Symbiosis in Health: The Powerful Alliance of AI and Propensity Score Matching in Real World Medical Data Analysis

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This research paper systematically explores the emerging symbiotic relationship between AI and PSM in real-world medical data analysis using Synthetic Thematic Analysis and bibliometric techniques on 433 Scopus-indexed publications. The study identifies four dominant thematic clusters including prediction, cancer management, diagnosing, and deep learning. It also highlights two principal modes of integration: AI enhancing PSM and PSM strengthening AI-based analyses. The authors conclude that the convergence of AI and PSM is a rapidly growing global research trend that improves causal inference and methodological rigor in observational medical studies. Overall, this paper is well-written however, I have a few concerns regarding the contents presented in the paper that are listed below:

1. The study relies heavily on bibliometric and thematic mapping, offering limited methodological or technical analysis of how AI and PSM are actually implemented in practice.
2. STA is presented as a key methodological contribution, but its reproducibility, sensitivity, and comparative advantage over established review frameworks (e.g., PRISMA, scoping reviews) are not rigorously justified.
3. The exclusive reliance on Scopus may introduce indexing bias and omit relevant studies indexed in PubMed, Web of Science, IEEE Xplore, or arXiv.
4. Heavy dependence on author keywords risks misclassification, since many studies use generic terms such as “machine learning” without specifying actual algorithms.
5. AI methods are frequently grouped under broad labels (AI, ML, DL), limiting interpretability and weakening conclusions about algorithmic trends. For reference you can read the paper to https://doi.org/10.3389/frai.2023.1202990.
6. There is no information about how the authors finalized an article or group of articles if a conflict raise among authors. How they did this?
7. The review synthesizes publications descriptively but does not assess study quality, bias risk, or validity of causal claims.
8. Growth trends and country productivity are interpreted positively without discussing publication inflation, citation bias, or research redundancy.
9. The four identified thematic clusters show substantial conceptual overlap, raising questions about the discriminative power of the clustering approach.
10. Known limitations of combining AI with PSM (e.g., unmeasured confounding, model-induced bias, overfitting in propensity estimation) are not sufficiently critiqued. Also,
    1. Despite extensive analysis, the manuscript provides limited concrete guidance for clinicians or researchers on when and how to combine AI and PSM effectively.
    2. The review risks overstating the extent to which AI-enhanced PSM can approximate randomized controlled trials without discussing underlying assumptions.
11. Limitation section is missing in the paper

Comments on the Quality of English Language

1. Several sections reiterate similar points about “symbiosis” and “methodological rigor,” reducing conciseness and readability. 
2. The paper needs a thorough proofread to fix these issue and fix grammatical and other syntax errors.

Author Response

STA is presented as a key methodological contribution, but its reproducibility, sensitivity, and comparative advantage over established review frameworks (e.g., PRISMA, scoping reviews) are not rigorously justified.

 

The aim and advantages of STA were added to the Introduction

 

The exclusive reliance on Scopus may introduce indexing bias and omit relevant studies indexed in PubMed, Web of Science, IEEE Xplore, or arXiv.

 

The analysis was limited to the reviewed research literature. Scopus includes PubMed and almost all of the journals covered by WoS.- The exclusion of possible publications was discussed in the study limitations

 

 

Heavy dependence on author keywords risks misclassification, since many studies use generic terms such as “machine learning” without specifying actual algorithms.

The keyword Machine learning was part of the search string

AI methods are frequently grouped under broad labels (AI, ML, DL), limiting interpretability and weakening conclusions about algorithmic trends. For reference you can read the paper to https://doi.org/10.3389/frai.2023.1202990.

 

The acronyms AI, ML, Dl can not be used  because  they might be part of nonrelevant search  terms or might stand for nonrelevant keywords. For example ML might stand for Maximum Likelihood, Millilitre, Money Laundering etc. However we added the above point to limitation

 

There is no information about how the authors finalized an article or group of articles if a conflict raise among authors. How they did this?

