Re: Ontolex/Lime: minutes of last meetings and some updates

On Fri, Jul 4, 2014 at 4:42 PM, Armando Stellato <stellato@info.uniroma2.it>
wrote:

> Dear all,
>
>
>
> since we’ll be working via email in these weeks, just a quick minute of
> what has been said at the last call (in particular, what is pending
> decision), and then one updates and…request for opinions on open aspects:
>
>
>
> As a general observation, we are at a good point. In the last call wee
> agreed on the overall structure, and also agreed on which part of the
> terminology can be improved.
>
>
>
> As for the last time, instead of presenting the model, I present a small
> example of its use, as it is shorter to be shown and more intuitive to be
> followed:
>
>
>
> /** inside the void file of the Lexicalization
>
> myItLex:myItalianLexicalizationOfDat
>
>   a lime:Lexicalization;
>
>   lime:lang "it";  // important to be here, this is the focus of search
> by agents!!! Not the lexicon!
>
>   lime:lexicalizedDataset :dat ;
>
>   lime:lexicalModel ontolex: ;
>   lime:lexicon :italianWordnet;
>   lime:resourceCoverage [   // see discussion later in sections 5
>     lime:class owl:Class;
>     lime:percentage …;
>     lime:avgNumOfEntries …
>   ].
>
>
>
> We already agreed in previous calls to leave aside discussion on the
> percentages/averages vs counts as the last thing so, obviously, these two
> properties:
>
>     lime:percentage …;
>     lime:avgNumOfEntries …
>
> may change also depending on which values they will host.
>
>
>
> *lime:lang* has already been agreed which can be replaced with some
> ontolex:lang. Actually, the general trend is to reinvent a lang property
> (exactly, by changing only the namespace) for each vocabulary, so to
> identify its specific use. So, for instance, dcat has its own one, with its
> dedicated domain and range, and so we could, by setting up domain of
> lime:lang to lime:Lexicalizaton. Apart from that, I’ve no strong objection
> against reusing another one.
>
One of the specific issues here is that it would be good to have an
"ontolex-all" ontology, and thus we should avoid any inter-module name
classes. Perhaps though the solution is to use the Dublin Core property and
add appropriate axioms to the definition of Lexicalization/Lexicon,
(Lexicalization ⊑ ∃ dc:language.String)

>
>
> *lime:lexicalizedDataset*: we more or less agreed on its name, providing
> that the term *Dataset* was proven to be including ontology vocabularies.
> In the meanwhile I did check on some mailing lists, and the reply from
> Richard Cyganiak (one of the authors of void) is affirmative: Dataset does
> include ontology vocabularies. This is his reply on the LOD ml:
> http://lists.w3.org/Archives/Public/public-lod/2014Jul/0012.html
>
> Note: I think there was also a proposal (maybe from Philipp) to use
> *targetDataset*. Not sure which one won, however, targetDataset is for me
> fine as well: more, if we have a Lexicalization, it **almost**
> immediately follows that its target is the dataset to be lexicalized, so
> maybe even nicer to use *targetDataset*. The only formal opposition to
> that would be that a Lexicon is a dataset too, and a lexicalization exactly
> binds a Lexicon and a Dataset to be lexicalized, so targetDataset would be
> slightly ambiguous.
>
My principal concern is the ambiguity as lexicons are also datasets... the
name of the group is OntoLex what is the problem of not just use the term
ontology to refer to what we are lexicalizing (even if some of the targets
may not be true ontologies)?

>
>
> *lime:resourceCoverage: *we agreed on its structure: it allows to
> factorize all the elements of a lexicalization in a single point (the
> Lexicalization object) and then have multiple partitions identified by it.
> However, we also agree that we may try to look for a better name :-)
> Suggestions?
>
> Actually this may be depending on that final decision on
> percentages/averages vs counts. resourceCoverage is evoked in my mind
> (though may be changed as well) if, like in this example, we have
> percentages/averages. With counts, I would be ever more tempted to look for
> something else.
>
Shall we not follow VoID here and call the object a "partition"?

>
>
> Oh, one last thing, which was left over from discussion: *LexicalLinkSet*
> s.
>
> I get back an example from a previous email: suppose that I’m (implicitly)
> lexicalizing an ontology by writing links between LexicalConcepts of
> WordNet (synsets) and the resources of the ontology. We thus have links
> between semantic entities on both sides (Lexicon and Dataset) so this
> cannot be expressed through a Lexicalization object (unless we want to
> count the non-OWL inferable lexical derivations of this semantic linking).
> So, we have the properties in ontolex core for that and I assume thus this
> is relevant for our model, and then probably it would be important to tell
> it somehow in the metadata. That’s where I suggested this
> lime:LexicalLinkSet as a subclass of void:LinkSet.
>
>
>
> Think that’s all,
>
>
>
> Armando
>
>
>
>
>