Using Machine Learning and Routing Protocols for Optimizing Distributed SPARQL Queries in Collaboration
Round 1
Reviewer 1 Report
The manuscript under review presents research findings on applying contemporary machine-learning methods and routing protocols to optimize data processing in the Internet of Things (IoT). The subject matter is both relevant and timely. The manuscript's structure is coherent, with the introduction highlighting the significance of the problem addressed and the primary contributions of the authors' research. The core content of the manuscript is presented accurately. I have no concerns or questions about the main body of the document. The authors have effectively illustrated both the theoretical and practical aspects. However, I do have minor remarks regarding the formatting of the manuscript, which I list below:
1. The abstract requires revision. It should more accurately convey the manuscript's content, highlighting the relevance and significance of the problem addressed, the authors' contributions to resolving this issue, the results achieved, and their importance.
2. Abbreviations, such as SIMORA, DBMS, etc., should be defined the first time they appear in the text. Kindly make this amendment.
Overall, I think that this manuscript is suitable for acceptance following minor revisions.
Author Response
1. The abstract requires revision. It should more accurately convey the manuscript's content, highlighting the relevance and significance of the problem addressed, the authors' contributions to resolving this issue, the results achieved, and their importance.
- We have enhanced the abstract.
2. Abbreviations, such as SIMORA, DBMS, etc., should be defined the first time they appear in the text. Kindly make this amendment.
- We provide the long form of each abbreviation next to their first use.
Reviewer 2 Report
The authors used machine learning methods for join order optimization in SPARQL queries, focusing on many joins. The proposed approach for reducing network traffic for IoT applications includes an operator placement strategy and a machine learning approach for join order optimization.
The paper is well-structured. The authors extensively reviewed the related research works. On that basis, they identified and formulated a research problem and proposed a new approach. The proposed method is well described in the paper. The effectiveness of the proposed approach was verified experimentally with the use of SIMORA simulator (which simulates DBMS in a randomized topology) and selected scenarios. To sum up, the paper presents an interesting approach, which can be applied in IoT systems.
Please address the following issues:
1. Please analyze the weak points of the proposed method.
2. Please enlarge the last section (5) and include future work plans.
3. Do the results come from many experiments? If yes, then they should undergo the statistical analysis.
Author Response
1. Please analyze the weak points of the proposed method.
- The first weak point (We need to keep the training phase short. Otherwise, the changes in the dataset will immediately render the newly trained model useless.) was already present in the text.
- We have added a new section about the weak points after the definition of our algorithm.
2. Please enlarge the last section (5) and include future work plans.
- We have extended that paragraph.
3. Do the results come from many experiments? If yes, then they should undergo the statistical analysis.
- the intermediate results and the network traffic are (due to the use of a simulator) exact values, such that the experiments do not need repetitions.
Reviewer 3 Report
Following are my comments for the article.
Clarification: It would be helpful to provide a brief explanation of what SPARQL queries are for readers who may not be familiar with the term.
Conciseness: Try to condense the abstract by removing redundant phrases. For example, "Choosing a good join order is crucial when evaluating SPARQL queries, as it significantly affects the number of intermediate results" could be simplified to "Optimal join order selection is crucial for SPARQL query evaluation."
Structure: Consider organizing the abstract into distinct sections for clarity. For instance, you could have sections for the problem statement, the NP-hard nature of the problem, the role of machine learning, and the importance of network topology.
Specifics: Provide a bit more detail on the "new techniques for collaboration between routing, application, and machine learning." What are the key innovations or contributions of the research?
Clarity: Ensure that the abstract maintains a clear flow of ideas, making it easy for readers to understand the research problem, its significance, and the proposed solutions.
Avoid jargon: While technical terms are essential, try to minimize jargon and provide explanations or context where necessary to make the abstract accessible to a broader audience.
Acronyms: It would be better to define all the acronyms.
Illustrations: Figure quality can be improved and some figures must be redesigned with more information.
Results: The results must be clearly explained.
Comparison: It is necessary to compare proposed algorithm with the methods recently published in 2021, 2022 and 2023.
Conclusion: Conclusion must be coherent with abstract and must conclude work.
Minor English editing is required.
Author Response
Clarification: It would be helpful to provide a brief explanation of what SPARQL queries are for readers who may not be familiar with the term.
- We added (SPARQL is particularly well suited to the IoT environment because it can handle various data structures.) to the abstract.
- Additionally ( …. the simple concept of triples in the resource description framework (RDF) leads to a higher number of joins in SPARQL queries.) in the motivation shows the main differences to SQL, namely the storage format(triples) and the consequence of many joins in the queries.
Conciseness: Try to condense the abstract by removing redundant phrases. For example, "Choosing a good join order is crucial when evaluating SPARQL queries, as it significantly affects the number of intermediate results" could be simplified to "Optimal join order selection is crucial for SPARQL query evaluation."
- The article is about reducing the network traffic by reducing the intermediate results, which could be done by using another (better) join order. Therefore we believe that this has to be mentioned in the abstract.
Structure: Consider organizing the abstract into distinct sections for clarity. For instance, you could have sections for the problem statement, the NP-hard nature of the problem, the role of machine learning, and the importance of network topology.
- The template from MDPI explicitly denies the use of paragraphs “(Do not insert blank lines, i.e. \\)“
Specifics: Provide a bit more detail on the "new techniques for collaboration between routing, application, and machine learning." What are the key innovations or contributions of the research?
- The entire collaboration between routing and application, and the collaboration between routing and machine learning is completely new, as current hardware completely prevents us from using such information.
Clarity: Ensure that the abstract maintains a clear flow of ideas, making it easy for readers to understand the research problem, its significance, and the proposed solutions.
- We have enhanced the abstract.
Avoid jargon: While technical terms are essential, try to minimize jargon and provide explanations or context where necessary to make the abstract accessible to a broader audience.
- We have enhanced the abstract.
Acronyms: It would be better to define all the acronyms.
- we provide the long form of each acronym near the first use of them.
Illustrations: Figure quality can be improved and some figures must be redesigned with more information.
- All the figures are explained in the text. Which information is missing in the figures?
Results: The results must be clearly explained.
- There is an entire evaluation chapter. Which part is not explained?
Comparison: It is necessary to compare proposed algorithm with the methods recently published in 2021, 2022 and 2023.
- There is no work, which includes the routing protocols directly in their algorithms.
- Additionally in the related work, we are showing, that approaches from RDBMs (SQL) can not be used for SPARQL(because they require too much memory), so they can not be compared.
Conclusion: Conclusion must be coherent with abstract and must conclude work.
- We have enhanced the conclusion and added future work plans to that section.
Round 2
Reviewer 3 Report
Dear Authors,
Please address my previous comments. Only abstract has been changed, however previous comments were related to whole article, not only abstract. Thanks.
Moderate English language editing is required.
Author Response
Dear Reviewer,
We have revised the entire article again regarding your previous comments and language.
Best Regards
Round 3
Reviewer 3 Report
Thanks for addressing my previous concerns. It would be better to improve figure 2.
Minor English editing is required.
Author Response
We have improved Figure 2 and the English language throughout the article.