 

All papers from the corpus  were included in the analysis, there were no conflicts regarding the selection of the papers included in the thematic analysis

 

The review synthesizes publications descriptively but does not assess study quality, bias risk, or validity of causal claims.

 

The paper is not a review paper. The aim of the study presented was to thematically analyse all publications concerning  PSM and AI indexed in Scopus

 

Growth trends and country productivity are interpreted positively without discussing publication inflation, citation bias, or research redundancy.

 

The study is focused on thematic analysis, thus only fundamental descriptive bibliometric data are presented

 

The four identified thematic clusters show substantial conceptual overlap, raising questions about the discriminative power of the clustering approach.

 

Use of AI in PSM applications in medicine is conceptually relatively small, consequently the clusters in the bibliometric map are thematically related. The bibliometric mapping/clustering was done with one of the most recognised  software tools namely VOSViewer which was successfully used in the-.thousands of studies

 

Known limitations of combining AI with PSM (e.g., unmeasured confounding, model-induced bias, overfitting in propensity estimation) are not sufficiently critiqued.

Added to limitations

Despite extensive analysis, the manuscript provides limited concrete guidance for clinicians or researchers on when and how to combine AI and PSM effectively.

 

Added to the discussion

 

The review risks overstating the extent to which AI-enhanced PSM can approximate randomized controlled trials without discussing underlying assumptions.

 

Limitation section is missing in the paper

 

The limitation are presented at the end of discussion section

Several sections reiterate similar points about “symbiosis” and “methodological rigor,” reducing conciseness and readability. 

 

Corrected

 

The paper needs a thorough proofread to fix these issue and fix grammatical and other syntax errors.

 

The MDPI services were used to proofread the paper

Reviewer 2 Report

Comments and Suggestions for Authors

The paper is a biblimetric paper with no interest for the applied sciences journal

Author Response

The paper is a biblimetric paper with no interest for the applied sciences journal

 

The paper is not a bibliometric paper, the induced bibliometric map serves as a strategic roadmap. But even pure descriptive bibliometrics enebles institutions and researchers to benchmark performance, identify emerging high-impact trends, and optimize funding or publication choices- Ultimately, it transforms raw publication data into actionable insights that guide the identification of untapped research gaps.

 

However as mentioned above we used Synthetic Thematic Analysis which  transformed fragmented research into applicative actionable models by integrating cross-disciplinary publications used to solve complex health problems, STA bridges the gap between theory and practice by creating causal frameworks and "master landscapes" that might  guide policymakers and R&D teams in making data-driven strategic decisions. Distilling hundreds of  papers, STA enabled to identify research gaps, trends, future research possibilities and maturity levels of different PSM and AI technologies with much fewer resources.

Reviewer 3 Report

Comments and Suggestions for Authors

In the present study, analysis of the use of Artificial Intelligence (AI) and Propensity Score Matching (PSM) is described. The study uses Synthetic Thematic Analysis (STA), derived from synthetic knowledge synthesis, to systematically review the existing literature on AI and PSM in medicine.  Publications were harvested from the Scopus database using a comprehensive search string limited to the Medical subject area. A corpus of 433 documents was analyzed using bibliometric tools (Bibliometrix and VOSViewer) to map the research landscape, identify thematic clusters based on author keywords, analyze collaboration patterns, and synthesize findings from highly prolific publications. STA identified four main thematic clusters: Prediction, Cancer Management, Diagnosing, and Deep Learning. In general, the paper is well written in English and it presents an informative state of the art, which is pertinent to introduce the problem.

