Systematic Literature Review on Evolvable Knowledge Graphs in Manufacturing

Thanks for revising the manuscript. This revision demonstrates substantial improvements across multiple dimensions, addressing major concerns raised in previous review while making meaningful contributions to the systematic literature review methodology and theoretical understanding of evolvable knowledge graphs.

The comprehensive search strategy spanning three rounds (March 2022, September 2023, April 2025) across six major databases provides thorough temporal coverage despite the inherent challenges of the rapidly evolving field. Your transparent documentation of exclusion counts and systematic screening processes demonstrates methodological rigor that meets good standards for systematic literature reviews.

The distinction between dynamic knowledge graphs (data value updates) and evolvable knowledge graphs (structural/semantic learning-based changes) represents a valuable theoretical contribution to the field. The conceptual framework has been further clarified through refined definitions, addressing previous concerns about definitional ambiguity. The three-tier categorization system for learning approaches (human-guided, machine-driven, human-machine collaborative updates) with subcategories provides a comprehensive framework that offers practical guidance for implementation decisions.

The updated Knowledge Graph Development Process (KGDP) model incorporates meaningful refinements based on systematic analysis. The integration of learning feedback loops and expanded knowledge utilization stages addressing application requirements, algorithm deployment, and user interaction processing represents a significant advancement that addresses gaps in existing process models.

The reorganization of Section 3 with improved alignment to the KGDP framework substantially improves the structure of the manuscript and its readability. The addition of Industry 4.0 and 5.0 definitions, consistent abbreviation introduction, and enhanced figure presentations with explanatory captions, is now much clearer.

Minor Technical Corrections

- please note that there appears to be a missing citation on line 14 of page 7 at the end of the sentence beginning "The Cohen's Kappa values in the first round...."

- while the theoretical contributions are conceptually sound, they would benefit from future empirical validation through case studies or expert evaluation. The distinction between dynamic and evolvable knowledge graphs, while theoretically valuable, remains primarily definitional without demonstrated practical utility differences. This limitation is understandable given the emerging nature of the research domain but should be acknowledged for future work.

- the manufacturing-specific focus provides valuable depth but limits immediate applicability to other semantic web domains. Consider including brief discussion of how your theoretical framework principles might generalize to other contexts, even if empirical validation remains beyond the current scope.

I could not find any mention to my comments in the authors' cover letter.

Thus, it was very difficult for me to check whether they addressed them in the paper.

I could manage to do it for most comments, i.e. the ones that requested some changes in the paper, and I can say that the authors at least partially addressed my comments.

Not sure if they addressed my main concern though, I think they could have easily discussed it in the cover letter: "it seems to me that, by selecting only the manufacturing domain, they may be missing important work about similar/different methodologies for KG construction applied to different domains, but they often talk about KG construction tasks in general, with no actual discussion about how some tasks have peculiar characteristics when applied to the manufacturing domain (see for example section 3.3.1)".

While I thank the authors for revising and improving the manuscript, several comments raised by the reviewers of the first version remain insufficiently addressed and the following list of opportunities for improvement remain relevant:

O1: Notion and use of evolving knowledge graphs. As outlined by all 3 reviewers of the first version, the distinction between dynamic and evolvable KGs was not clear and should be reconsidered. While the authors adopt an own definition, it would be more useful to set the use of the terms in context of related work, specifically:
Polleres, A., Pernisch, R., Bonifati, A., Dell'Aglio, D., Dobriy, D., Dumbrava, S., ... & Wachs, J. (2023). How does knowledge evolve in open knowledge graphs?. Transactions on Graph Data and Knowledge, 1(1), 11-1.
(Sidenote: while the authors provide a reference to this paper, they do not set their research in the context of this paper).
Specifically, I question whether the statement "the transition from dynamic KGs to evolvable KGs highlights a shift from changing to learning" actually makes sense. In fact, even with classic logical reasoning, KGs can be automatically extended through inference, and this is neither new nor indicates a change. Therefore, I am sceptical if this understanding holds.

O2: Insufficient concrete insights / take-aways.
While the authors refine their originally proposed KGDP, the findings for the overall research community are limited. The drawn conclusions in Section 4 are very general and little surprising. A more deep discussion and especially specific recommendations for the manufacturing domain would be required here, instead of focusing on the general topic of evolving KGs. Purely from the based review, one cannot claim to draw conclusions for the entire topic when focusing on this specific domain.

O3: Several inconsistencies appeared during the revision. For example, the authors renamed KGCP (knowledge graph construction process) to KGDP (knowledge graph development process) in the Introduction, but did not align Section 1.2 with this terminology.

O4: Not addressed (minor) comments from prior review.
I will simply copy/paste all comments that were not answered in the revision and are still deemed relevant for a final version.

- Data quality in knowledge graphs. Although a more thorough discussion about the topic is added, relevant related work is still missing, specifically: Zaveri, A., Rula, A., Maurino, A., Pietrobon, R., Lehmann, J., & Auer, S. (2015). Quality assessment for Linked Data: A Survey: A systematic literature review and conceptual framework. Semantic web, 7(1), 63-93.
- EC6: clarify how poor quality is determined.
- Table 4: information on how many papers were excluded per EC should be added. In addition, it should be mentioned whether the exclusion criteria were applied in the respective order or whether more than one could be assigned to one result.
- Fig. 4, 5, and 6 should be ordered by decreasing number of publications. In addition, for each Figure it should be mentioned whether a publication has a single assignment or multiple possible ones.
- References are still not always correct, e.g., [9] is a journal and no conference (but contains field Conference name) and [60] contains strange fields like "Artwork size" and bytes.