A Study of the Evolution of Haze Microblog Concerns Based on a Co-Word Network Analysis

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article “A Study of the Evolution of Haze Micro-blog Concerns Based on the Co-word Network Analysis” present a study that analyze through topic modeling social media posts related to haze phenomenon.

The article explain the methodology used to analyze social media posts about air pollution and discuss the results achieved.

The author need to employ state of the art topic modeling techniques such as BERTopic and/or Top2Vec for topic detection instead of the current methods employed. Given that these transformer-based embeddings methods handle social media content in a more effective way since they can identify semantic relationships between words and phrases from contextual understanding, and can also give significant advantages for analyzing topic evolution.

In the literature review section the authors give an overview of existing literature on topic modeling, however the transformer-based embeddings methods are missing.

In the figures’ captions the authors should no state the source as “Author’s construction” since is given that the authors have created the figures.

Tables 3 to 10 should be attached as supplementary material and use a more compact way to represent the results or at least show only one of the locations.

Figures 4-10 are too many consider remove some of them. At line 214 the authors should use “as shown in Figure 1” and start the paragraph In a better way.

 

 

Comments on the Quality of English Language

Please rephrase text from 219 to 223, the paragraph does not flow and it is not clear.

At line 365 use Sankey diagram instead of Sankey map.

Please rewrite text from 327 to 330, Sankey diagrams do not have horizontal or vertical axes, the connection between nodes (not blocks) are called links.

Author Response

Responses to reviewer’s comments

We would like to express our sincere appreciation for your professional and insightful remarks on our manuscript. The feedback has been invaluable in enhancing the quality of our work. We have carefully considered each comment and have made significant revisions accordingly. We believe these amendments comprehensively address the concerns you have raised. In the submitted revised version, changes to our manuscript were all highlighted within the document by using red-colored text. The revisions and itemized responses to editors and reviewers are presented as follows. Responses to the comments are highlighted in blue for clarity. 

For the manuscript, we have conducted further optimizations. We have engaged professionals to polish the language of the manuscript, elevating the level of the writing. In the Introduction, we revised specific sections to better highlight the study's scientific questions, research objectives, and contributions. For the Literature review, we organized relevant studies from the perspectives of research content and perspective to strengthen the theoretical foundation. In the Material and methods, we refined the details, including data preprocessing steps, the advantages and limitations of the methods, and the accuracy of terminology. Additionally, we adjusted the structure of the paper for better clarity. In the Discussion section, we provided a deeper analysis of the results and their practical implications, improving the overall completeness of the manuscript. We also proposed several policy recommendations for environmental management based on the findings. Lastly, we thoroughly outlined the study’s Limitations, offering insights for future research and guidance for scholars interested in this topic. Point-by-point response to comments and suggestions for authors are as follows:

Reviewer(s) Comments:
Reviewer: 1

Comments to the Author
The article“A Study of the Evolution of Haze Micro-blog Concerns Based on the Co-word Network Analysis”present a study that analyze through topic modeling social media posts related to haze phenomenon. The article explain the methodology used to analyze social media posts about air pollution and discuss the results achieved.

Response: Thank you for your feedback. We appreciate your insightful comments of the positive aspects of our methodology and acknowledge the need for revisions. The comments offered have been immensely helpful. We have responded to every question, indicating exactly how we addressed each concern or problem and describing the changes we have made. The revisions have been approved by all authors. The point-to-point responses to your comments are listed below in blue.

Comments 1: The author need to employ state of the art topic modeling techniques such as BERTopic and/or Top2Vec for topic detection instead of the current methods employed. Given that these transformer-based embeddings methods handle social media content in a more effective way since they can identify semantic relationships between words and phrases from contextual understanding, and can also give significant advantages for analyzing topic evolution.

Response 1: Thank you very much for your thoughtful and insightful suggestion. We appreciate your recommendation to consider advanced topic modeling techniques such as BERTopic and Top2Vec. These transformer-based methods offer significant advantages in capturing semantic relationships and analyzing topic evolution, which are indeed valuable for social media content analysis.

Our current study employs a community-based co-word network approach, which involves TF-IDF keyword extraction, co-occurrence matrix construction, and the Louvain algorithm for community detection. This approach was specifically chosen for its strong visualization capabilities. It allows us to effectively present the importance of topic keywords and their co-occurrence relationships in an intuitive manner, which is crucial for understanding the focus of discussions within each topic community.

While we recognize the advanced capabilities of BERTopic and Top2Vec, our chosen method aligns closely with our research objectives. It excels in visualizing topic networks and providing a clear representation of keyword significance and co-occurrence relationships. The precision can still be improved and strengthened for more accurate topic modeling and evolutionary analysis. This is particularly useful for determining the central theme within a themed community and its evolution over time.

Our team is considering comparing different topic modeling methods to quantitatively measure the topic recognition accuracy and effectiveness of different methods, providing theoretical references for method selection in similar research. Your feedback is invaluable and will guide us in considering additional methodologies to complement our current approach.

Thank you once again for your valuable input, which will help improve the rigor and depth of our research. 

Comments 2: In the literature review section the authors give an overview of existing literature on topic modeling, however the transformer-based embeddings methods are missing.

Response 2: Thank you for your valuable suggestion. We have carefully revised the literature review section to include a discussion of transformer-based embeddings methods, such as BERTopic and Top2Vec. See lines 198-207. These methods represent a significant advancement in topic modeling, particularly in terms of their ability to capture deeper semantic relationships and improve accuracy in topic detection.

We also highlighted that these methods offer higher precision and effectiveness in topic identification compared to traditional approaches. While our current study focuses on a community-based co-word network approach due to its strong visual representation capabilities, we fully recognize the potential of transformer-based methods and will consider incorporating them into future research to further enhance our topic modeling analysis.

We appreciate your insightful feedback, which has helped us strengthen the comprehensiveness of our literature review.

Comments 3: In the figures’ captions the authors should no state the source as “ Author’ s construction” since is given that the authors have created the figures.

Response 3: Thank you very much for your careful review and insightful suggestions. We appreciate your attention to detail regarding the figure captions. In response to your comment, we have removed the phrase "Author's construction" from all figure captions.

We have updated the captions to ensure clarity and consistency throughout the manuscript. Should you have any further suggestions or require additional modifications, please do not hesitate to let us know.

Thank you once again for your valuable feedback.

Comments 4: Tables 3 to 10 should be attached as supplementary material and use a more compact way to represent the results or at least show only one of the locations.

Response 4: Thank you for your valuable feedback and thoughtful suggestion. In response, we have revised the presentation of our data to enhance clarity and conciseness. We have retained only one table, which serves as a representative example, to streamline the content and focus on key results. This adjustment aims to present the data in a more compact and accessible manner while ensuring that the information remains comprehensive and useful.

We hope this revision addresses your concern effectively. If you have any further recommendations or require additional modifications, please let us know.

Thank you once again for your input.

Comments 5: Figures 4-10 are too many consider remove some of them. At line 214 the authors should use “as shown in Figure 1” and start the paragraph In a better way.

Response 5: Thank you for your constructive feedback. We appreciate your suggestion regarding the figures. In response, we have streamlined the number of illustrations by removing some of the figures to enhance the clarity and focus of our presentation. This reduction helps ensure that each figure contributes effectively to the overall narrative.

Additionally, we have updated line 214 to reference "as shown in Figure 1" as recommended. We have also restructured the paragraph to improve its clarity and coherence, ensuring a smoother flow of information. Please refer to the revised content on lines 240-251.

