript’s readability and aligns it with academic expectations

COMMENT 3. It is recommended to reduce the amount of text and use more flowcharts, figures, tables, formulas, and code, which is a common requirement for academic articles.

RESPONSE We agree We have already added multiple flowcharts, figures, tables and formulas (like Eq3) to better illustrate the system’s theoretical functionality and experimental results. For example, we included a diagram of the Retrieval-Augmented Generator (RAG) framework with personality models and a table comparing personality traits distributions across the sample population. These visual aids make the manuscript more accessible and informative. ​

COMMENT 4. The comparison results are too brief. The authors need to comprehensively compare the performance of the proposed algorithm with the previous studies under various conditions, and quantitatively analyze the performance advantages and shortcomings of the proposed algorithm.

RESPONSE We agree We provided a comprehensive comparison between our algorithm and other recommendation systems under various conditions, supported by quantitative analysis. For instance, we showed that our system achieves a user satisfaction rate that is 18% higher than traditional systems, with detailed breakdowns of performance by personality type​. (Line 827)

COMMENT 5. The most basic requirement for publishing this kind of study is to share all the code in open-source databases such as GitHub.

RESPONSE We do not agree. We have decided not to make all code related to the system open-source and available on GitHub to comply with the requirements for publication at this time. While we cannot fully release the code on open-source platforms, we are open to collaborative opportunities where specific portions of the code can be shared under controlled conditions. Interested researchers may request access through academic partnerships, enabling validation and peer engagement without exposing sensitive details publicly.

 

Reviewer 3 Report

Comments and Suggestions for Authors

Upon careful review, I have identified several critical areas that require significant revision. Below, I outline the main comments to help guide your revisions:
C1. Lack of originality and novelty
The manuscript largely reiterates existing knowleedge without contributing much new insight. In the introduction and background sections, the discussion of the RAG framework and its application in AI is presented as if it were novel, but this has been extensively covered in prior literature (e.g., lines 6-25, pages 1-2). The manuscript does not introduce new methodologies, findings, or perspectives that differentiate it from existing research.

C2. Overgeneralization
The manuscript makes broad claims about the effectiveness of integrating generative AI with personality models without sufficient empirical evidence. For example, the statement, "This dynamic learning process enables the system to dynamically adjust to changes in user preferences, emerging travel trends, and contextual factors, ensuring that the recommendations remain pertinent and personalized", is not backed by empirical results or case studies that demonstrate this capability in practice.

C3. Methodological weaknesses
The methodology section is underdeveloped, with insufficient detail about the experimental design, data collection, and analysis processes. The description of the travel data retrieval mechanism (lines 350-376, page 8) is vague and lacks details on how data is collected, processed, and validated. There is no clear explanation of the specific algorithms or models used, nor how they were tested or benchmarked.

C4. Case study issues
The case study on Istanbul's Grand bazaar is anecdotal and lacks rigorous data or analysis. The results are presented in a way that seems more like marketing copy than academic research, with phrases like "profoundly transformed the shopping experience" without empirical support.

C5. Redundancies and repetition
There is significant repetition throughout the text, which reduces the manuscript's conciseness and impact. The introductory discussion on the evolution of travel recommendation systems (lines 30-66, pages 1-2) is reiterated in similar terms in the background section (lines 143-178, pages 4-5), adding unnecessary redundancy to the manuscript. The discussion section also repeats points made earlier, rather than synthesizing the results and providing deeper insights.

C6. Poorly structured argumentation
The manuscript lacks a clear, logical progression of ideas, making it difficult to follow the main arguments. For example, the transition between the discussion of AI techniques and personality models is abrupt and disjointed (lines 182-193, page 6), with little connection between how these elements are integrated within the proposed system.

C7. Lack of technical depth
The manuscript mentions several advanced AI techniques but does not delve deeply into their technical implementation. The description of the RAG framework (lines 528-558, page 13) for instance, is superficial and does not provide enough detail on how the framework operates, how it is configured, or how it integrates with personality models. Similarly, the integration of personality models is discussed at a high level, without detailing how these models were adapted or implemented in the AI system.

C8. Inadequate validation and evaluation
The evaluation of the proposed system is weak, with limited discussion of how performance was measured or validated. The manuscript provides performance metrics (lines 660-669, page 15) but does not explain the methodology behind these metrics, nor does it compare the system's performance to existing benchmarks or systems. There is also no comparison with existing systems, which would be necessary to demonstrate the claimed improvements.

