Review Comment:
This paper describes a platform called metaphactory that enables creating and managing knowledge graphs connected to heterogeneous data sources, and building application UIs to interact with these knowledge graphs and associated data sources.
The paper addresses an important problem and fits the scope of this special issue pretty well. I have little doubt that the described platform is useful to the type of organization targeted. However, I am having a very hard time seeing a research contribution in the paper as it is written, that would be of clear interest to the academic community. The lack of positioning with respect to related work is symptomatic of this problem, but had this been the only issue, I would have recommended a minor revision. The problem with the paper in its current form, however, is much deeper, as detailed below. Writing systems papers for academic venues is particularly challenging. There are many pitfalls, which the paper fails to completely avoid.
After reading the introduction, I was still not sure what is the contribution w.r.t the state of the art. What specific challenges have been addressed? How? This remains unclear as the introduction consists mostly of a wishlist at a very high level of abstraction. A clear contribution statement would be the minimum, but I believe the paper would benefit significantly from a complete rewrite of the introduction to make these things crystal clear.
It isn't clear where the Architecture section is going. What is novel here? This is a very flat, somewhat lengthy description, without any clear rationale for the choices made and what particular challenges have been tackled.
Section hints at where the research effort had focused (page 4), but this never developed, at least not to a significant extent. Not to mention that part of what seems to be novel in this research has already been the subject of publications at ISWC (albeit as posters or demos, as far as I understand). Related to the lack of positioning w.r.t related work, Section 3 fails to pinpoint specific limitations of existing solutions and to explain Ephedra addresses those limitations. The description remains too abstract. For instance, the last paragraph of Section 3 (page 5) hints at some interesting aspects of hybrid query processing, but remains far too vague for this to be informative to the interested readers.
Section 5, which deals with the user interface aspects of the platform, is maybe the weakest. Everything that is stated in this section is already known. The paper completely ignores (except for two anecdotal references) the very large body of work on visualization and user interfaces for semantic web/linked data, even failing to acknowledge the challenges related to generating meaningful representations of semantic data. While the issue of UI design and development in the context of such a platform is indeed key, the paper fails to provide any convincing solution, evidence, or even insightful discussion beyond what has already been said many times in the literature. One exception to this could be Section 5.2 about GraphScope, but again, this section give far too little information about it. It isn't clear how it works, and how users interact with it.
Finally, the use cases bear relatively little value. While it is certainly a good thing that the platform is used in various projects, it is unclear how the descriptions provided in Section 6 provide actual validation (from an academic/research perspective) of the approach investigated in this platform.
All of that said, my review is more about the paper's contents (how it is framed) rather than the platform itself, which might very well be worthy of publication in a scientific journal. The presentation of the work, however, needs significant revision.
|