An RDF/OWL2 model to represent complex, annotated and temporally limited relations between concepts in a cultural heritage context

This paper aims at describing a so called model for representing complex relationship, annotated with provenance information and time constraints in the cultural heritage domain.

Although addressing a concrete modelling issue for this domain (and for other domain as well I suppose), in my opinion the paper suffers from several limitations:
1. I would say that the paper describes some modeling tricks rather than a model since no metamodel or fine description of the components of the model are provided. The access to the used datasets is not given as well which also limits the understanding of the modeling choices that were made.
2. The design of the model can be, to some extent, compared to an ontology building problem. To this end, the methodology that has been followed to construct this model is not sufficiently described. Even if there are some aspects that are mentioned (like the two or three competency questions of the section 1.2 lines 13-16), the main steps of the ontology construction methodology are not explained.
3. It is also unclear to me what has been reused from CIDOC, what is coming from RDF and what is coming from OWL2, what components of the OWL Time ontology has been reused. As an example I can cite section 3.1 lines 40-42 “… and the second can be inferred”. As in this section the authors are mentioning RDF I cannot understand how such property can be inferred as in RDF the only thing that can be inferred is: a node is linked to another node. This kind of sentence shows some lack and is making the reading really confuse.
4. The related work is focusing on semantic web technologies only (which is fair enough regarding the conceptual choices) but how other communities like conceptual modelling are approaching the problem? I would be surprised if others have never faced that particular problem.
5. The model is not evaluated. This is problematic for a journal paper moreover, in the abstract the authors stated “we are convinced that the resulting model …”. How can readers be convinced of the validity of the model if it is not evaluated? As evaluation is part of their future work, I encourage the authors to resubmit this paper with experimental results and a more solid evidence of the added value of their model.

The contribution is relevant for the special issue on Semantic Web for Cultural Heritage and has a reasonably well-defined contribution that is of interest to many, but there is room for improvement. On one hand, the contribution is an implementation of n-ary relations as classes in RDF and OWL-2 property chains. On the other hand, the contribution is an implementation of temporality based on CIDOC CRM and OWL-Time.

The introduction to the article is rather lengthy and somewhat focused on the particular project this work is developed in the context of. A slightly more general presentation of the problem would increase the value of this and maybe attract more readers and citations.

For the background study, the authors have investigated CIDOC CRM , the Wikidata model, and the OWL-Time ontology. I would also expect to see a discussion of Yago which is a “time and space aware” ontology. The reference list is somewhat short, and not convincing that the authors have fully explored relevant research – particularly when it comes to implementations of n-ary relations where most references are to W3C papers. Even if this is considered to be a practitioner’s papers, it should clearly position itself within the related field of research and should have a more well-defined study on related work. It is also a bit confusing that some context and related work is covered in section 2, 3 and 4, but the more formal presentation of related work is given in section 6.

The main contributions are presented through discussions and examples. A more formal presentation of the proposed model is needed to pinpoint what exactly is proposed – otherwise, this work is a limited experience that is difficult for others to reuse in another context or to elaborate on.

The presentation is somewhat verbose and descriptive – and very readable with only a few cases where this reviewer finds odd phrases. Some more clearly stated issues and principles could significantly increase the visibility of the contribution and be helpful in judging if the solution fits the issues they attempt to address. As of now it the problem is described but I miss a clearer statement for instance in the form of a list of requirements.
The use of figure is good but some figures appear rather repeatable (why separate figures for 2 and 3?).

The contribution lacks a presentation of a systematic evaluation, although there are hints in the conclusion that some evaluation has been performed. A separate evaluation section with a more elaborate presentation of experience and any more formal evaluations that have been performed would contribute significantly to the value of this article.

This manuscript was submitted as 'full paper' and should be reviewed along the usual dimensions for research contributions which include (1) originality, (2) significance of the results, and (3) quality of writing.

Regarding the above mentioned dimensions the paper is well written and presents a proposed model for representing information about cultural heritage. Combining known solutions such as CIDOC, Wikidata and OW-Time the proposed model efficiency represents cultural knowledge. Although the resulting model is complex this is necessary when representing complex relations requiring extensive use of reification and Allen-based temporal relations.
On the other hand significance of the proposed model is borderline because although efficient with respect to representation (as illustrated by the representation example) scalability may be a weakness of the model (because of the complexity of the model) when representing large datasets. Authors don't provide evidence supporting the contrary i.e. scalability has been achieved (i.e. reasoning/querying experiments over large, possible synthetic data).
The main weakness of this work is that novelty is limited. Combining and applying existing solutions for specific use cases may be a good match for in-use conference tracks/workshops/journals but not so much in case of research oriented journals/conferences.