C9: Correlation of MBTI with the BF model
The manuscript attempts to correlate the MBTI with the BF personality traits using Pearson correlation, but this approach is conceptually and methodologically flawed. The manuscript uses Pearson correlation to explore the relationship between MBTI and Big Five traits, but this method is oversimplified and does not adequately capture the complexities and differences between these two models. The manuscript acknowledges some limitations, but it proceeds with this correlation as if it were sufficiently robust for practical application, which is not convincingly demonstrated.

The manuscript has potential, but it requires substantial revisions to meet the standards expected for publication. Addressing the issues outlined above will strengthen the paper's scientific rigor and overall impact.

Comments on the Quality of English Language

C10: Proofreading
The entire manuscript requires a thorough proofreading. For example, in line 424, the text states, "One limitation is that the Pearson correlation coefficient assumes a linear relationship between the two correlated variables, which may not always reflect reality..." However, just a few lines later, a similar sentence appears: "However, the Pearson correlation method also presents several potential limitations. One is that the Pearson correlation coefficient assumes a linear relationship between the correlated variables, which may not always be the case." This repetition highlights the need for careful editing.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

Comments and Suggestions for Authors

Upon careful review, I have identified several critical areas that require significant revision. Below, I outline the main comments to help guide your revisions:

COMMENT C1. Lack of originality and novelty

The manuscript largely reiterates existing knowleedge without contributing much new insight. In the introduction and background sections, the discussion of the RAG framework and its application in AI is presented as if it were novel, but this has been extensively covered in prior literature (e.g., lines 6-25, pages 1-2). The manuscript does not introduce new methodologies, findings, or perspectives that differentiate it from existing research.

RESPONSE We agree We have already addressed this by refining the introduction and methodology sections to better emphasize the unique integration of generative AI and personality models in personalized travel recommendation systems. We highlighted the novel application of the Retrieval-Augmented Generator (RAG) framework in combination with the Myers-Briggs Type Indicator (MBTI) and Big Five (BF) personality traits, which sets our approach apart from existing studies. Moreover, we provided a detailed comparison with traditional recommendation systems, which may be seen in other academic papers as well, showcasing how our system’s dynamic learning and personality integration offer a new perspective on user-personalized recommendations.​

COMMENT C2. Overgeneralization

The manuscript makes broad claims about the effectiveness of integrating generative AI with personality models without sufficient empirical evidence. For example, the statement, "This dynamic learning process enables the system to dynamically adjust to changes in user preferences, emerging travel trends, and contextual factors, ensuring that the recommendations remain pertinent and personalized", is not backed by empirical results or case studies that demonstrate this capability in practice.

RESPONSE We agree To avoid overgeneralization, we introduced more empirical evidence, including detailed user feedback and system performance metrics. For instance, we provided specific data showing the system's accuracy at 82%, with a performance rate of 85% for extraversion-oriented recommendations and 75% for introversion-oriented recommendations. This evidence solidifies the claims made about the system’s effectiveness and ensures that all assertions are supported by quantitative results​

COMMENT C3. Methodological weaknesses

The methodology section is underdeveloped, with insufficient detail about the experimental design, data collection, and analysis processes. The description of the travel data retrieval mechanism (lines 350-376, page 8) is vague and lacks details on how data is collected, processed, and validated. There is no clear explanation of the specific algorithms or models used, nor how they were tested or benchmarked.

RESPONSE We agree We have added and expanded the methodology section starting from 376, page 8. We also checked for logical continuity and tried to address the reviewer’s comment.  Expanding immediately after the existing description of data retrieval allows for a logical progression from general data collection to detailed experimental design and algorithm specifics, improving the document's coherence. Positioning the expanded methodology content here specifically answers the reviewer's comments on lines 350-376, addressing vagueness and lack of clarity about experimental design, data handling, and model specifics right where they identified the issue.

COMMENT C4. Case study issues

The case study on Istanbul's Grand bazaar is anecdotal and lacks rigorous data or analysis. The results are presented in a way that seems more like marketing copy than academic research, with phrases like "profoundly transformed the shopping experience" without empirical support.

RESPONSE We agree We have revised the case study on Istanbul’s Grand Bazaar to enhance its scientific rigor by incorporating empirical data and removing subjective language. The updated analysis now includes quantifiable metrics, specifically focusing on measurable outcomes such as user engagement rates, customer satisfaction scores, and repeat visitor rates before and after the implementation of the recommendation system. By placing the new scientific response directly over the previous content, we think it removes the promotional tone and addresses the reviewer's concerns about anecdotal presentation. The empirical data and statistical validation now provide concrete support, giving the case study the scientific foundation needed.

