RE: Coining a specific vocabulary for synsets in the OntoLex model

Hi all,

 

quick response on this. I’ll address tomorrow a more argumented reply on the 3-entities path regarding WordNet, as I’m in a full day meeting today.

 

Yes I agree that we should introduce a specific name in our model for Synset. 

Firstly, the modelling proposed for WordNet is based on existing modelling (i.e. lemon (1.0) and SKOS) so hence the usage of skos:Concept

As for  a new class I am not so keen on the name SemanticIndex, I would assume that the best would simply be to call it Synset, so as to ease adoption among the wider community. Semantic index I dislike as it is for associated with Latent Semantic Indexing, and in this sense more of a signature of a concept than a concept itself.

                Armando: Agree with you. I just put it as a name for that “element-in-the-middle”, but surely SemanticIndex is not really nice :-D. Still I do not like Sense or LexicalSense. I’ve anticipated something about that in the email of yesterday, however I’ll come to this in the more elaborated answer.

I don't think we should explicitly say the ontolex:LexicalForm is a skosxl:Label. In fact, the lexical form represents the orthographic union of surface forms of words, that is the same form (at least according to the lemon definition, itself based directly on the LMF definition) can have multiple strings (e.g., spelling variants, version in other writing systems, pronunciations, segmentations, etc.) unlike a SKOS-XL label.

                Armando: From what I got from the WordNet example, I supposed that we have a LexicalEntry and then each entry has several LexicalForms, but that each of these forms in turn has a single written representation. However, obviously there could be other assumptions that are out of the example, like this one that you presented. We should then discuss if we want to have this form actually represented by more strings. Also, this would clash with other models, even simple SKOS, where a spelling variant is normally associated to a different alt label. I’m not strongly voting for one or the other, but I would suggest to dedicate a separate discussion to this.

Best,

Armando