Review Comment:
The paper presents Gavel, a tool which supports developing and reasoning over ontologies written in FOWL, a first order extension of OWL. Gavel provides an Protege extension which allows the user to add FOL axioms to OWL ontologies and it performs reasoning over such heterogeneous ontologies by translating all data to TPTP and then executing it in a theorem prover. The authors describe their implementation and show various examples for which the addition and/or evaluation of FOL axioms to OWL ontologies provide an added value.
As such, the paper is well written and very interesting. I also think that there are OWL use cases that could benefit from the addition of FOL formulas. However, I still see two problems with the paper and the approach:
Firstly, the authors do not even mention that OWL-DL is designed for the Web and that many design decisions like for example the use of IRIs reflect that. So I wonder, what happens if we combine FOWL ontologies with really similar FOL annotations which refer to different contexts (think for example in some concept “name” which relates to a foaf:name in one ontology and to some ex:name in another). How do we combine and exchange FOWL ontologies in general? Also: the representation of the FOL formula in a literal might ensure compatibility with OWL, but it also makes it difficult for other standards and implementations (e.g. SHACL, N3, …) to access this data in order to (partly) perform FOL reasoning as well.
Secondly, I would like to learn more about the concrete mapping. The authors provide a link, but did they also formally show the correctness of the mapping? Does the mapping preserve equivalence and non-equivalence.
Most likely, the authors only forgot to add references to address my second problem. Therefore, I am more concerned about the first one and would like to see at least some considerations of the Web-aspect of OWL. But I think, a short discussion treating these issues can be easily added. I therefore recommend a minor revision.
Some more comments:
- Figure 1: the second clif expression denotes symmetry while it should be antisymmetry in order to get a strict order? Maybe I misinterpreted the expression?
- Introduction: you claim that OWL-DL is easier to understand for users than FOL
(CLIF, TPTP). Of course that claim is subjective, but I doubt that given for example that restriction posed on the axioms which are just there to ensure decidability. I would instead add as an advantage the compliance to Semantic Web principles.
- If you do OWL reasoning via Gavel-OWL, do you also have a mapping from (restricted) FOL to OWL to give back results?
- page 9 end, there seems to be a wrong full stop (“. and used…)
- related work: you discuss the problem that OWL-DL combined with swrl is not decidable and that therefore dl-safe swrl was introduced. After that, you talk about swrl+owl which is less expressive than FOL, do you refer to dl-safe swrl or general swrl in that sentence? I am also confused y the wording of the paragraph, since the lack of decidability is presented as a problem while it is clear, that FOL is also not decidable. So how is that part meant?
|