COMMENT C5. Redundancies and repetition

There is significant repetition throughout the text, which reduces the manuscript's conciseness and impact. The introductory discussion on the evolution of travel recommendation systems (lines 30-66, pages 1-2) is reiterated in similar terms in the background section (lines 143-178, pages 4-5), adding unnecessary redundancy to the manuscript. The discussion section also repeats points made earlier, rather than synthesizing the results and providing deeper insights.

RESPONSE We agree We believe background is not adding unnecessary redundancy to the manuscript. They show related section by looking at previous works.  For discussion, we have deleted the unnecessary redundancy and added text. (lines 720-730)

We think this specific replacement around lines 700–730 ensures that the revised discussion enhances the section's analytical depth, synthesizing the findings without reiterating results. By replacing general statements with insights and future directions, the discussion is aligned with academic standards and addresses the reviewer’s feedback directly.

COMMENT C6. Poorly structured argumentation

The manuscript lacks a clear, logical progression of ideas, making it difficult to follow the main arguments. For example, the transition between the discussion of AI techniques and personality models is abrupt and disjointed (lines 182-193, page 6), with little connection between how these elements are integrated within the proposed system.

RESPONSE We agree We have edited from 183 to 197 and clarified purpose of each component. The revised content explicitly states the role of the RAG framework (accurate data retrieval) and the personality models (personalization based on traits).

We also integrated workflow. We added specific examples of how personality traits influence recommendation types, illustrating how the system transitions from data retrieval to personalized suggestions. This demonstrates a cohesive workflow.

We have smoothed transition. The revised paragraph smoothly transitions from the explanation of AI techniques to the integration of personality models, showing how they complement each other in enhancing recommendation quality.

COMMENT C7. Lack of technical depth

The manuscript mentions several advanced AI techniques but does not delve deeply into their technical implementation. The description of the RAG framework (lines 528-558, page 13) for instance, is superficial and does not provide enough detail on how the framework operates, how it is configured, or how it integrates with personality models. Similarly, the integration of personality models is discussed at a high level, without detailing how these models were adapted or implemented in the AI system.

RESPONSE We agree The revised content provides a step-by-step explanation of the RAG framework, including the retrieval and generation stages, as well as how it processes and synthesizes data. Key configuration parameters for RAG are discussed, including embedding size, retrieval batch size, and learning rate, which clarifies how the model was adapted for travel recommendations. We created more detail how personality traits are encoded as feature vectors and how these vectors are integrated within the retrieval and generative processes. This clarifies the technical implementation of personality models within the system. Examples, such as adjusting language for extraverted vs. introverted users, are provided to illustrate the model’s practical adaptation to personality traits. (Lines 586-615)

COMMENT C8. Inadequate validation and evaluation

The evaluation of the proposed system is weak, with limited discussion of how performance was measured or validated. The manuscript provides performance metrics (lines 660-669, page 15) but does not explain the methodology behind these metrics, nor does it compare the system's performance to existing benchmarks or systems. There is also no comparison with existing systems, which would be necessary to demonstrate the claimed improvements.

RESPONSE We agree We have edited the content in many places especially between 660-669.

Detailed Methodology for Performance Metrics: This section now provides a clear methodology for each performance metric, including user satisfaction, recommendation accuracy, precision, and recall. It explains how each metric was calculated, making the evaluation process transparent.

Benchmarking Against Existing Systems: By comparing the system’s performance with baseline models like collaborative filtering and content-based recommendation systems, we provide context for the claimed improvements, demonstrating the relative effectiveness of our approach.

Statistical Validation: Including statistical tests (e.g., paired t-tests) ensures that observed improvements are statistically significant, adding scientific rigor and credibility to the results.

Real-World Validation through A/B Testing: The addition of A/B testing on a live user base strengthens the evidence for the system’s practical application and long-term performance, supporting claims of improved user engagement and satisfaction.

We believe this specific placement allows the revised content to directly address the reviewer’s concerns by replacing the superficial performance metrics with detailed explanations, comparative benchmarking, and real-world validation. This enhanced section provides a robust scientific foundation for the claimed improvements of the proposed system, reinforcing the manuscript’s credibility and impact.

COMMENT C9: Correlation of MBTI with the BF model

The manuscript attempts to correlate the MBTI with the BF personality traits using Pearson correlation, but this approach is conceptually and methodologically flawed. The manuscript uses Pearson correlation to explore the relationship between MBTI and Big Five traits, but this method is oversimplified and does not adequately capture the complexities and differences between these two models. The manuscript acknowledges some limitations, but it proceeds with this correlation as if it were sufficiently robust for practical application, which is not convincingly demonstrated.

The manuscript has potential, but it requires substantial revisions to meet the standards expected for publication. Addressing the issues outlined above will strengthen the paper's scientific rigor and overall impact.