We believe these changes address your concerns and improve the readability of the manuscript. Should you have any further suggestions or require additional adjustments, please let us know.

 

Thank you once again for your valuable input.

 

Comments 6: (1) Please rephrase text from 219 to 223, the paragraph does not flow and it is not clear.

• At line 365 use Sankey diagram instead of Sankey map.
• Please rewrite text from 327 to 330, Sankey diagrams do not have horizontal or vertical axes, the connection between nodes (not blocks) are called links.

 

Response 6: (1) Thank you for your suggestions on our manuscript. We have removed the original wording and placed the description of the flowchart before Figure 1. See lines 240-251.

• Thank you for your suggestions on our manuscript. The term "Sankey map" at line 365 has been replaced with "Sankey diagram" to maintain accuracy in terminology.
• Thank you for your suggestions on our manuscript. We have rewritten the text from lines 327 to 330 to correct the references to Sankey diagrams. Specifically, we have removed mentions of horizontal or vertical axes and used the correct terms, See lines 416-421. It is worth noting that the blocks in the article refer to the rectangular color blocks representing the topic community in the Sankey diagram, which are more like blocks than nodes.

 

Thank you again for your helpful suggestions. We believe these revisions address your concerns and enhance the accuracy of our manuscript. If further adjustments are needed, please let us know.

 

We hope that these revisions adequately address the concerns raised by the reviewers and align with the expectations of the editorial team. We sincerely appreciate your time and effort in evaluating our work and look forward to the opportunity for further improvement.

 

Best regards,

Haiyue Lu et al.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors,

The manuscript is framed around the novel application of topic evolution analysis to haze-related social media discussions, emphasizing its ability to track topic development over time and space. The findings suggest that topic evolution analysis can provide valuable insights for urban management and policy-making, particularly in addressing environmental concerns like haze pollution. The authors advocate for the use of social media data in public opinion research and recommend further studies to refine the methodology, offering a more dynamic and comprehensive understanding of how these concerns evolve over time and across different regions. By addressing these areas below, the manuscript could be significantly strengthened, making its contribution to the field even more impactful.

1. The paper could be significantly enhanced by incorporating more quantitative analysis, such as statistical measures to assess the effectiveness of the co-word network in comparison to other methods. This might include comparing the accuracy or relevance of topics identified using different techniques.

2. The limitations section (Lines 671-677) should be expanded to address potential biases in social media data, such as the demographic skew of Weibo users or the influence of censorship. Acknowledging these limitations would add depth and credibility to the analysis.

3. It is recommended to include a more detailed exploration of the evolution of topics over shorter time intervals in the discussion. This would help identify the immediate impact of unexpected events on public attention.

4. Adding a sentiment analysis module is suggested to examine changes in public sentiment across different topics, which would provide insights into the psychological impact's depth and breadth.

5. Since most key users on the selected social platforms are concentrated in China, it would be beneficial to integrate data from other global social media platforms (e.g., Twitter or Facebook) to verify the generalizability of the research findings.

6. Considering that only Chinese keywords were selected for analysis, it is advisable to include the analysis of multilingual microblogging data to broaden the applicability and influence of the research results.

7. Incorporating more recent studies on social media analysis in environmental science would strengthen the theoretical foundation of the research.

8. Provide a more detailed description of the preprocessing steps, including any challenges encountered and how they were addressed, to enhance the transparency and reproducibility of the study.

9. Further exploration of the reasons behind the polarization of haze issues between heavily and lightly polluted areas is needed, potentially considering socio-economic factors.

10. Clarify why the study period was limited to 2013-2019 and whether the inclusion of more recent data could yield additional insights.

11. Discuss how the findings could be applied to other environmental issues beyond haze, such as water pollution or climate change, to expand the scope of the research.

12. Explain the process of data verification from Weibo, particularly how the authenticity and relevance of the analyzed posts were ensured.

13. Address any limitations of the Louvain algorithm in detecting topic communities, especially when handling large-scale social media data.

14. Explore the potential inclusion of other temporal analysis methods, such as time-series analysis, to complement the cosine similarity approach.

15. Discuss how data sparsity was managed in the text clustering method and whether this could impact the study's findings.

16. Ensure that all references are current and relevant, particularly in the rapidly evolving fields of environmental science and social media analysis.

Comments on the Quality of English Language

Moderate editing of English language required.

Author Response

Responses to reviewer’s comments

We would like to express our sincere appreciation for your professional and insightful remarks on our manuscript. The feedback has been invaluable in enhancing the quality of our work. We have carefully considered each comment and have made significant revisions accordingly. We believe these amendments comprehensively address the concerns you have raised. In the submitted revised version, changes to our manuscript were all highlighted within the document by using red-colored text. The revisions and itemized responses to editors and reviewers are presented as follows. Responses to the comments are highlighted in blue for clarity. 

For the manuscript, we have conducted further optimizations. We have engaged professionals to polish the language of the manuscript, elevating the level of the writing. In the Introduction, we revised specific sections to better highlight the study's scientific questions, research objectives, and contributions. For the Literature review, we organized relevant studies from the perspectives of research content and perspective to strengthen the theoretical foundation. In the Material and methods, we refined the details, including data preprocessing steps, the advantages and limitations of the methods, and the accuracy of terminology. Additionally, we adjusted the structure of the paper for better clarity. In the Discussion section, we provided a deeper analysis of the results and their practical implications, improving the overall completeness of the manuscript. We also proposed several policy recommendations for environmental management based on the findings. Lastly, we thoroughly outlined the study’s Limitations, offering insights for future research and guidance for scholars interested in this topic. Point-by-point response to comments and suggestions for authors are as follows:

Reviewer(s) Comments:
Reviewer: 1

Comments to the Author
The manuscript is framed around the novel application of topic evolution analysis to haze-related social media discussions, emphasizing its ability to track topic development over time and space. The findings suggest that topic evolution analysis can provide valuable insights for urban management and policy-making, particularly in addressing environmental concerns like haze pollution. The authors advocate for the use of social media data in public opinion research and recommend further studies to refine the methodology, offering a more dynamic and comprehensive understanding of how these concerns evolve over time and across different regions. By addressing these areas below, the manuscript could be significantly strengthened, making its contribution to the field even more impactful.

Response: Thank you for your feedback. We appreciate your insightful comments of the positive aspects of our methodology and acknowledge the need for revisions. The comments offered have been immensely helpful. We have responded to every question, indicating exactly how we addressed each concern or problem and describing the changes we have made. The revisions have been approved by all authors. The point-to-point responses to your comments are listed below in blue.

Comments 1: The paper could be significantly enhanced by incorporating more quantitative analysis, such as statistical measures to assess the effectiveness of the co-word network in comparison to other methods. This might include comparing the accuracy or relevance of topics identified using different techniques.

Response 1: Thank you very much for your insightful suggestion. We truly appreciate your thoughtful feedback and acknowledge the value that more quantitative analysis would bring to the study. While the current research has primarily emphasized the qualitative strengths of the co-word network approach, particularly its ability to visually and intuitively map relationships between topics and keywords, we agree that incorporating statistical measures would provide a more comprehensive evaluation of the methodology.

At this stage, our focus has been on the interpretability and practical applicability of the co-word network, especially in the context of public attention analysis, where the visual representation of topic communities plays a crucial role. However, we recognize that conducting a quantitative comparison, such as using metrics like precision, recall, topic coherence, or clustering accuracy, would allow us to rigorously assess the effectiveness of our method in relation to others, such as BERTopic or Top2Vec, which rely on transformer-based embeddings, which has a good theme modeling effect.