My main concerns with the present research are the following:

a) Please include keywords.
b) Quality of Figure 1 must be improved.
c) Please combine Figure 2 and Figure 3 by using a bar plot representation.
d) Please improve quality of Figure 5. The journals are not completely visible.
e) Please add a paragraph about the main Editorial publishing papers in the topic.
f) According to lines 212-220. Please add a discussion about the use of Machine Learning techniques, software/version, topic of application. Similar to Machine Learning, add discussion involving deep learning techniques. In Table 5, a deductive thematic analysis is presented, but, to consider a proper analysis more specific information is required.
g)  A more detailed discussion about advantages and disadvantages of integration and usage of AI in the domain of propensity score matching is required.
h) In order to increase the quality of the present paper, future trends or ideas that drive the future of the topic are required.

Author Response

Please include keywords.

 

Added

 

Quality of Figure 1 must be improved.

 

Done

 

Please combine Figure 2 and Figure 3 by using a bar plot representation.

 

Figures 2 and 3 are original outputs from the Bibliometrix software and can not be presented in an alternative way

 

Please improve quality of Figure 5. The journals are not completely visible

 

 

The figure 5 is also the original figure from the Bibliometrics sofware and can not be improved. A Supplementary file listing  the journals from the Figure 5 was added

 

 

Please add a paragraph about the main Editorial publishing papers in the topic.

There were no editorial papers

 

According to lines 212-220. Please add a discussion about the use of Machine Learning techniques, software/version, topic of application. Similar to Machine Learning, add discussion involving deep learning techniques. In Table 5, a deductive thematic analysis is presented, but, to consider a proper analysis more specific information is required.

 

The discussion of the deductive analysis results was added

 

A more detailed discussion about advantages and disadvantages of integration and usage of AI in the domain of propensity score matching is required.

 

Added to the discussion

 

In order to increase the quality of the present paper, future trends or ideas that drive the future of the topic are required.

 

The future trend section has been expanded

Reviewer 4 Report

Comments and Suggestions for Authors

The manuscript presents a synthetic thematic and bibliometric review of the intersection between AI and Propensity Score Matching (PSM) in medical research.

Comments (answers can [and should] be used to improve the manuscript):
1. See https://www.mdpi.com/journal/applsci/instructions - The abstract should be a single paragraph and should follow the style of structured abstracts, "but without headings"...
2. This part of the abstract is weird: '...in complex areas like cardiovascular disease, cancer, and diagnostics'; diagnostics of what?
3. Avoid defining abbreviation multiple times. For example, AI and PSM are defined a lot of times. Definition is required only for the first time in the text.
4. "...into following more specific research question" -> questionS
5. Define explicitly the study objective at the end of the introduction.
6. Unpack 'STA'. The manuscript would benefit from a clearer conceptual explanation of how STA differs from, for example, classical systematic reviews, scoping reviews, and standard bibliometric thematic analyses.What are the epistemological assumptions of STA? How subjective is the synthesis step when naming clusters and themes? How reproducibility is ensured when moving from keyword networks to thematic interpretations? A short comparative paragraph would significantly strengthen methodological transparency.
7. "5. Synthesize the most prominent publications" -> this is step 6.
8. The search is limited to Scopus and the Medical subject area, together. This choice introduces potential bias. Relevant AI-driven causal inference studies published in computer science or statistics journals and conferences may be excluded. Why have you not considered including other digital libraries / article databases like Web of Science, PubMed, and IEEE Xplore?
9. Consider "Bibliometrix was run in RStudio version 2025.09.1+401, which was utilising the R programming language version R-4.5.1 for Windows. VosViewer was run under the latest version of the programming language, Java 8 Update 461, using the Java Runtime Environment (JRE) 1.8.0_461."; this information is irrelevant.
10. How do you know "with no papers scheduled for 2026"? The authors do not have access to the schedule of journals or conferences.
11. Remove titles from figures, because all of them have already a caption.
12. The narrative text 'Synthesis of publications' in table 4 is boring. This content could be summarized or better presented.
13. Table 5 is extremely confusing. There are different overlaps between terms. For example, in 'AI algorithms and services', artificial intelligence is a core concept related to machine learning, which also includes deep learning. It's hard to separate them. This also includes the different algorithms and methods in the same column of the table.
14. Figures 7 and 8 need to be explained in the text, because they are not self-explanatory.