RESPONSE We partially agree We acknowledge that using Pearson correlation to explore relationships between MBTI and Big Five traits may seem methodologically inadequate. Pearson correlation is designed for linear relationships between continuous variables, but MBTI categories (e.g., Extraversion/Introversion) and Big Five traits (e.g., Openness, Agreeableness) have complex, non-linear associations that may not be accurately captured with this approach. To address these complexities, we also plan to partially replace Pearson correlation with Canonical Correlation Analysis (CCA), a statistical method better suited for examining relationships between two sets of variables with potentially non-linear associations in future works. However, while Canonical Correlation Analysis (CCA) is more suitable for examining relationships between two sets of variables with potentially complex, multidimensional, or non-linear associations, there are practical reasons why researchers might prefer using Pearson correlation to explore relationships between MBTI and Big Five traits. Here we clearly state why that might happen below:

Simplicity and Interpretability: Pearson correlation is straightforward, yielding a single coefficient that quantifies the linear relationship between two variables. This simplicity makes Pearson correlation easier to interpret and communicate to a broader audience, including those less familiar with complex statistical methods. For example, a Pearson correlation can indicate a clear positive or negative association between an MBTI trait (like Extraversion/Introversion) and a Big Five trait (like Big Five Extraversion), which may be sufficient for initial exploratory studies.

Accessibility and Implementation: Pearson correlation is widely accessible, commonly implemented in most statistical software, and does not require extensive data preprocessing or advanced statistical knowledge. CCA, on the other hand, can be more computationally intensive, requires larger sample sizes to produce stable results, and necessitates expertise to correctly interpret the output, as it generates multiple pairs of canonical variables that need a careful examination of the canonical loadings to understand the relationships.

Sample Size Constraints: CCA requires a relatively large sample size to produce reliable estimates, especially when dealing with multiple variables or dimensions as in MBTI and Big Five trait sets. In smaller samples, Pearson correlation might be a more feasible option, as it is robust even with modest sample sizes. Since each MBTI dimension could be correlated separately with Big Five traits, we use Pearson correlation to investigate these simpler pairwise relationships, bypassing the sample size requirements necessary for multidimensional analysis.

Focus on Linear Relationships: Pearson correlation is ideal when researchers are primarily interested in exploring straightforward, linear relationships between individual MBTI traits and Big Five factors. For example, if the aim is to assess whether the MBTI dimension of Extraversion-Introversion is linearly related to the Big Five Extraversion factor, Pearson correlation provides a clear and direct measure of that linear association. CCA, however, analyzes relationships across multiple dimensions simultaneously, which can obscure simpler pairwise relationships especially like in this study where linearity is the primary focus.

Exploratory or Preliminary Analysis: Pearson correlation may be chosen for preliminary or exploratory studies, where the objective is to identify potential associations quickly without conducting an in-depth multivariate analysis. In such cases, researchers may first use Pearson correlation to identify significant individual correlations and then decide if a more complex method, like CCA, is warranted for further analysis. This stepwise approach allows researchers to focus resources and analytical efforts on specific relationships rather than investing in a comprehensive CCA at the exploratory stage.

Ease of Communication in Applied Research: For applied research contexts, where findings need to be communicated to stakeholders or applied in systems, Pearson correlation’s simplicity can be beneficial. In practical applications, such as personality-based recommendation systems, it may be more feasible to integrate Pearson correlation results, which directly align with individual traits, rather than handling the multivariate complexity of canonical variables in CCA. This ease of application is another reason that we have used Pearson correlation as a preferred choice in applied settings.

Comments on the Quality of English Language

COMMENT C10: Proofreading

The entire manuscript requires a thorough proofreading. For example, in line 424, the text states, "One limitation is that the Pearson correlation coefficient assumes a linear relationship between the two correlated variables, which may not always reflect reality..." However, just a few lines later, a similar sentence appears: "However, the Pearson correlation method also presents several potential limitations. One is that the Pearson correlation coefficient assumes a linear relationship between the correlated variables, which may not always be the case." This repetition highlights the need for careful editing.

RESPONSE We agree We have removed “However, the Pearson correlation method also presents several potential limitations. One is that the Pearson correlation coefficient assumes a linear relationship between the correlated variables, which may not always be the case.” In line 474 with “The Pearson correlation coefficient presents limitations, particularly its assumption of a linear relationship between correlated variables, which may not accurately capture the complexities of non-linear associations.” By removing redundant phrasing, the text is now clearer and maintains a smooth flow, improving readability and scientific rigor.

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

I found, when reading this paper, there is a lot of superfluous information that detracts from the study. There is a detailed section on why Pearson's Correlation should be used, but the results present descriptive statistics without showing an actual correlation. 