 

It is worth noting that our team is considering comparing different topic modeling methods to quantitatively measure the topic recognition accuracy and effectiveness of different methods, providing theoretical references for method selection in similar research. we intend to expand our analysis by applying quantitative measures. This would allow us to compare the accuracy, relevance, and topic coherence of the co-word network with other advanced topic modeling techniques. By doing so, we hope to better balance the interpretability of the visual method with the performance advantages of statistical measures, ultimately leading to a more robust and well-rounded analysis.

We greatly appreciate your suggestion and will incorporate this direction into our future work to further enhance the analytical depth and robustness of our study.

Comments 2: The limitations section (Lines 671-677) should be expanded to address potential biases in social media data, such as the demographic skew of Weibo users or the influence of censorship. Acknowledging these limitations would add depth and credibility to the analysis.

Response 2: Thank you for your suggestions on our manuscript. We have expanded the limitations section, elaborated on the potential biases generated by using social media data such as Weibo, and provided feasible future research directions. See lines 912-938.

Comments 3: It is recommended to include a more detailed exploration of the evolution of topics over shorter time intervals in the discussion. This would help identify the immediate impact of unexpected events on public attention.

Response 3: Thanks for your valuable feedback. In our study, we chose winter as the time scale for topic identification and modeling, primarily because winter is the peak season for haze and is highly representative of public concerns regarding air quality. Analyzing topic evolution over multiple years allows us to effectively capture the long-term attention and changing trends of the public on air quality issues. The winter haze phenomena are typically closely related to climate conditions, pollutant emissions, and people's living habits. Therefore, by analyzing data from this specific time period, we can identify significant trends and patterns.

However, we recognize that our study's limitation lies in not focusing on a finer time scale, such as quarterly, monthly, or even daily intervals. We discuss this in detail in the limitations section. See lines 901-911. At the same time, we also elaborated on the impact of the event on public attention in the discussion section, such as policy releases, severe pollution incidents, etc.

Future studies will consider using more granular data to examine real-time concerns, hot events, and other factors that influence public attention during haze occurrences. For instance, analyzing Weibo data over shorter periods will help us understand the immediate impact of specific events—such as extreme weather, policy changes, or public health crises—on public attention. This approach will not only reveal shifts in public sentiment and opinion but also provide more targeted recommendations for policymakers.

Comments 4: Adding a sentiment analysis module is suggested to examine changes in public sentiment across different topics, which would provide insights into the psychological impact's depth and breadth.

Response 4: Thank you for your valuable suggestions regarding the addition of a sentiment analysis module. We appreciate the insight into how sentiment analysis could enhance the understanding of changes in public sentiment across different topics.

In a related study, we have already conducted sentiment analysis on Weibo posts related to haze, specifically to assess the impact of haze pollution on public mental health. This analysis provides a detailed evaluation of the psychological effects of haze on the public, and we believe it complements the findings of the current study. Please refer to the literature for details: Lu, H.; Rui, X.; Gemechu, F. G.; Li, R. Quantitative Evaluation of Psychological Tolerance Under the Haze: A Case Study of Typical Provinces and Cities in China with Severe Haze. Int. J. Environ. Res. Public Health. 2022, 19, 6574. https://doi.org/10.3390/ijerph19116574.

Comments 5: Since most key users on the selected social platforms are concentrated in China, it would be beneficial to integrate data from other global social media platforms (e.g., Twitter or Facebook) to verify the generalizability of the research findings.

Response 5: Thank you for your valuable feedback. We understand the importance of verifying the generalizability of our research findings across different social media platforms. Indeed, integrating data from global platforms such as Twitter or Facebook could provide additional insights into how haze pollution is perceived and discussed in various regions.

However, it is important to note that haze pollution is highly variable in its occurrence, often influenced by seasonal and geographic factors. For instance, in China, haze is particularly prevalent during winter, a pattern that may not align with seasonal or spatial patterns in other countries. As a result, directly comparing or generalizing findings from Weibo to other platforms might not fully capture these regional differences.

Nevertheless, we recognize the potential benefits of expanding our research to include global social media platforms. Such an approach could offer a more comprehensive view of public discourse on environmental issues like haze pollution and enhance the generalizability of our findings.

Moreover, incorporating data from multiple platforms could be particularly valuable for studying global events with higher temporal and spatial consistency, such as the COVID-19 pandemic. This would allow for a more nuanced understanding of how global environmental and health crises are discussed and perceived across different regions and cultures.

We appreciate your suggestion and will certainly consider including a comparative analysis involving multiple platforms in our future research endeavors. This would help assess the broader applicability of our findings and provide a more complete perspective on public reactions to both local and global issues.

Thank you once again for your insightful comments.

Comments 6: Considering that only Chinese keywords were selected for analysis, it is advisable to include the analysis of multilingual microblogging data to broaden the applicability and influence of the research results.

Response 6: Thank you for your insightful feedback regarding the inclusion of multilingual content in our study. We appreciate your suggestion to broaden the analysis beyond Chinese-language posts.

Our research primarily focused on Chinese-language content due to its predominance on Weibo and its relevance to the majority of the discussions about haze pollution. This approach allowed us to delve deeply into the most widely shared and discussed views on this topic within the primary language of the platform.

However, we recognize that including posts in other languages could offer significant benefits. Analyzing multilingual data could capture subtle differences and perspectives that might be missed when focusing solely on Chinese-language content. This could reveal additional layers of sentiment and discussion, enriching our understanding of the public discourse surrounding haze pollution.

Furthermore, incorporating posts in different languages would enable us to study the influence of various cultural backgrounds on the discourse. Different cultural contexts may shape how individuals perceive and discuss environmental issues, providing a more nuanced view of global attitudes and responses. This broader perspective could enhance the generalizability and depth of our findings, offering a more comprehensive view of the issue at hand.

We are considering expanding our research to include multilingual data in future studies. This will involve developing methods for effectively identifying and analyzing non-Chinese posts while ensuring their relevance and accuracy. Such an approach will not only broaden the applicability of our research but also contribute to a more inclusive and culturally aware analysis.

Thank you once again for your valuable suggestion. We are committed to improving the scope and impact of our research, and your feedback will play a crucial role in guiding our future work.

Comments 7: Incorporating more recent studies on social media analysis in environmental science would strengthen the theoretical foundation of the research.

Response 7: Thank you for your suggestion to incorporate more recent studies on social media analysis in environmental science. We have reviewed relevant literature and provided a comprehensive overview in the revised manuscript. Please refer to lines 123-177 for our discussion of recent studies and their contributions to strengthening the theoretical foundation of our research.

Comments 8: Provide a more detailed description of the preprocessing steps, including any challenges encountered and how they were addressed, to enhance the transparency and reproducibility of the study.

Response 8: Thank you for your valuable feedback. We have expanded the description of the data preprocessing steps in the Methods section to enhance the transparency and reproducibility of the study. This includes a detailed account of the preprocessing procedures, the challenges encountered during this process, and the strategies we employed to address these challenges. These additions aim to provide a clearer understanding of our methodology and ensure that our approach can be effectively replicated. See lines 286-304.

Comments 9: Further exploration of the reasons behind the polarization of haze issues between heavily and lightly polluted areas is needed, potentially considering socio-economic factors.

Response 9: Thank you for your insightful comment. In response, we have expanded the discussion in the section analyzing the topic popularity related to haze on Weibo. We now include a detailed examination of the differences in haze pollution levels across research areas. This expanded analysis covers disparities from both natural and socio-economic perspectives, offering a more comprehensive exploration of the reasons behind the polarization of haze issues between heavily and lightly polluted areas. See lines 699-714.