Author Response

See https://www.mdpi.com/journal/applsci/instructions - The abstract should be a single paragraph and should follow the style of structured abstracts, "but without headings"...

Corrected

 

This part of the abstract is weird: '...in complex areas like cardiovascular disease, cancer, and diagnostics'; diagnostics of what?

 

Corrected

 

Avoid defining abbreviation multiple times. For example, AI and PSM are defined a lot of times. Definition is required only for the first time in the text.

 

Corrected

 

 "...into following more specific research question" -> questionS

 

Corrected

Define explicitly the study objective at the end of the introduction.

 

Done

Unpack 'STA'. The manuscript would benefit from a clearer conceptual explanation of how STA differs from, for example, classical systematic reviews, scoping reviews, and standard bibliometric thematic analyses.What are the epistemological assumptions of STA? How subjective is the synthesis step when naming clusters and themes? How reproducibility is ensured when moving from keyword networks to thematic interpretations? A short comparative paragraph would significantly strengthen methodological transparency.

 

The introduction was expanded with a short description of the  STA

Synthesize the most prominent publications" -> this is step 6.

 

Corrected

The search is limited to Scopus and the Medical subject area, together. This choice introduces potential bias. Relevant AI-driven causal inference studies published in computer science or statistics journals and conferences may be excluded. Why have you not considered including other digital libraries / article databases like Web of Science, PubMed, and IEEE Xplore

 

 Scopus includes PubMed and almost all of the journals covered by WoS and iEEE.- The exclusion of possible publications was discussed in the study limitations

 Consider "Bibliometrix was run in RStudio version 2025.09.1+401, which was utilising the R programming language version R-4.5.1 for Windows. VosViewer was run under the latest version of the programming language, Java 8 Update 461, using the Java Runtime Environment (JRE) 1.8.0_461."; this information is irrelevant.

Deleted

How do you know "with no papers scheduled for 2026"? The authors do not have access to the schedule of journals or conferences.

 

Deleted

 

11. 
    Remove titles from figures, because all of them have already a caption.

 

The figures all original outputs from the sofware tools and the titles can not be removed

The narrative text 'Synthesis of publications' in table 4 is boring. This content could be summarized or better presented.

 

Table 4. presents the  content of the  influential publications identified in Table 3.  The  narrative text  in the table was rewriten

 

Table 5 is extremely confusing. There are different overlaps between terms. For example, in 'AI algorithms and services', artificial intelligence is a core concept related to machine learning, which also includes deep learning. It's hard to separate them. This also includes the different algorithms and methods in the same column of the table.

The column was renamed and further explained in the discussion section

 

Figures 7 and 8 need to be explained in the text, because they are not self-explanatory.

 

Figure 7 and 8 were explained in more detail

 

Reviewer 5 Report

Comments and Suggestions for Authors

The manuscript presents a novel topic that addresses the current integration of medical statistics and data science. The systematic review of AI combined with Propensity Score Matching (PSM) is timely. However, I believe there is room for improvement regarding terminological accuracy, methodological details, and the depth of discussion. Below are five specific suggestions for revision:

 

1.Terminological Accuracy and Spelling Upon reviewing the full text, I identified several serious errors in terminology definitions and spelling oversights that significantly undermine the professionalism of the manuscript. These must be corrected before publication:

• Line 199 Error: The text states "...research on artificial intelligence (AI) and patient safety management (PSM)." This is a fundamental error. In the context of this paper, PSM refers to Propensity Score Matching, not "patient safety management." This error must be corrected immediately.
• Line 61 Spelling Error: The text reads "When thus combined, AI and PMS enhance...", where PSM is misspelled as PMS. Please proofread the entire manuscript to fix such typos.
• Line 57 Syntax Suggestion: The phrase "...(the use of neural networks, decision trees, using AI to...)" appears redundant and awkward. I suggest rewriting it as: "utilizing neural networks and decision trees to automate variable selection..."