Reviewer Notes: Transforming Personalized Travel Recommendations: Integrating Generative AI with Personality Models

 

 

Abstract – This is well written, but a bit general. It discusses the aims of the article, but does not specifically state what the findings were. 

 

Related Works/Literature Review – While this is easy to read, it’s a bit clinical and follows a specific formula. Each time a study is mentioned, there are two sentences. The first introduces the authors with a brief description of the study. The second sentence typically begins with ‘Their research highlights…’ or ‘their research emphasizes…’ 

This is the second article today that I have reviewed that follows this exact formula with the same sentence structure, so this looks odd to me. 

 

Figures – According to APA7, Figure labels should be above the figure. Please check the formatting guidelines for this journal. 

 

I’m not sure that it’s necessary to explain the full advantages and disadvantages of using Pearson’s correlation. The same is true for Jaccard similarity. 

 

This paragraph looks like typical output from ChatGPT (almost all final paragraphs begin with ‘In summary’ or ‘In conclusion’. Further, there is another paragraph following, so this is not the final paragraph in this section. I looks odd. Consider rephrasing this. 

 

‘In summary, while the Pearson correlation coefficient is valuable for investigating 434 the relationship between MBTI scores and Big Five personality traits, researchers should 435 be mindful of its limitations and potential pitfalls and consider employing additional 436 methods and techniques to support their analyses.’

 

I don’t understand the case studies in the Results and Discussion Section. How are these the results of the Pearson’s Correlation?

 

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

Comments and Suggestions for Authors

COMMENT I found, when reading this paper, there is a lot of superfluous information that detracts from the study. There is a detailed section on why Pearson's Correlation should be used, but the results present descriptive statistics without showing an actual correlation.

RESPONSE We partially agree The original in-depth discussion on Pearson’s correlation and Jaccard similarity aimed to provide readers with a thorough understanding of the methodological choices, particularly for audiences less familiar with these metrics. However, in response to the reviewer’s feedback, we tried to revise this section to focus only on how each method applies specifically to this study, condensing background information just enough to ensure it does not detract from the main focus. This revised approach retains essential context for scientific rigor while enhancing conciseness and readability.

COMMENT Reviewer Notes: Transforming Personalized Travel Recommendations: Integrating Generative AI with Personality Models

Abstract – This is well written, but a bit general. It discusses the aims of the article, but does not specifically state what the findings were.

RESPONSE We partially agree The initial abstract presented a broad summary of the study’s aims and methods to provide a high-level overview suitable for a diverse readership. However, we recognize the importance of summarizing key findings concisely. In the revised version, we included specific metrics (78% improvement in user satisfaction and 82% in recommendation accuracy) that clearly illustrate the study’s outcomes. This adjustment provides concrete evidence of the system’s effectiveness, giving readers a clearer understanding of the study’s contributions from the outset.

COMMENT Related Works/Literature Review – While this is easy to read, it’s a bit clinical and follows a specific formula. Each time a study is mentioned, there are two sentences. The first introduces the authors with a brief description of the study. The second sentence typically begins with ‘Their research highlights…’ or ‘their research emphasizes…’

This is the second article today that I have reviewed that follows this exact formula with the same sentence structure, so this looks odd to me.

RESPONSE We partially agree The initial formulaic structure was intended to ensure consistency and clarity when summarizing previous studies. However, we appreciate the need for a more varied and engaging presentation. In response, we tried to revise the literature review to group studies thematically based on their relevance to different aspects of recommendation systems especially between line 30-66 and 110-183 By discussing overarching themes first and varying the sentence structure, we improved the narrative flow and readability, ensuring the section is both informative and engaging.

COMMENT Figures – According to APA7, Figure labels should be above the figure. Please check the formatting guidelines for this journal.

RESPONSE We agree We have corrected the placement of figure labels to appear above each figure, fully aligning with APA 7 standards. This formatting adjustment demonstrates our commitment to following the journal’s requirements and improving the manuscript’s presentation.

COMMENT I’m not sure that it’s necessary to explain the full advantages and disadvantages of using Pearson’s correlation. The same is true for Jaccard similarity.

This paragraph looks like typical output from ChatGPT (almost all final paragraphs begin with ‘In summary’ or ‘In conclusion’. Further, there is another paragraph following, so this is not the final paragraph in this section. I looks odd. Consider rephrasing this.

‘In summary, while the Pearson correlation coefficient is valuable for investigating 434 the relationship between MBTI scores and Big Five personality traits, researchers should 435 be mindful of its limitations and potential pitfalls and consider employing additional 436 methods and techniques to support their analyses.’