Comments 10: Clarify why the study period was limited to 2013-2019 and whether the inclusion of more recent data could yield additional insights.

Response 10: Thank you for your insightful question. The study period of 2013-2019 was carefully selected to capture a critical window in China's air pollution crisis, marked by significant shifts in public awareness and government policy. This period includes the implementation of the Air Pollution Prevention and Control Action Plan in 2013, which played a pivotal role in shaping both public discourse and policy response to haze. During these years, public attention to air quality reached its peak, offering a robust dataset for topic modeling and evolution analysis. By focusing on this era, we ensured that our study captured the most dynamic phase of public engagement and governmental action, providing meaningful insights into the development of key themes and concerns.

We acknowledge, however, that the inclusion of more recent data (post-2019) could further enrich the analysis by revealing how discourse has evolved, especially in light of ongoing environmental regulations, technological advancements in air quality monitoring, and changes in societal priorities. As policies have continued to evolve, there may be shifts in public sentiment or new concerns emerging, such as the intersection of air pollution with broader environmental issues like climate change. While we focused on 2013-2019 to ensure a manageable and coherent study scope, future research could certainly build upon this work by extending the timeframe to explore these recent trends.

We appreciate the reviewer's insightful suggestion and will consider it for further research. Expanding the data to cover more recent years could provide valuable additional context for understanding long-term trends and how evolving policies have shaped public perceptions of air pollution.

Comments 11: Discuss how the findings could be applied to other environmental issues beyond haze, such as water pollution or climate change, to expand the scope of the research.

Response 11: Thank you for your insightful comments on this paper. I appreciate your suggestion to expand on how the research findings could be applied to other environmental issues beyond haze. Our response is as follows, and we have also expanded the conclusion section to briefly discuss the scalability of this study for addressing other environmental issues, see lines 866-872.

Building on the methods employed in this study, I believe the framework of community-based co-word network analysis and Sankey diagrams can be effectively extended to a range of other environmental concerns, such as water pollution, climate change, and air quality management. The approach used to analyze haze--related discussions on Weibo provides valuable insights into public perceptions and the evolution of topics over time. This flexibility in topic identification and trend visualization could be instrumental in understanding public discourse on a variety of environmental issues.

For instance, water pollution, particularly with regard to regional incidents like chemical spills or waste management crises, often triggers spikes in public concern and online dialogue. Applying the co-word network analysis in such cases could reveal how public sentiment shifts in response to government interventions, media coverage, or local advocacy. In a practical sense, this would allow policymakers to track the efficacy of their communication strategies and identify emerging issues before they escalate.

Similarly, climate change discussions are often multifaceted, spanning topics from policy debates to renewable energy, and even disaster preparedness. The Sankey diagram’s ability to depict the evolution of these intertwined discussions can provide clarity in tracking how public priorities shift, for example, from mitigation strategies to adaptation measures following extreme weather events. This visualization would be invaluable in creating more targeted public outreach or policy response frameworks.

Moreover, extending the methodology to broader air quality management issues, such as particulate matter (PM2.5) or industrial emissions, would help capture the more localized, real-time concerns that differ across regions. By identifying how specific environmental concerns evolve over time, policymakers and researchers could develop regionally tailored strategies for communication, intervention, and public education.

The patterns observed in haze-related discussions might also reveal commonalities in how environmental concerns are perceived across different contexts, contributing to a more holistic understanding of public environmental awareness. By applying this methodology to a broader set of topics, we can generate cross-comparative insights that help policymakers not only address current challenges but also anticipate future ones more effectively.

I hope this expansion offers a clearer view of the potential broader applicability of the methods employed in this study. Thank you once again for your valuable feedback.

Comments 12: Explain the process of data verification from Weibo, particularly how the authenticity and relevance of the analyzed posts were ensured.

Response 12: Thank you for your suggestion. Specifically, we implemented several key steps to ensure the authenticity and relevance of the analyzed posts from Weibo, and added relevant descriptions in the methods section, see lines 277-281.

When using Houyi Collector to collect Weibo data, we took systematic measures to ensure the authenticity and relevance of the collected data. First, in terms of keyword selection, we carefully screened based on the research topic and conducted advanced searches with keywords such as "haze" and "air" to ensure that the keywords used not only cover the core research issues, but also avoid overly broad or vague words. The accuracy of keywords is a key step to ensure that the collected data is highly relevant to the research objectives, and it can effectively filter out irrelevant or marginalized information.

Secondly, we control the timeliness of data collection by limiting the time range. In order to ensure that the collected data is relevant to the current research period, we set a specific time interval. This not only ensures the timeliness of the data, but also avoids collecting historical data or redundant information that is irrelevant to the current analysis. For example, we focus on collecting data in specific years in winter to ensure the pertinence of the data.

After data collection, we also carried out a series of data screening and filtering. First, we eliminated advertisements, robot-generated content or noise information through factors such as content review, user activity, and account authentication status. For suspicious low-quality data, we further conducted manual inspections to ensure that the Weibo content was authentic and reliable. In addition, we used data preprocessing and filtering mechanisms to delete duplicate, invalid, or content that deviated greatly from the topic. This multi-level filtering and screening measure ensured the overall quality of the data.

Comments 13: Address any limitations of the Louvain algorithm in detecting topic communities, especially when handling large-scale social media data.

Response 13: Thank you for highlighting this important consideration. The Louvain algorithm was chosen for its efficiency and scalability in handling large networks, such as those generated from social media data. However, we acknowledge several limitations of the Louvain algorithm when detecting topic communities, especially in the context of large-scale social media datasets.

First, while the algorithm is computationally efficient, it can experience performance degradation as the network size increases. In large networks, it may merge smaller communities into larger ones, potentially leading to less granular community detection and the loss of fine details in the topic clusters. To address this, we employed TF-IDF keyword extraction to ensure that the input data was refined, allowing the algorithm to detect more meaningful and distinguishable topics.

Another known limitation of the Louvain algorithm is the resolution limit, which makes it less effective at identifying smaller communities within large networks. This limitation can result in the merging of smaller but significant clusters, potentially obscuring niche topics. To mitigate this, we conducted manual reviews of the detected communities and, where necessary, applied additional analysis to refine or split broader communities into more specific topics.

Additionally, the Louvain algorithm relies on modularity optimization, which, while useful, may not always capture non-hierarchical structures in networks. Despite these limitations, we found that the algorithm provided a good balance between speed and accuracy for our study, achieving good results in detecting topic communities, given the scale of the dataset and the goals of the analysis. However, we recognize that alternative algorithms, such as Infomap or Label Propagation, could offer different perspectives, particularly in terms of detecting smaller communities or uncovering hierarchical structures.

In future research, we plan to explore these alternative community detection methods to improve the granularity and accuracy of topic identification, particularly when dealing with large-scale social media data.

Comments 14: Explore the potential inclusion of other temporal analysis methods, such as time-series analysis, to complement the cosine similarity approach.

Response 14: Thank you very much for your valuable suggestion. We sincerely appreciate your insightful feedback. While the current study focuses on the cosine similarity approach to analyze topic evolution over time, we fully recognize the potential advantages of integrating additional temporal analysis methods, such as time-series analysis, to provide a more comprehensive understanding of public attention and topic dynamics.

