Review Comment:
This paper presents a serialization of EUCISE XML schema and UML diagrams as an OWL ontology. Although there has been a cosiderable effort for the development of the ontology, it is not well presented in the paper. For this reason the paper should be revised.
Specifically, authors in introduction state that the motivation for developing the proposed ontology, is to benefit from semantic interoperatbility and reasoning capabilities. In my humble opinion, there should be some section in the paper, proving that the proposed ontology indeed satisfies the corresponding goals, i.e. presenting SPARQL queries, correspondences to concepts or imports to other ontologies and experimental results on reasoning tasks.
Another important issue, is that section 2 outlines the EUCISE2020 main concepts (what about the relations?) but these are never explicitly described in the paper. Section 3 describes mappings between UML and OWL, but it is not clear how this section is related to the rest of the work, i.e. how have these mappings been used for the development of the ontology, in the NeOn methodology? There seem to be some misconceptions in this section, e.g. in Table 1 the mapping for UML instance to OWL individual should be "ex:instance rdf:type ex:Cls . ex:Cls rdf:type owl:Class" instead of "ex:instance rdf:type owl:Class" (the latter describe that ex:instance is a Class). In general, the examples would have been more helpful if terms of the ontology had been used instead of "ClassA", "ClassB", etc. Also, authors should justify the need for association Classes/Roles, explain how these are used and how they affect the performance of query answering and reasoning tasks.
Section 4 presents the OWL implementation of EUCISE (i.e. the main contribution of this work) and it should delve into more details, i.e. describe the core concepts and roles, provide examples and explain the use of predefined individuals. Did authors consult also other documents (including other ontologies) for the development of this ontology, or only EUCISE? If they have not consulted other documents, how do they guarantee semantic interoperability? Is this ontology only related to SKOS schema? What about other domain specific concepts (e.g. Agent, Location, Event)? The Use Case presented in section 5, could have been used as a running example in the section presenting the core concepts and relations of the proposed ontology, i.e. it could explain why association classes are necessary (and why OWL relations do not suffice).
Also important issue, is that the proposed ontology is not compared to the related work and the need for another ontology in the domain is not justified. Authors should answer at least the following in this section:
b) Does the proposed ontology overlap or complement other ontologies in related work?
c) Where does this ontology excel and where does it fail?
Finally, the ontology evaluation section is not actually evaluating the ontology. Authors have tested the correctness (consistency) of the ontology using common reasoners, but they have not presented performance results on reasoning tasks using a populated with real data version of the ontology. The figures alone presented for attribute and inheritance richness, do not justify that the proposed ontology excels over the other ontologies in the related work.
|