 2.In the "Limitations" section, the authors should explicitly discuss the potential Selection Bias caused by the current search strategy. Limiting the search strictly to the medical field may exclude foundational methodological papers that propose "using deep learning for propensity score estimation". This omission could affect the comprehensive assessment of the "AI-assisted PSM" direction.

 

3.The current "Results" and "Discussion" sections are predominantly Descriptive, focusing on listing and describing the literature. The manuscript lacks a deep technical Evaluative analysis regarding the mechanisms of combining AI and PSM. I recommend adding a critical analysis of how these methods improve upon traditional statistics, rather than just stating that they are used.

 

4.Table 4 occupies an excessive amount of space and essentially consists of a simple stacking of abstracts from multiple papers. This results in a poor reading experience and lacks true Synthesis. I suggest significantly reconstructing Table 4. Please avoid copying large blocks of abstracts for each paper; instead, summarize the key contributions, methodologies, or findings concisely to aid the reader's understanding.

 

5.To enhance the discussion of the application background, particularly regarding Dynamic Learning, please reference and discuss the latest literature (2025 and relevant references). Incorporating recent studies on dynamic adaptation and intelligent control will provide a more comprehensive context. Examples of relevant literature include:"ISCC: Intelligent Semantic Caching and Control for NDN-Enabled Industrial IoT Networks""Unsupervised domain adaptation for lithology classification using dynamic entropy-based prototype learning"

Author Response

Terminological Accuracy and Spelling Upon reviewing the full text, I identified several serious errors in terminology definitions and spelling oversights that significantly undermine the professionalism of the manuscript. These must be corrected before publication:

 

Done

 

Line 199 Error: The text states "...research on artificial intelligence (AI) and patient safety management (PSM)." This is a fundamental error. In the context of this paper, PSM refers to Propensity Score Matching, not "patient safety management." This error must be corrected immediately.

Corrected

Line 61 Spelling Error: The text reads "When thus combined, AI and PMS enhance...", where PSM is misspelled as PMS. Please proofread the entire manuscript to fix such typos.

Done

Line 57 Syntax Suggestion: The phrase "...(the use of neural networks, decision trees, using AI to...)" appears redundant and awkward. I suggest rewriting it as: "utilizing neural networks and decision trees to automate variable selection..."

 

Done

 

In the "Limitations" section, the authors should explicitly discuss the potential Selection Bias caused by the current search strategy. Limiting the search strictly to the medical field may exclude foundational methodological papers that propose "using deep learning for propensity score estimation". This omission could affect the comprehensive assessment of the "AI-assisted PSM" direction.

 

Done

 

The current "Results" and "Discussion" sections are predominantly Descriptive, focusing on listing and describing the literature. The manuscript lacks a deep technical Evaluative analysis regarding the mechanisms of combining AI and PSM. I recommend adding a critical analysis of how these methods improve upon traditional statistics, rather than just stating that they are used.

 

Table 4 occupies an excessive amount of space and essentially consists of a simple stacking of abstracts from multiple papers. This results in a poor reading experience and lacks true Synthesis. I suggest significantly reconstructing Table 4. Please avoid copying large blocks of abstracts for each paper; instead, summarize the key contributions, methodologies, or findings concisely to aid the reader's understanding.

 

Done

 

To enhance the discussion of the application background, particularly regarding Dynamic Learning, please reference and discuss the latest literature (2025 and relevant references). Incorporating recent studies on dynamic adaptation and intelligent control will provide a more comprehensive context. Examples of relevant literature include:"ISCC: Intelligent Semantic Caching and Control for NDN-Enabled Industrial IoT Networks""Unsupervised domain adaptation for lithology classification using dynamic entropy-based prototype learning

 

In the Discussion sections we discussed only our own results, no similar studies have been published yet, and only four review papers relating AI and PSM but in concrete and very specific medical problems. The dynamic adaptation and intelligent control has not yet been used in PSM and the authors can not see how it can employed in AI driven PSM

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors did not incorporated the suggestions that I suggested for.