Time-series analysis could indeed offer richer insights by identifying underlying patterns in the data, such as trends, seasonal variations, or anomalies over different time periods. This method could reveal cyclical behaviors in public discussions or pinpoint specific time frames where significant shifts occur, which might be overlooked when relying solely on cosine similarity. Additionally, time-series analysis could be especially beneficial in detecting sudden changes in public sentiment or attention toward environmental issues, which may not manifest as gradual changes in topic similarity.

In fact, in another study we have conducted, we applied time-series analysis to examine the relationship between public attention, smog pollution levels, and emotional responses. Our findings demonstrated a high degree of temporal consistency and correlation among these three elements, particularly during periods of severe smog pollution. In that study, we also performed topic keyword analysis on Weibo texts during major smog events to explore the public's key concerns at such critical moments. These insights highlighted how public attention shifts in tandem with the severity of smog pollution and the emotional reactions it evokes. (For details, please refer to:Lu, H.; Rui, X.; Gemechu, F. G.; Li, R. Quantitative Evaluation of Psychological Tolerance Under the Haze: A Case Study of Typical Provinces and Cities in China with Severe Haze. Int. J. Environ. Res. Public Health. 2022, 19, 6574. https://doi.org/10.3390/ijerph19116574. )

Although time-series analysis was not implemented in the current paper, we believe that its potential for enhancing the temporal dimension of topic evolution analysis is clear. We plan to integrate this method in future work to complement cosine similarity and provide deeper insights into the temporal fine-grained fluctuations of public discourse.

Once again, thank you for your excellent suggestion, which will help shape the direction of our ongoing and future research efforts. We are confident that incorporating time-series analysis in future studies will further enrich our understanding of the temporal patterns in public attention and environmental discussions.

Comments 15: Discuss how data sparsity was managed in the text clustering method and whether this could impact the study's findings.

Response 15: Thank you for your insightful feedback. We acknowledge that data sparsity is a common challenge in text clustering, particularly when dealing with large and diverse datasets such as social media posts. In our study, we managed data sparsity through the following methods(Additionally, relevant descriptions have been added to the Material and methods section, addressing management measures for data sparsity, see lines 327-331):

In managing data sparsity within the text clustering method using the community-based co-word network approach, we employed several strategies to ensure robust topic identification. Specifically, we utilized TF-IDF for keyword extraction, which effectively prioritizes terms that hold greater significance within the context of the dataset while diminishing the influence of common, less informative words. This method allows us to focus on keywords that are more representative of the discussions surrounding haze on Weibo.

Moreover, we established a minimum frequency threshold for keyword inclusion, which serves to filter out low-frequency terms that could contribute to data sparsity. By setting this threshold, we aimed to retain only those keywords that exhibit sufficient relevance and frequency, thereby enhancing the quality of the co-occurrence relationships we analyzed. This approach not only mitigates the risk of overlooking important topics due to insufficient representation of certain keywords but also helps in creating a clearer and more coherent visualization of the relationships between significant terms.

However, it is crucial to acknowledge that this exclusion of infrequent terms may impact the overall findings. While we strive for a more focused analysis, the omission of certain low-frequency keywords may lead to an incomplete understanding of emerging topics or nuanced discussions within the Weibo data. These infrequent terms could represent specific, localized concerns or less mainstream perspectives that, if excluded, might skew the analysis of public sentiment and topic evolution.

Ultimately, by addressing data sparsity through these methods, we aimed to enhance the reliability and clarity of our clustering results. Still, we recognize the inherent limitations that may arise from this filtering process, and we encourage future research to explore these less common terms to gain a more comprehensive view of the environmental discourse on social media. This could involve integrating additional data sources or employing alternative methods that can better capture the full spectrum of public discussions related to haze and other environmental issues.

Comments 16: Ensure that all references are current and relevant, particularly in the rapidly evolving fields of environmental science and social media analysis.

Response 16: Thank you for your suggestion. We have updated several references to ensure that the literature is current and relevant. These updated references are highlighted in red within the text for clarity. However, we have retained some older references because they are foundational in establishing key concepts and methodologies that remain relevant to our study. These classic works provide essential context and continue to be widely cited in the field.

Comments 17: Moderate editing of English language required.

Response 17: Thank you very much for your valuable suggestion. We sincerely appreciate your careful review and insightful comments. In response, we have conducted a thorough and meticulous revision of the manuscript to improve not only the clarity and fluency of the English language but also to ensure that the overall structure and presentation of the content are more cohesive and polished.

We have carefully examined each section to refine the wording, enhance readability, and correct any ambiguities. Our goal was to make the manuscript more accessible to a broader audience while maintaining the accuracy and precision of the scientific content. Additionally, we have paid particular attention to aligning the tone and style throughout the manuscript to ensure a smooth and consistent reading experience.

We believe these revisions significantly improve the quality of the paper, and we are grateful for your constructive feedback, which has been instrumental in guiding these improvements. Thank you once again for your time and effort in reviewing our work.

 

We hope that these revisions adequately address the concerns raised by the reviewers and align with the expectations of the editorial team. We sincerely appreciate your time and effort in evaluating our work and look forward to the opportunity for further improvement.

 

Best regards,

Haiyue Lu et al.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

1. It is suggested that the author clearly state the research purpose and scientific questions of this article in the introduction.

2. The literature review only summarizes the progress of research methods and does not cover other aspects such as research content, research scale, research data, research perspective, etc.

3. The innovation of this article is insufficient, and the relevant methods have been widely applied. The author's contribution needs to be further condensed.

4. There are significant issues with the structure of the paper, and the introduction of the research area should not be placed in the results section. The results and discussion should be separated.

5. Why did the study area choose the six provinces and regions mentioned in the article instead of others, and what is their representativeness? Suggest explaining clearly.

6. Which Weibo data or data was selected in the article, and what should be detailed about its representativeness? In addition, the number of people using Weibo is still a minority, and its user base seems to be declining year by year, replaced by WeChat Moments and other platforms. Detailed analysis should be conducted on the rationality and representativeness of the data.

7. What insights does the research result provide for haze control or environmental protection? Suggest refining.

8. Is there a temporal and spatial consistency between the occurrence of significant haze pollution events and public attention? Suggested analysis.

9. The article actually lacks discussion, echoes the solution of scientific problems, compares with similar research, and points out the shortcomings of this study. Suggest adding.

10. What are the inspirations for similar research that should broaden the international perspective of this study?

Comments on the Quality of English Language

Moderate editing of English language is required.

Author Response

Responses to reviewer’s comments

We would like to express our sincere appreciation for your professional and insightful remarks on our manuscript. The feedback has been invaluable in enhancing the quality of our work. We have carefully considered each comment and have made significant revisions accordingly. We believe these amendments comprehensively address the concerns you have raised. In the submitted revised version, changes to our manuscript were all highlighted within the document by using red-colored text. The revisions and itemized responses to editors and reviewers are presented as follows. Responses to the comments are highlighted in blue for clarity. 

For the manuscript, we have conducted further optimizations. We have engaged professionals to polish the language of the manuscript, elevating the level of the writing. In the Introduction, we revised specific sections to better highlight the study's scientific questions, research objectives, and contributions. For the Literature review, we organized relevant studies from the perspectives of research content and perspective to strengthen the theoretical foundation. In the Material and methods, we refined the details, including data preprocessing steps, the advantages and limitations of the methods, and the accuracy of terminology. Additionally, we adjusted the structure of the paper for better clarity. In the Discussion section, we provided a deeper analysis of the results and their practical implications, improving the overall completeness of the manuscript. We also proposed several policy recommendations for environmental management based on the findings. Lastly, we thoroughly outlined the study’s Limitations, offering insights for future research and guidance for scholars interested in this topic. Point-by-point response to comments and suggestions for authors are as follows:

Reviewer(s) Comments:
Reviewer: 1

Comments 1: It is suggested that the author clearly state the research purpose and scientific questions of this article in the introduction.

