============================================================================
REVIEWER #1
============================================================================
---------------------------------------------------------------------------
Comments
---------------------------------------------------------------------------
A well written paper that deals with significant issues in ontology retrieval.
The tool is well-defined and explained, it is based on an acceptable platform
such as Gate but more results are needed in the final edition. Moreover, some
important references are missing.
============================================================================
REVIEWER #2
============================================================================
---------------------------------------------------------------------------
Comments
---------------------------------------------------------------------------
Introduction:
- The paper is a little unclear whether the system supports natural language in
general, or a well-defined subset of it (the Controlled Language). Then, you
also say that keyword search is supported. It'd be nice having the full paper
using more precise wording (replacing for instance, "CLOnE QL,for querying
knowledge stores in natural language" by something more concise). The full
paper shall also cover a description of the controlled language that is being
covered, and also describe to which extent users must respect the syntactic
constructions of this language. At the end of Sect. 2 you say, i.e., that users
have the freedom to enter queries of any length and FORM. Do such queries
translate nicely into formal (SQL) ones?
Related Work:
- There _must_ be a massive amount of related work in the area of natural
language interfaces to databases (not specifically RDF stores). How does their
work relate to yours? How do ontologies, stored in your RDF store, make a
difference, compared to "just" having databases schemes (which are also related
to each other). This relates to the reasoning mechanisms (that you haven't
really described in the abstract) but also to the syntactic/semantic processing
of these DB folks.
CLOnE QL Implementation:
- It seems that the OntoResChunk A. identifies relations between identified
ontology resources. If yes, please say so, and also which kind of relations you
are able to parse and translate.
- The full paper should give more details about the reasoning comp. of the
knowledge store.
- Are the highest scoring queries, as computer from user input, shown to the
user, say for validation? In fact, do the users see figures 2 and 3 when
interacting with the system? Do they know, for instance, that "parameters" map
to "Resource Parameter"
- are prepositions part of the controlled language, e.g., is there a
construction of the form "X in Y" that may translate into relations like
"contains..."
Evaluation:
That's the weakest part of the paper. You should have hinted at some evaluation
results (there was space for another good paragraph!).
Minor spelling errors:
- "claims to have A Google-like..." (page 1)
- "on its the position" (sic) (page 2)
- hypenation of know\-ledge
============================================================================
REVIEWER #3
============================================================================
---------------------------------------------------------------------------
Comments
---------------------------------------------------------------------------
In Section 2, you mention that a disadvantage of SemSearch is that it
does not consider properties. In Section 3, you mention two factors
for scoring retrieved relations: similarity of relation name and
chunk, and position in property hierarchy (if existing). For those not
familiar with ontology terminology, it would probably be clearer if
you use either the term relation or property.
For the evaluation (not included in this abstract, is it done yet?),
you could possibly consider also a more general ontology than the GATE
ontology, to make the results more legible for non-GATE users.