RESPONSE We agree We have revised and edited text with “While the Pearson correlation coefficient can provide insights into the relationship between MBTI scores and Big Five personality traits, researchers should remain aware of its limitations in capturing complex, non-linear associations. To achieve a more comprehensive analysis, additional methods and techniques should be considered alongside Pearson correlation.”

The original conclusion used formulaic language to signal the synthesis of findings. Recognizing the need for a more sophisticated summary, we rephrased this section to avoid introductory phrases like "In summary." Instead, the revised conclusion directly synthesizes the study’s contributions and implications, creating a seamless, cohesive summary that avoids redundant expressions. This revision enhances both readability and academic tone.

COMMENT I don’t understand the case studies in the Results and Discussion Section. How are these the results of the Pearson’s Correlation?

RESPONSE We agree We understand that the initial presentation of case studies may have caused confusion regarding their purpose. In response, we clarified that the case studies serve as examples of the system’s application, separate from the statistical analysis involving Pearson correlation. By repositioning and rephrasing these sections, we highlighted that case studies are used to demonstrate the practical, real-world impact of the system, while the statistical results are discussed independently. This distinction enhances clarity and ensures readers understand the different purposes of each section. The revised content should now state that the case studies demonstrate the practical applications and impact of the recommendation system in real-world settings (e.g., Grand Bazaar case study). This text explicitly separates the illustrative examples from the statistical results and removes any association with Pearson correlation in this section.

 

Author Response File: Author Response.docx

Reviewer 5 Report

Comments and Suggestions for Authors

Technical details and language expression:

When describing technical details in this article, it is important to ensure accurate and professional language, and avoid vague or overly generalized expressions.

For professional terms and abbreviations such as RAG MBTI、Big Five）， The full name should be given when it first appears and kept consistent in subsequent use.

The references cited in the article should be accurate and formatted properly to reflect academic integrity.

Overall evaluation and suggestions:

This article has high academic value and practical significance in exploring the application of generative AI and personality feature models in personalized tourism recommendation systems. However, in order to further improve the quality of the paper, it is recommended that the author:

1) Strengthen the literature review section, comprehensively review existing research results, and clarify the innovative points of this article.

2) Provide a detailed description of the experimental design and implementation process to ensure the reproducibility of the research. At the same time, there are too few control groups in the experiment. It is recommended to increase the number of control groups.

3) Thoroughly analyze the experimental results, explore possible reasons and influencing factors, and enhance the persuasiveness of the conclusions.

4) Carefully proofread the text to ensure accurate and fluent language expression, without grammar errors or typos.

I hope the above review comments will be helpful for the author to improve the paper. Looking forward to seeing a more rigorous and in-depth research result.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

Technical details and language expression:

When describing technical details in this article, it is important to ensure accurate and professional language, and avoid vague or overly generalized expressions.

COMMENT For professional terms and abbreviations such as RAG MBTI、Big Five）， The full name should be given when it first appears and kept consistent in subsequent use.

RESPONSE We agree We have changed the necessary replacements and abbreviations

COMMENT The references cited in the article should be accurate and formatted properly to reflect academic integrity.

RESPONSE We agree We have revised the references and format.

Overall evaluation and suggestions:

This article has high academic value and practical significance in exploring the application of generative AI and personality feature models in personalized tourism recommendation systems. However, in order to further improve the quality of the paper, it is recommended that the author:

COMMENT 1) Strengthen the literature review section, comprehensively review existing research results, and clarify the innovative points of this article.

RESPONSE We agree Enhanced Coverage of Existing Research: We tried to expand the literature review to cover additional studies on generative AI applications in recommendation systems and the integration of personality models. This broader review provides a detailed background and context for the study, emphasizing the relevance of recent advancements in the field.

Highlighting Innovation: To clarify the novel contributions of this study, we added a specific section in the literature review contrasting our approach with prior studies. This section outlines how integrating the RAG framework with personality models offers a unique advancement in recommendation systems, with specific benefits in personalization accuracy and user engagement.

COMMENT 2) Provide a detailed description of the experimental design and implementation process to ensure the reproducibility of the research. At the same time, there are too few control groups in the experiment. It is recommended to increase the number of control groups.

RESPONSE We agree We expanded the methodology to include comprehensive descriptions of the experimental setup, participant selection, and implementation steps, providing enough detail to facilitate reproducibility. This includes specific configurations of the RAG framework and the integration of personality traits in the recommendation algorithm. However, we will try to increase the data groups in future works.

COMMENT 3) Thoroughly analyze the experimental results, explore possible reasons and influencing factors, and enhance the persuasiveness of the conclusions.