Response 1: Thank you very much for your constructive feedback. In response to your suggestion, we have made revisions to the introduction to explicitly articulate the research purpose and the scientific questions addressed in this study. These changes aim to provide a clearer context and better highlight the objectives of our research. Please refer to lines 82-101 for the updated section. We appreciate your input, which has greatly contributed to improving the clarity and focus of our manuscript. 

Comments 2: The literature review only summarizes the progress of research methods and does not cover other aspects such as research content, research scale, research data, research perspective, etc.

Response 2: Thank you very much for your constructive feedback. We have carefully revised the literature review to address the points you raised. In addition to summarizing advancements in research methods, we have now incorporated a broader range of aspects, including research content, data, and perspectives. This expanded review provides a more comprehensive overview of the application of social media in environmental research. You can find these updates in lines 123-177 of the manuscript. We believe these revisions significantly enhance the context and relevance of our study and appreciate your valuable suggestions for improving the depth of our literature review.

Comments 3: The innovation of this article is insufficient, and the relevant methods have been widely applied. The author's contribution needs to be further condensed.

Response 3: Thank you for your valuable feedback regarding the innovation and contribution of our manuscript. In response to your comments, we have undertaken a comprehensive revision to better articulate the unique contributions and innovations of our study. We have carefully clarified the section outlining our research’s novelty and significance to emphasize its distinct contributions in the field. The revised section now provides a more focused and detailed discussion of how our study advances the existing knowledge and methods. Please refer to lines 103-117 of the updated manuscript for a thorough presentation of these refined points. We appreciate your constructive feedback, which has greatly assisted us in strengthening the overall impact of our research. 

Comments 4: There are significant issues with the structure of the paper, and the introduction of the research area should not be placed in the results section. The results and discussion should be separated.

Response 4: Thank you for your insightful feedback on the structure of our manuscript. We have carefully revised the manuscript to address the structural issues you highlighted. Specifically, we have relocated the introduction of the research area from the results section to the Material and methods section. Additionally, we have restructured the results and discussion sections to clearly separate them, ensuring that each section is distinctly focused and logically organized. These changes aim to enhance the clarity and coherence of the manuscript. We appreciate your guidance, which has been invaluable in improving the overall quality of our paper.

Comments 5: Why did the study area choose the six provinces and regions mentioned in the article instead of others, and what is their representativeness? Suggest explaining clearly.

Response 5: Thank you for your thoughtful question regarding the selection of the study areas, at the same time, we have included relevant descriptions in the Material and methods section, briefly explaining the reasons for selecting the study area, as indicated in lines 257-260.

First, the seven regions (Beijing, Shandong, Liaoning, Hebei, Shanxi, Tianjin, and Inner Mongolia) were selected based on their importance and representativeness in northern China. These regions have played a key role in economic development and industrialization, and have historically been severely affected by haze. In particular, as the national capital, Beijing has a leading role in environmental policies and public attention for the whole country, while provinces such as Hebei and Shanxi face more severe air pollution problems due to heavy industry and coal mining. Although Shandong is located in East China, its economic activities, industrial structure, and environmental connections with North China make it equally important in smog governance research. Inner Mongolia occupies an important position in energy production and resource development, and is also closely related to environmental issues in North China.

Second, these seven regions are similar in climate and geographical environment, and are all affected by similar climatic conditions and pollution sources. This geographical consistency makes them comparable in terms of haze formation mechanisms and governance strategies, facilitating systematic comparative studies.

Third, public attention and public opinion on smog issues are particularly evident in these seven regions. Through topic identification and evolution analysis on social media (such as Weibo), we can better capture the changes in public sentiment and its response to policies, thereby providing data support for the optimization of governance measures.

Finally, the selection of these seven regions also helps to provide an empirical basis for future policy recommendations. By analyzing their common characteristics and differences, effective governance strategies can be identified, which in turn can provide reference for haze governance in other regions.

In summary, these seven provinces and regions were selected as research objects not only because of their importance in the northern region, similar climatic conditions and significant public attention, but also because of their representativeness, which makes the research results have wide applicability and reference value. Such a selection provides a solid foundation for this study and helps to reveal key issues and effective strategies in haze governance. 

Future research will aim to expand the scope of the study to encompass the entire country or even multiple countries. This approach will help verify the universality of the research methods and results, as well as provide new insights for global environmental problem management. We appreciate your interest in the rationale behind our study area's selection.

Comments 6: Which Weibo data or data was selected in the article, and what should be detailed about its representativeness? In addition, the number of people using Weibo is still a minority, and its user base seems to be declining year by year, replaced by WeChat Moments and other platforms. Detailed analysis should be conducted on the rationality and representativeness of the data.

Response 6: Thank you for raising important questions regarding the representativeness and selection of data in our study.

In the paper, we describe the data used in detail, including air quality data and Weibo data:see line 271-285. Among them, air quality data can reflect the air quality status of different cities in a specific time period, and has high authenticity and representativeness; when collecting Weibo data, we collect it with as comprehensive keywords as possible, with high temporal resolution and large quantity, and also with high representativeness.

In addition, we recognize that although the user base of Weibo has declined in recent years, Weibo can still effectively reflect public sentiment and discussion during specific events. Weibo's information dissemination method is more open, and users can post content freely without being restricted by friend relationships, which allows information to spread rapidly and is suitable for discussions of hot events. In contrast, the content dissemination of WeChat Moments is mainly limited to friends, and the scope of dissemination is relatively closed.

In addition, Weibo supports a variety of content formats, such as text, pictures, videos and links, with a word limit of 240 words, which is suitable for publishing long discussions or real-time news, while the content format of WeChat Moments is mainly short text and pictures, focusing more on sharing personal life. In terms of interaction, Weibo promotes extensive discussion through forwarding, commenting and liking, and the forwarding function allows information to spread quickly, while the interaction in WeChat Moments is mainly through likes and comments, and the forwarding function is relatively limited.

Although the number of Weibo users has declined, it still has a stable user base, especially in discussions among young people and specific topics, and its activity is still very high. In summary, the combination of Weibo and air quality data provides a comprehensive perspective for studying air pollution issues, so we believe that the data used is reasonable and representative, and will further clarify this in the paper to enhance the persuasiveness and representativeness of the study.

In summary, while Weibo may not represent the entire social media landscape, it provides valuable insights into public discourse on environmental issues in northern China. We appreciate your feedback and will consider expanding our analysis to include data from other platforms in future research to enhance the representativeness and generalizability of our findings.

Comments 7: What insights does the research result provide for haze control or environmental protection? Suggest refining.

Response 7: Thank you very much for your insightful suggestion. We have carefully refined and expanded the section addressing the practical insights our research offers for haze control and environmental protection. This can now be found in the conclusion section, specifically in lines 873-888.

In this section, we emphasize how the findings from our social media analysis can inform targeted policy-making, raise public awareness, and enhance the implementation of environmental measures in key regions affected by haze. Additionally, we highlighted how these insights could be applied to other environmental issues, offering broader value beyond the scope of haze control. These insights are crucial not only for improving the effectiveness of haze control efforts in northern China but also for informing broader environmental protection initiatives.

We deeply appreciate your valuable feedback, which has helped us enhance the relevance and clarity of our study’s contributions.

Comments 8: Is there a temporal and spatial consistency between the occurrence of significant haze pollution events and public attention? Suggested analysis.