This research article has performed thematic analysis of literature to systematically evaluate the existing literature on AI and PSM in medicine. This paper proposes an AI-driven framework that leverages deep learning techniques to enhance predictive performance in a domain-specific application, aiming to address limitations of traditional and shallow machine learning approaches. While the manuscript presents promising results with in-depth analysis, the level of methodological novelty appears limited. The proposed architecture and training strategy are largely consistent with previously published approaches. The contents in the paper are well-improved however, I have a few concerns about my last comments and have new comments as well. Some of my round-1 comments are not properly addressed in the revised version that I would recommend the authors to address while following are some of my new concerns on the paper.

1. The manuscript lacks a sufficiently detailed description of the datasets used for:
   1. experimentation, including their size, distribution, potential class imbalance, and acquisition process.
   2. Providing this information is essential to assess the reproducibility, reliability, and generalizability of the proposed approach.
2. Although performance comparisons are included, the selection of baseline methods does not fully reflect the current state of the art.
   1. Incorporating more recent and competitive models would significantly strengthen the experimental validation and better the proposed method within the existing literature.
3. The evaluation primarily focuses on accuracy-based metrics, which may not fully capture the effectiveness of the proposed framework, particularly in real-world or imbalanced scenarios.
   1. The authors are encouraged to include additional metrics (e.g., robustness, computational complexity, inference time) to enhance the practical relevance of the study.
   2. There is no information that the data used in the literature are balanced or imbalanced. If imbalanced then how the authors in these studies performed analysis.
   3. Whether some data augmentation is performed or not to address data-imbalance, if yes then what strategies they followed.
   4. Data interpretation is reported or not? Because explainability is one of the key aspects in healthcare domain.
4. The reported performance improvements are notable; however, the manuscript does not sufficiently address the risk of overfitting. More rigorous validation strategies, such as cross-validation, external testing datasets, or ablation studies, would improve confidence in the model’s generalization capability.
5. The discussion section largely reiterates the experimental results without critically analyzing the limitations of the proposed approach. A more in-depth discussion of potential weaknesses, failure cases, and directions for future work would enhance the scientific rigor and transparency of the manuscript.

Comments for author File: Comments.pdf

Author Response

I would like to thank the reviewer fort he comments however I am not sure that I am able to adequately respond to them, because we didn’t use a conventional data - database but a bibliographic - database (Scopus) without any classes and we didn’t perform classification or any type of “classical” machine learning analyisi

 

The manuscript lacks a sufficiently detailed description of the datasets used for experimentation, including their size, distribution, potential class imbalance, and acquisition process.

The site of the database in the number of papers is already given in the paper, but as said above there are no classes, no imbalance, distribution, etc

Although performance comparisons are included, the selection of baseline methods does not fully reflect the current state of the art. Incorporating more recent and competitive models would significantly strengthen the experimental validation and better the proposed method within the existing literature.

 

We didn’t perform  performance analysis, because we didn’t use any use any models and thus didn’t used any metrics, data augmentation, etc there is is not needed for them in  the type of study we used

Reviewer 2 Report

Comments and Suggestions for Authors

I confirm my opinion that this paper is not suited to the applied sciences journal, it is bibliometric politically driven paper with no scientific contribution

Author Response

I would like to thank the reviewer for this opinion, with which we do not agree. The applied contribution of our study ican be that the  study operationalizes the AI-PSM symbiosis, into actionable evidence.  By identifying specific thematic clusters, most used approaches, etc it provides a methodological roadmap for clinicians to adopt AI-driven treatment planning, effectively bridging the vital "research-to-practice" gap and significantly improving high-stakes healthcare decision-making processes.

 

Reviewer 3 Report

Comments and Suggestions for Authors

My comments have been properly addressed.

