Summary of voting

Dear all,

  here an overview of the interpretation of the voting we reached last 
Friday during our regular telco:

1) Do you agree with the core model's structure as currently defined?

There is clear agreement here, with the absolute majority agreeing on 
the model. Nevertheless, there were a few comments by Thierry that we 
could discuss directly on the telco. Armando still does not like the 
label "contains". Actually, I do not like it either and could propose 
"subsumes" again, which is a bit more intensional that "contains", which 
evokes the "set-containment" notion a bit too much for my taste. If many 
people like subsumes more than contains, we might reconsider it. For now 
I propose we live with "contains".

2) Do you agree with the definitions of the core elements as given?

After the discussion and the comments in the wiki, we decided to change 
the definitions as follows (see minutes):

Lexical Concept: the mental representation of the shared meaning of a 
collection of senses that can be exchanged in many contexts without 
substantially changing the meaning.

Lexicon: A lexicon represents a collection of lexical entries which can 
be used to refer to the elements of the vocabulary of an ontology.

Lexical Entry: A single unit of analysis in the lexicon, i.e. a 
collection of forms that can be derived morphologically from each other.

Form: The form represents a single inflected unit of a lexical entry 
with a single pronunciation, although potentially many orthographies.

Lexical Sense: A language-specific meaning of a lexical entry which 
abstracts from specific occurrences of the lemma. This class is a 
lexical sense as it represents a sense of a lexical entry.

3) Should we link directly to other models instead of using informal 
comments to link?

The majority is clearly in favour of more formal links, preferrably to 
W3C models. We should work out in detail in the next telcos to which 
models we will then formally link. I will make a more specific proposal 
on this soon.

4) Do you agree that the property|reference|does not have a formal 
range, but can refer to anything that has a URI?

The majority of us has voted for not specifying the range of reference 
in more detail, but the clear understanding is that the range should be 
some ontological entity. Nevertheless, formally the range is going to be 
owl:Thing anyway, so I propose that we add this explicitly so that 
anyone that uses the model is aware that the referenced URI then will 
formally be treated as something that has a model-theoretic 
interpretation according to OWL. Does this make sense to all of you?

In this sense I would propose to modify the definition of "reference" as 
follows:

Def. reference: the relation between a lexical sense of a lexical entry 
and the ontology entity that represents the denotation; the range of the 
reference is thus expected to be an ontological entity with a 
model-theoretic interpretation.

5) Is the name of any property or class unsuitable and requires change?

The majority has agreed on the naming. We should revisit the "contains" 
relation in due time and also discuss in more detail the name of the 
inverse properties.

6) Should the core be defined in separate namespace from further modules?

Most people have voted for having separate namespaces for submodules of 
the model. I think we should accept this and go on under this premise. 
However, I would like to stress once more that I am sure this will 
affect the usability of the model. Think of how cumbersome it can be to 
query DBpedia with 2-3 different namespaces. I always get a query wrong 
to DBpedia in the first place. One namespace would be clearly easier to 
use. The argument of having a too complex model is not a real issue IMHO 
as people will use the model mainly by copying code from somewhere. Only 
few experts will really inspect the model as a whole in Protégé or 
similar tools and they will appreciate to see the model in its 
complexity I speculate.

In any case, let's take a modular approach for now and develop different 
modules. It will be relatively easy to create one merged / monolithic 
model in case it is needed later. So I am not too worried about this 
right now.

7) Should the model be named 'OntoLex' or 'lemon' (Lexicon Model for 
Ontologies, seehttp://lemon-model.net/), on which the current model is 
based?

See my comment on a separate email on this. We should consider the 
potential loss of a growing community when coming up with YALM 
(Yet-Another-Lexical-Modell). I hope to get some consensus here. 
Otherwise, the slight majority has voted for using "ontolex" instead of 
"lemon".

8) Do you believe it would be advantageous to aim to transition the 
OntoLex W3C Community Group into a W3C Working Group?

There seems to be some agreement that there is no pressing need to go 
for a WG right now. Once we have a stable community of users, we might 
decide to go this route and profit form the feedback we get from these 
early adopters in designing the standard.

I think this is a faithful summary of the discussion we had last Friday.

Feel free to comment.

More on the roadmap for the next months in a separate email.

Philipp.






-- 

Prof. Dr. Philipp Cimiano

Phone: +49 521 106 12249
Fax: +49 521 106 12412
Mail: cimiano@cit-ec.uni-bielefeld.de

Forschungsbau Intelligente Systeme (FBIIS)
Raum 2.307
Universität Bielefeld
Inspiration 1
33619 Bielefeld