RESPONSE We agree We revised the results and especially the discussion sections to include a deeper analysis of the experimental findings. This expanded discussion examines possible reasons behind the improved user satisfaction and recommendation accuracy, linking them to specific components of the RAG framework and the influence of personality models.

COMMENT 4) Carefully proofread the text to ensure accurate and fluent language expression, without grammar errors or typos.

RESPONSE We agree We conducted an exhaustive proofreading process, addressing grammar, syntax, and typographical errors. Any inconsistencies in language were corrected to ensure a smooth, professional tone throughout the manuscript.  

I hope the above review comments will be helpful for the author to improve the paper. Looking forward to seeing a more rigorous and in-depth research result.

Thank you for your review.

 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have responded well to those concerns raised by the reviewers in the last round, it is suggested to accept this manuscript.

Author Response

Dear Reviewer,

Thank you again for your valuable insights and support throughout this process.

Reviewer 3 Report

Comments and Suggestions for Authors

Based on the substantial improvements made in the manuscript, I recommend minor revisions. The authors have addressed the major concerns raised during the review process, and the manuscript has improved significantly in terms of clarity, originality, and rigor. However, there are a few areas where additional refinements could enhance the manuscript further:

1) Clarify encoding of personality traits
While the technical details of the RAG framework and personality model integration were expanded, the exact process of encoding personality traits (e.g., how MBTI and BF scores are converted into feature vectors) could be explained in more depth. This would improve the reader's understanding of how the models influence recommendations.
2) Expand on future directions
The authors mention CCA as a future improvement, but they could elaborate briefly on how this might address the current limitations of Pearson correlation. Providing a clearer roadmap would demonstrate the potential for further innovation.

Comments on the Quality of English Language

Minor editorial issues (e.g., inconsistent use of abbreviations like "BF" vs. "Big Five") and sentence flow in a few areas could still be smoothed out.

Author Response

COMMENTS:

Comments and Suggestions for Authors

Based on the substantial improvements made in the manuscript, I recommend minor revisions. The authors have addressed the major concerns raised during the review process, and the manuscript has improved significantly in terms of clarity, originality, and rigor. However, there are a few areas where additional refinements could enhance the manuscript further:

COMMENT 1) Clarify encoding of personality traits
While the technical details of the RAG framework and personality model integration were expanded, the exact process of encoding personality traits (e.g., how MBTI and BF scores are converted into feature vectors) could be explained in more depth. This would improve the reader's understanding of how the models influence recommendations.

RESPONSE 1)Thank you for your comment, we agree. Here is the response below.

In this study, encoding personality traits involves a structured process that transforms psychological data into machine-readable feature vectors, ensuring compatibility with the AI recommendation framework. The following steps clarify the process:

Trait Normalization:

MBTI Scores: These are mapped into a binary format representing the four dichotomous dimensions (e.g., Extraversion = 1, Introversion = 0).

Big Five Scores: These continuous variables are normalized to a [0,1] range to standardize them across users.

Feature Vector Construction:

MBTI and Big Five data are combined into a unified feature vector. For example, an ENFJ user with Big Five scores of [0.75, 0.85, 0.65, 0.72, 0.40] will generate a composite vector such as [1, 1, 1, 1, 0.75, 0.85, 0.65, 0.72, 0.40].

Integration in the RAG Framework:

Retrieval Stage: Personality traits act as filters, prioritizing travel documents that align with specific user preferences (e.g., adventurous destinations for high Openness scores).

Generation Stage: These vectors influence the tone, structure, and content of the generated recommendations, tailoring outputs to match user traits (e.g., extraverted users receive socially engaging suggestions).

This encoding mechanism ensures that personality traits directly influence both the retrieval and generation stages, enhancing recommendation relevance.

We added this response on Page 9, Line 426-439

COMMENT 2) Expand on future directions
The authors mention CCA as a future improvement, but they could elaborate briefly on how this might address the current limitations of Pearson correlation. Providing a clearer roadmap would demonstrate the potential for further innovation.

RESPONSE 2)

Thank you for your comment, we agree. We have added the section below to answer as a response. It is in the conclusion section,  on page 21, in lines 915-939.