Author Response

Thanks for reviewing the paper

Reviewer 4 Report

Comments and Suggestions for Authors

The authors addressed most of my concerns, but some issues remain:

My comment: Avoid defining abbreviation multiple times. For example, AI and PSM are defined a lot of times. Definition is required only for the first time in the text.
Authors' reply: Corrected
My reply: This problem remains in different parts of the text.

My comment: Remove titles from figures, because all of them have already a caption.
Authors' reply: The figures all original outputs from the sofware tools and the titles can not be removed
My reply: Any image can be edited.

My comment: The narrative text 'Synthesis of publications' in table 4 is boring. This content could be summarized or better presented.
Authors' reply: Table 4. presents the  content of the  influential publications identified in Table 3.  The  narrative text  in the table was rewriten
My reply: It was rewritten, but it is still boring and verbose. It is just a summary of the study, which can be easily accessed by reading the paper abstract. This long, narrative text inside a table is not useful.

My comment: Table 5 is extremely confusing. There are different overlaps between terms. For example, in 'AI algorithms and services', artificial intelligence is a core concept related to machine learning, which also includes deep learning. It's hard to separate them. This also includes the different algorithms and methods in the same column of the table.
Authors' reply: The column was renamed and further explained in the discussion section
My reply: This issue remains. Renaming the heading is not a possible solution to address it.

Author Response

Reviewer 4

 

My comment: Avoid defining abbreviation multiple times. For example, AI and PSM are defined a lot of times. Definition is required only for the first time in the text.
Authors' reply: Corrected

My reply: This problem remains in different parts of the text.

 

We removed the remaining abbreviations definitions

 

My comment: Remove titles from figures, because all of them have already a caption.
Authors' reply: The figures all original outputs from the sofware tools and the titles can not be removed
My reply: Any image can be edited.

 

We apologize for the wrong wording, of course every image can be edited, but what we want to say was that we would not like to “mess” with the original  images. Actually many journals forbid to that. However, we deleted the titles in the images

 

My comment: The narrative text 'Synthesis of publications' in table 4 is boring. This content could be summarized or better presented.
Authors' reply: Table 4. presents the  content of the  influential publications identified in Table 3.  The  narrative text  in the table was rewriten
My reply: It was rewritten, but it is still boring and verbose. It is just a summary of the study, which can be easily accessed by reading the paper abstract. This long, narrative text inside a table is not useful.

 

We appreciate the reviewers comment however we do not agree with his view of the usefulness of the column 'Synthesis of publications'. The aim of this column is to enable readers to get a fast but holistic  overview of the influential research without searching for single publications and reading the abstracts. In that way readers can easily identify which themes, topics , publications etc are really interesting for their further research.

 

My comment: Table 5 is extremely confusing. There are different overlaps between terms. For example, in 'AI algorithms and services', artificial intelligence is a core concept related to machine learning, which also includes deep learning. It's hard to separate them. This also includes the different algorithms and methods in the same column of the table.

Authors' reply: The column was renamed and further explained in the discussion section
My reply: This issue remains. Renaming the heading is not a possible solution to address it.

 

The Table presents the actual  terms authors used in their abstracts and keyords. Some authors used generic claimslike We used artificial intelligence/machine learning but some used more granulate terms. This s now explained in the rewritten text

Reviewer 5 Report

Comments and Suggestions for Authors

1. The abstract is too long, and the structure should be reasonably combined, not too detailed
2. Figure 5.What does the icon on the right mean?
3. Table 4 should be appropriately summarized
4. Format references, etc., please take the format seriously

Author Response

1. The abstract is too long, and the structure should be reasonably combined, not too detailed

 

The abstract has been shortened to less than 200 words and rewritten

2. Figure 5.What does the icon on the right mean?

 

That the analysis was performed b Bibliometrix software

3. 
   Table 4 should be appropriately summarized

 

The summary of table 4 was added

4. Format references, etc., please take the format seriously

 

The references were reformatted with Zotero using MDPI stylesheet