OWL Representation of the Common Information Sharing Environment Data Scheme for the Maritime Domain

In this article, the authors present an OWL transformation model for a maritime environmental data schema (CISE) specified in XML, which represents ships and other vessels and their involvement in events. The goal is to use CISE data in a semantic model in order to do consistency checks, ontology based queries and reasoning on the data. The authors describe their reingineering approach including the mapping of XSD schemes to OWL classes, properties and statements, as well as associations (complex relations), enumeration classes and constraints. Some general manual as well as automated evaluation was done.

While I can see that this approach to generate CISE ontologies makes sense and might be useful and needed for the maritime domain, I think the article suffers from a lack of a clear argumentation from goals/requirements to design choices and evaluations. It is a compilation of many practical (and probably meaningful) steps which would deserve a better motivation. There are also some questionable choices in ontology design regarding events and association classes. The authors should go through their paper and address these issues one by one.

- First, regarding motivation and introduction, it should be made more clear how the suggested approach deviates from earlier CISE ontologies (namely the one by Riga et al 2021 [2]). To what extent do the authors address the gaps in this previous ontology? Why is the previous ontology not available (can you contact the previous authors?). The mentioned goal is to improve the automation of the CISE -> ontology transformation which still requires manual work. However, the degree of automation in the new ontology seems still limited (manual validation 5.3 etc). The authors should therefore explain and show/test how their approach improves automation with respect to earlier approaches. The authors say they expand the CISE vocabulary by land surveillance terms etc, but never really show that they did this (remove?). The authors say they integrate different approaches, but I missed an explanation of what can be done with the integrated approach that cannot be done with the previous ones. In general I also missed a clear goal/question in the introduction which is more specific than just interoperability. And furtherdown in the text, I missed a consistent referral back to the problems raised in the introduction to show how the proposed transformation approach/ontology accounts for them.

- Section 2 looks extensive, but it should show what the research gap concerning other approaches consists in. How does the suggested approach differ/build on the others?

- In Section 5.2, again the authors reuse rules from [2] but dont say to what extent they go beyond those rules. Furthermore, they introduce two new ontological notions "Association Class" and "Enumeration", but without properly motivating the need for these notions. It should be illustrated why we need these based on examples and they should be compared against existing similar ontological concepts. For example, association classes are similar to "reified relations" in RDF (https://www.w3.org/TR/swbp-n-aryRelations/) and there is a lot of literature about the latter. Does this approach differ and if not, shouldnt the standards be reused? Similarly, the notion of ObjectEvent as an association class repeats the idea of "participation" in event ontologies (see e.g. Rodriguez et al: What to consider about events: A survey on the ontology of occurrents). It also seems that the modeling choices can be simplified here: In standard event ontologies, objects can participate in events, and events themselves have time periods. Thus no need to first introduce ObjectEvent as a association class, and then link its instances to an actual event via "hasEvent" and to a period. Rather, events can be directly linked to periods and objects participating in them. Furthermore, "InvolvedEventRel" and "InvolvedObjectRel" seem to be equivalent and thus unnecessary just due to the fact that events are not properly modeled (probably there is a reason why these are not distinguished in the CISE model...). More generally, how does the proposed Event model compare to e.g. dul:Event? Regarding enumerations and numeric constraints, it remains uncelar why they are needed and what precisely the problem is that is addressed here.

The conclusion lacks a discussion regarding the goals of the introduction. The authors write in their conclusion that their proposal is the first attempt to enhance CISE using semantics, but what about the ontology in [2]? In general, the article needs to be thoroughly revised an rewritten to make a good case for the surely useful work that has been done.

The ontology seems complete and is published using a standard specification scheme with links to various RDF serializations. The long-term availability is unclear, though (maybe add to github or Figshare?).

The paper describes the work done by the authors in the context of the European project H2020 EFFECTOR, and in line with this journal and the funding agency requirements, the output of the work is publicly available. The chosen sharing support is appropriate and the shared resources appears to be complete and sufficiently documented to enable experiment replication. The shared resources include the software code for the automated conversion of the XML specifications of the two most relevant maritime data sharing models in the European ecosystem in OWL, and the ontologies resulting from such conversion.

The two ontologies, which have been manually validated by subject matter experts, are potentially very relevant for the operational community. To improve their potential reusability and facilitate the alignments with other relevant ontologies, the authors have manually defined the necessary correspondences. The obtained ontologies appear to be of good quality with respect to standard ontology evaluation approaches.
Within the project the two ontologies have been used to define a data integration layer for partners datasets. However, how this semantic integration layer contributed to the definition of the project maritime services is discussed in vague terms. The only two event/anomaly discovery specifications remain at the level of toy examples, and the potential advantages (and the shortcomings) of the ontologies in supporting event/anomaly discovery with respect to other data management related solutions are not discussed. Neither the overall data management infrastructure of the project with the integration with the two ontologies, nor the user evaluation, are presented or referred. Note that the integration of this work with the information system infrastructure of the project is given as future work in the conclusions, but the project appears to have ended in the second half of 2022.

The other contribution of the work is a XML to OWL transformation process, developed to automate the creation of the two ontologies, whose software is also shared on GitHub. The transformation process has been refined to take into consideration the end users' validation of the ontologies, to improve the accuracy of the matching between the original data sharing model and the ontologies. Despite it's not possible to guarantee that the software could be re-applied as it is for future versions of the same data models, the transformation, which is proposed as an improvement of the state of the art of generic XML to OWL transformation, seems promising in this respect. As a potential improvement of the current paper, the authors should discuss not only the pros of the approach, but better present what are, from their perspective, the expected limitations. Some considerations concerning future work partially address this aspect but could be expanded to better grasp the value of the proposed approach.

Overall, and despite few typos, the paper is easy to follow, and the design choices applied seem reasonable and effective despite not always completely justified. Figures and tables support the paper content. As a minor note, either the size or the resolution of Figure 1 must be improved, the text of the small boxes is not readable in the printed version of the paper.

Beside a suggested revision to improve the aspects above, I would also encourage the authors to revise carefully the references and the list of relevant related projects.
Most of the references are quite old, and somehow clustered around the work of few research groups. Old and less relevant references could be removed at the advantage of more recent work. Similarly, the list of related projects is a bit outdated. Recent maritime-related projects have also addressed aspects of interoperability that could be relevant to this work.