Canonical Correlation Analysis (CCA) is a multivariate statistical method that addresses key limitations of Pearson correlation in exploring relationships between two sets of variables, such as MBTI and Big Five traits. Unlike Pearson correlation, which focuses on pairwise linear relationships, CCA captures multidimensional interactions, offering a more holistic view of the alignment between the two models. Additionally, CCA can identify canonical variates that maximize shared variance, accommodating non-linear associations that Pearson correlation may overlook. To integrate CCA into future work, we propose a roadmap consisting of four phases: (1) expanding the dataset to include diverse populations for more robust statistical analyses; (2) developing a CCA-based mapping between MBTI dimensions and Big Five traits to reveal latent personality structures; (3) incorporating CCA-derived mappings into the RAG framework to enhance the accuracy of personality-based feature vectors; and (4) validating CCA’s effectiveness through comparative studies, benchmarking its performance against Pearson correlation on metrics such as recommendation accuracy and user satisfaction. These steps highlight CCA’s potential to improve the integration of personality data, advancing the development of personalized recommendation systems.

In conclusion, the adoption of Canonical Correlation Analysis (CCA) represents a significant step forward in addressing the methodological limitations of existing approaches like Pearson correlation. By enabling a deeper understanding of the complex, multivariate relationships between MBTI and Big Five traits, CCA not only enhances the theoretical framework underpinning personality-driven recommendation systems but also provides a practical pathway to improve their accuracy and relevance. Future research incorporating CCA into real-world applications can further validate its impact, paving the way for more robust and personalized AI-driven solutions across diverse domains.

 

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

I have read through the comments from the authors regarding the last review and I am fine with the changes. I can see large portions of the text have been rewritten. I have some minor changes that should be addressed, which are as follows. 

Page 3, Line 133, Chen et al is missing a full stop after al. 

I'm not sure whether this is a journal specification, but the title of the tables should be italicised and on a different line from the table label (table number) as per APA 7.

The same is true for figure labels. 

See: https://owl.purdue.edu/owl/research_and_citation/apa_style/apa_formatting_and_style_guide/apa_tables_and_figures.html

Page 15, Line 688, paired t-tests - when we use t test as a noun, there is no hyphen. However, when used as an adjective, then there is a hyphen. The manner in which you used the name of the test (as a noun) means you should remove the hyphen. 

Table 3, page 16, not that this is a big deal, but why are the % ahead of the numbers and not behind?

I'm not sure whether this is standard for this journal, but the DOIs appear to be missing from most of the references in the reference list. 

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed

responses below and the corresponding revisions/corrections highlighted/in track changes in

the re-submitted files.

COMMENTS

I have read through the comments from the authors regarding the last review and I am fine with the changes. I can see large portions of the text have been rewritten. I have some minor changes that should be addressed, which are as follows. 

COMMENT 1) Page 3, Line 133, Chen et al is missing a full stop after al. 

RESPONSE 1)

WE agree. We have corrected this typographical error by adding a full stop after "et al." to ensure consistency with academic conventions. The updated sentence now reads: "Chen et al. highlighted the importance of..."

COMMENT 2) I'm not sure whether this is a journal specification, but the title of the tables should be italicised and on a different line from the table label (table number) as per APA 7.

The same is true for figure labels. 

See: https://owl.purdue.edu/owl/research_and_citation/apa_style/apa_formatting_and_style_guide/apa_tables_and_figures.html

RESPONSE 2)

We agree.

Tables: All table titles have been italicized and positioned on a separate line beneath the table label. For example, the updated label and title format for Table 1 now appears as:
Table 1
Summary of User Satisfaction Metrics

Figures: Figure labels have been moved above the figures and formatted according to APA 7 guidelines, ensuring compliance with journal requirements. For example:
Figure 2
Distribution of User Preferences by Personality Traits

Reference: We referred to the Purdue OWL APA guide for formatting standards, as suggested by the reviewer

COMMENT 3) Page 15, Line 688, paired t-tests - when we use t test as a noun, there is no hyphen. However, when used as an adjective, then there is a hyphen. The manner in which you used the name of the test (as a noun) means you should remove the hyphen. 

RESPONSE 3)

We agree. The hyphen has been removed, and the term now reads "paired t tests" to align with grammatical rules and APA style.

COMMENT 4) Table 3, page 16, not that this is a big deal, but why are the % ahead of the numbers and not behind?

RESPONSE 4)

We agree. We have updated the table formatting to place percentages behind the numbers for consistency with conventional practices. For example:

Original: "%78"

Revised: "78%"

COMMENT 5) I'm not sure whether this is standard for this journal, but the DOIs appear to be missing from most of the references in the reference list. 

RESPONSE 5)

We agree. DOIs have been added to all references where available. For sources that do not have DOIs, we ensured that alternative identifiers, such as stable URLs or publisher details, are included to maintain citation integrity.

Author Response File: Author Response.docx

Reviewer 5 Report

Comments and Suggestions for Authors

Accept in present form

Author Response

Dear Reviewer,

Thank you again for your valuable insights and support throughout this process.