Review Comment:
The paper presents the Collection Ontology (CO), an OWL 2 DL ontology
(TBox + RBox) designed to represent collections (sets, bags and
lists).
The presentation is objective and clear, with a nice discussion of
related work (including RDF frameworks) and applications. The ontology
itself (and related software) is made freely available on the web.
The paper discusses the design decisions made in the development of
the CO. In this regard, I would like to offer a few questions and
remarks:
1. In Section 3 (P. 3, col. 2), the authors explain that "two sets
of people, composed exactly by the same people, [can be seen] as
two different research groups without contradictions." I agree
with this decision, but I think the explanation would be clearer
if the authors pointed out that identity of collections is not
*extensional*, which is the correct mathematical term for
considering two collections equal if and only if they have the
same elements.
2. Also in Section 3, the authors use the phrase "identity function"
repreatedly. I think the more common term in Conceptual Modeling
would be "principle of identity" instead.
3. On P. 5, col. 2, the phrase "inconsistent model" should be
replaced with "inconsistent ontology". A logician would point out
that only theories (e.g., ontologies) can be inconsistent, not
models (in the logical sense).
4. In the CO, can a list ever be empty? But then, what would be the
values of co:firstItem and co:lastItem? Cardinality restrictions
force these properties to have exactly one filler. I would like
to have this clarified and discussed in the paper.
5. Mathematically, any set is also a bag. Why are the corresponding
classes disjoint in the CO? Page 3 mentions that this was a
design decision: "from an implementation standpoint, the data
structures managing co:Set and co:Bag are very different." But
this seems to violate a principle of software design, because it
conflates the *interface* and the *implementation* of the co:Set
and co:Bag classes. In my opinion, it should be possible to treat
any set *implicitly* as a bag (i.e., co:set should be a subclass
of co:bag), and the differences in the implementation should be
hidden. As the ontology stands, the user must decide a priori
whether he will represent a set as an instance of co:set (and
never allow it to become a bag), or as an instance of co:bag with
all multiplicities equal to one (but allowing such multiplicities
to change eventually, in a dynamic context). I would like to see
a more detailed discussion of this decision in the paper.
The following typographical and grammatical errors should be fixed:
+ P. 1, col. 1: the need of describing -> the need to describe
+ P. 1, col. 2: how to answer particular queries CO collections
-> how to answer particular queries on CO collections
+ P. 1, col. 2, footnote: Working Draft
+ P. 2, col. 2: both RDF/XML and Turtle provide a compact
syntaxes -> both RDF/XML and Turtle provide compact syntaxes
+ P. 2, col. 2: OWL have no support -> OWL has no support
+ P. 2, col. 2: spatially locate -> spatially located
+ P. 3, col. 1: Gregor Cantor -> Georg Cantor
+ P. 4, col. 1: footnote marker 7 should be superscript
+ P. 4, col. 1: The class co:Item links exactly one resource
the effectively is contained in the bag -> The class co:Item
links exactly one resource that effectively is contained in
the bag
+ P. 4, col. 2: the belongingness to a bag -> membership in a
bag
+ P. 5, col. 1: Characterstics -> Characteristics
+ P. 5, col. 2: more then one list -> more than one list
+ P. 5, col. 2: third author of another.)
+ P. 5, col. 2: all the items that follows/precedes -> all the
items that follow/precede
+ P. 5, col. 2: it would violate one of the needed keep the
ontology in a DL framework -> it would violate one of the
constraints needed to keep the ontology in the DL framework
+ P. 6, col. 1, footnote 12: constraints, aware of of
+ P. 7, col. 1, footnote 14: because of -> because, accordin
-> according, a list do not -> a list does not
+ Fig. 4: exatcly -> exactly
| 