Response 8: Thank you very much for your valuable feedback and insightful suggestion. We sincerely appreciate your thorough review, which has provided us with the opportunity to further enhance our study.

To address the suggestion of analyzing the temporal and spatial consistency between significant haze pollution events and public attention, we would like to provide additional context. In another study, we explored the dynamic relationship between public attention, haze pollution, and public sentiment using time-series analysis. Our findings demonstrated that, over time, there is a high level of consistency and correlation among these three factors, particularly during periods of severe pollution. This analysis not only highlighted the temporal alignment but also provided insights into the emotional reactions of the public in response to worsening environmental conditions.

Furthermore, during severe haze pollution events, we conducted topic keyword analysis on Weibo posts to explore the focus of public discussions. The results showed that during these critical moments, the public's attention tends to shift towards specific issues such as health impacts, governmental response, and pollution control measures. This analysis offered a deeper understanding of the public's concerns during significant haze pollution events, reinforcing the relevance of spatio-temporal consistency between pollution and public attention.

However, it is important to note that our preliminary findings were primarily based on data from Beijing, with public attention being represented by the volume of Weibo posts. This limitation suggests that our analysis may not fully capture the broader regional or national response to haze pollution events. In future research, we plan to apply more precise methodologies to other regions to assess whether this spatio-temporal consistency holds true on a larger scale. By incorporating more accurate measurements of public attention, such as analyzing the popularity of specific topics over time, we aim to validate the spatial and temporal alignment between significant haze events and public concern in other affected areas.

This enhanced approach will allow us to better understand how public attention evolves in response to environmental crises and provide more robust evidence of the relationship between pollution events and social response.

Once again, we sincerely thank you for your invaluable suggestion. We are committed to incorporating these considerations into future research, and we believe that this will further strengthen the comprehensiveness and rigor of our study.

Comments 9: The article actually lacks discussion, echoes the solution of scientific problems, compares with similar research, and points out the shortcomings of this study. Suggest adding.

Response 9: Thank you for your valuable feedback. In response to your suggestion, we have carefully expanded the discussion section of the article. See lines 684-815. This now includes a more detailed analysis of the structural components, directly addressing the resolution of the scientific questions posed in the introduction. Additionally, we highlighted the unique contributions and theoretical references of our research compared to similar studies.

Furthermore, we have thoroughly revised and extended the limitations section. Here, we acknowledge various aspects where the study may have limitations, including potential biases in the data sources, constraints of the chosen methodology, and the need for broader generalization across different geographic and temporal contexts. By doing so, we aim to provide a balanced perspective that reflects both the strengths and areas for future improvement.

We believe these revisions offer a more comprehensive view of the study's outcomes and limitations, and we sincerely appreciate your thoughtful comments that have helped us enhance the overall rigor of the manuscript.

Comments 10: What are the inspirations for similar research that should broaden the international perspective of this study?

Response 10: Thank you very much for your insightful suggestion. We fully recognize the importance of broadening the international perspective in this study, and we have reflected on how our research can inspire similar studies in other regions or contexts. While our current focus is on China, the methods and findings could be applied more broadly to analyze public attention and environmental issues in different countries.

The inspirations for similar studies are mainly reflected in several aspects. First, cross-national comparative studies can reveal the successful experiences and challenges of different countries in haze governance. By analyzing the effective practices of different countries in air quality management and public participation, it can provide useful references for the governance strategies of the country. In addition, international cooperation is an important way to deal with global environmental problems. Participating in international environmental protection organizations and conferences can promote knowledge sharing and technical exchanges, and help better deal with environmental problems such as haze. Third, using an international perspective to analyze the model and mechanism of public participation can understand the public's response and behavior to environmental issues under different cultural backgrounds, thereby guiding the country's efforts to enhance public environmental awareness and participation. Finally, focusing on the relationship between climate change and air pollution provides a more comprehensive perspective for haze governance. While responding to climate change, many countries are also actively improving air quality. By learning the comprehensive governance models of these countries, it is possible to promote the synergy between haze governance and climate change response and achieve sustainable development.

We truly appreciate your thoughtful feedback, and we believe that these considerations will help to further enrich the study in future iterations.

Comments 11: Moderate editing of English language is required.

Response 11: Thank you very much for your valuable suggestion. We sincerely appreciate your careful review and insightful comments. In response, we have conducted a thorough and meticulous revision of the manuscript to improve not only the clarity and fluency of the English language but also to ensure that the overall structure and presentation of the content are more cohesive and polished.

We have carefully examined each section to refine the wording, enhance readability, and correct any ambiguities. Our goal was to make the manuscript more accessible to a broader audience while maintaining the accuracy and precision of the scientific content. Additionally, we have paid particular attention to aligning the tone and style throughout the manuscript to ensure a smooth and consistent reading experience.

We believe these revisions significantly improve the quality of the paper, and we are grateful for your constructive feedback, which has been instrumental in guiding these improvements. Thank you once again for your time and effort in reviewing our work.

 

We hope that these revisions adequately address the concerns raised by the reviewers and align with the expectations of the editorial team. We sincerely appreciate your time and effort in evaluating our work and look forward to the opportunity for further improvement.

Best regards,

Haiyue Lu et al.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors list BerTopic and top2vec together with LSTM, unfortunately the LSTM is not used for topic modelling.

At line 207 the authors state that "their drawbacks include the need for substantial labeled data, high computational resources, and the "black box" nature of the models, making the results harder to interpret." Embedding based topic modelling are unsupervised methods and they do not need labelled data, both method can be run on low end GPUs, moreover the models give the possibility to trace back the part of the documents belonging to a given topic, improving the interpretability of the results. Please revise

.

Author Response

Responses to reviewer’s comments

We would like to express our sincere appreciation for your professional and insightful remarks on our manuscript. The feedback has been invaluable in enhancing the quality of our work. We have carefully considered each comment and have made significant revisions accordingly. We believe these amendments comprehensively address the concerns you have raised. In the submitted revised version, changes to our manuscript were all highlighted within the document by using blue-colored text. The revisions and itemized responses to editors and reviewers are presented as follows. Responses to the comments are highlighted in red for clarity. 

For the manuscript, we further optimized the English writing. In the Introduction, we revised the description of the article's innovation to better highlight the scientific questions, research objectives, and contributions of the study. For the Literature review, we revised some of the statements about the characteristics of the methods to improve the accuracy of the content. In the Discussions and Conclusions sections, we appropriately optimized and deleted the content to improve the refinement and clarity of the article. The responses to the authors' comments and suggestions are as follows:

Reviewer(s) Comments:
Reviewer: 1

 

Comments 1: The author need to employ state of the art topic modeling techniques such as BERTopic and/or Top2Vec for topic detection instead of the current methods employed. Given that these transformer-based embeddings methods handle social media content in a more effective way since they can identify semantic relationships between words and phrases from contextual understanding, and can also give significant advantages for analyzing topic evolution.

 

Response 1: 

 

Thank you for your detailed review and valuable feedback on my manuscript. I sincerely apologize for the oversight in my original submission regarding the description of embedding-based topic modeling methods at line 207. You are absolutely correct in pointing out that these methods, such as Top2Vec and BERTopic, are unsupervised and do not require substantial labeled data. Moreover, they can indeed run on lower-end GPUs and offer the possibility of tracing back parts of documents to specific topics, which enhances the interpretability of the results.

 

I have carefully revised the manuscript to correct these inaccuracies, as shown in lines 201-204. I appreciate your insight in highlighting these important aspects, and the manuscript is now more accurate and precise thanks to your suggestions.

 

If there are any further points that need clarification or additional revisions, I am more than happy to address them.

 

Thank you again for your valuable comments and for improving the quality of my work. 

 

 

We trust that these revisions sufficiently address the reviewers' concerns and meet the editorial team's expectations. We are truly grateful for your time and consideration in reviewing our work, and we look forward to any further opportunities to refine and enhance the manuscript.

 

Best regards,

 

Haiyue Lu et al

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

This revision has addressed the problems in previous version. The authors have revised the paper according to my comments. The innovation of this work can be further explained in the introduction and method section.

Comments on the Quality of English Language

 Moderate editing of English language required.

Author Response

Responses to reviewer’s comments

We would like to express our sincere appreciation for your professional and insightful remarks on our manuscript. The feedback has been invaluable in enhancing the quality of our work. We have carefully considered each comment and have made significant revisions accordingly. We believe these amendments comprehensively address the concerns you have raised. In the submitted revised version, changes to our manuscript were all highlighted within the document by using blue-colored text. The revisions and itemized responses to editors and reviewers are presented as follows. Responses to the comments are highlighted in red for clarity. 

For the manuscript, we further optimized the English writing. In the Introduction, we revised the description of the article's innovation to better highlight the scientific questions, research objectives, and contributions of the study. For the Literature review, we revised some of the statements about the characteristics of the methods to improve the accuracy of the content. In the Discussions and Conclusions sections, we appropriately optimized and deleted the content to improve the refinement and clarity of the article. The responses to the authors' comments and suggestions are as follows:

Reviewer(s) Comments:
Reviewer: 1

 

Comments 1: This revision has addressed the problems in previous version. The authors have revised the paper according to my comments. The innovation of this work can be further explained in the introduction and method section.

 

Response 1: 

 

Thank you very much for your insightful and constructive feedback. We sincerely appreciate the time and effort you put into reviewing our paper. Your comments have been invaluable in improving the quality and clarity of our work.

 

In response to your suggestions, we have made the following revisions:

 

Innovation Points: We have carefully reviewed the three contributions outlined in the original version. After consideration, we consider that the third contribution, which pertains to the environmental management measures proposed based on our findings, does not strictly represent an innovative aspect of the research. Therefore, we have removed this point from the list of contributions. Instead, we have focused on the first two contributions as the key innovations of this study(lines 103-113 in the revised manuscript).

 

Method Section: To further emphasize the innovative aspects of our methodology, we have expanded the explanation of our approach in the method section (lines 236-253 in the revised manuscript). We now provide a more detailed account of the co-word network method and the introduction of the topic popularity index, highlighting how these methods contribute to a deeper understanding of public opinion dynamics. Additionally, we explain the use of cosine similarity to measure the temporal evolution of topics, which further supports the novelty of the analysis.

 

We are confident that these revisions address your concerns, and we have ensured that the innovative aspects of our work are now more clearly articulated.

 

Once again, we are grateful for your thoughtful comments and the opportunity to refine our paper. 

 

Comments 2: Moderate editing of English language required.

 

Response 2: 

 

Thank you for your valuable feedback, particularly regarding the need for moderate editing of the English language. We have carefully reviewed the entire manuscript and made a series of revisions to address issues related to grammar, sentence structure, and overall clarity. These edits were aimed at improving the readability and flow of the text, ensuring that our ideas are communicated more effectively and precisely.

 

Specifically, we focused on simplifying complex sentences, refining awkward or unclear phrasing, and enhancing the logical coherence between sections. In addition, we have carefully reviewed the use of technical terminology to ensure consistency and accuracy throughout the manuscript. We believe that these revisions have significantly improved the language quality of the paper.

 

We sincerely appreciate your time and effort in reviewing our manuscript and providing this constructive feedback. We are confident that the improvements made align with your expectations, and we look forward to any further suggestions or comments that could help us enhance the quality of our work.

 

Thank you once again for your thorough evaluation and guidance.

 

We trust that these revisions sufficiently address the reviewers' concerns and meet the editorial team's expectations. We are truly grateful for your time and consideration in reviewing our work, and we look forward to any further opportunities to refine and enhance the manuscript.

 

Best regards,

Haiyue Lu et al.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

1. All my concerns have been addressed.

2. The conclusion is lengthy.

Comments on the Quality of English Language

 Minor editing of English language is required.

Author Response

Responses to reviewer’s comments

We would like to express our sincere appreciation for your professional and insightful remarks on our manuscript. The feedback has been invaluable in enhancing the quality of our work. We have carefully considered each comment and have made significant revisions accordingly. We believe these amendments comprehensively address the concerns you have raised. In the submitted revised version, changes to our manuscript were all highlighted within the document by using blue-colored text. The revisions and itemized responses to editors and reviewers are presented as follows. Responses to the comments are highlighted in red for clarity. 

For the manuscript, we further optimized the English writing. In the Introduction, we revised the description of the article's innovation to better highlight the scientific questions, research objectives, and contributions of the study. For the Literature review, we revised some of the statements about the characteristics of the methods to improve the accuracy of the content. In the Discussions and Conclusions sections, we appropriately optimized and deleted the content to improve the refinement and clarity of the article. The responses to the authors' comments and suggestions are as follows:

Reviewer(s) Comments:
Reviewer: 1

 

Comments 1: The conclusion is lengthy.

 

Response 1: 

 

Thank you very much for your insightful feedback. We appreciate your careful review and constructive suggestions. In response to your comment regarding the length of the Conclusions, we have thoroughly revised this section to improve its clarity and conciseness. In addition, appropriate simplifications have been made to the Discussions and Limitations and Future work sections to improve the overall clarity of the article.

 

Specifically, we have:

 

Removed redundant information that repeated key points already discussed in earlier sections of the manuscript, ensuring that the conclusion provides a concise summary rather than reiterating detailed analysis.

 

Streamlined the discussion of key findings to focus on the most important insights and their implications, without introducing overly detailed explanations.

 

Shortened the section by eliminating any extraneous content, while still ensuring that the main contributions of our research are clearly articulated.

We believe these revisions have enhanced the overall structure and readability of the conclusion, making it more focused and impactful. Once again, we sincerely appreciate your helpful comments and believe that the revised version better aligns with your expectations.

 

Thank you for your time and effort in reviewing our manuscript. 

 

Comments 2: Minor editing of English language is required.

 

Response 2: 

 

Thank you for your valuable feedback, particularly regarding the need for moderate editing of the English language. We have carefully reviewed the entire manuscript and made a series of revisions to address issues related to grammar, sentence structure, and overall clarity. These edits were aimed at improving the readability and flow of the text, ensuring that our ideas are communicated more effectively and precisely.

 

Specifically, we focused on simplifying complex sentences, refining awkward or unclear phrasing, and enhancing the logical coherence between sections. In addition, we have carefully reviewed the use of technical terminology to ensure consistency and accuracy throughout the manuscript. We believe that these revisions have significantly improved the language quality of the paper.

 

We sincerely appreciate your time and effort in reviewing our manuscript and providing this constructive feedback. We are confident that the improvements made align with your expectations, and we look forward to any further suggestions or comments that could help us enhance the quality of our work.

 

Thank you once again for your thorough evaluation and guidance.

 

We trust that these revisions sufficiently address the reviewers' concerns and meet the editorial team's expectations. We are truly grateful for your time and consideration in reviewing our work, and we look forward to any further opportunities to refine and enhance the manuscript.

 

Best regards,

 

Haiyue Lu et al.

Author Response File: Author Response.docx