RE: Issues about the semantics of the ontology-lexicon interface [was: Re: Why not to shortcut the "sense" object]

A reply on the other points discussed so far.

In bold (not bald J ), the main sentence about my preferences. The rest is
motivations and rationale behind them (in my PoV). The use of terms such as
Sense, representedBy etc.. is not definitive. I'm using the ones currently
most used, and then provide my candidate ones.

 

---Creating the sense class---

Concerning Philipp's summary, firstly I agree with the decision (?not yet
approved, it seems?) of creating the intermediate Sense class: it's
obviously needed, either for making room for lexical senses (definitely to
be distinguished from ontology entities), or to be able to talk about senses
(reifications of the meaning function).

 

+1 from me (though it seems we all agree). In terms of motivation, I've seen
referring much debate about whether they are useful or not to define
something not representable by an ontology model (e.g. the discussion about
"Meaning and Semiotics - Issues for Modelling" of this August, and the
question: "Is there any semantic aspect of a word sense that cannot be
represented in an ontological model?" ). I don't think this is the main
motivation we may want to address.

While the above is interesting, I think that a mere direct connection
(OntologyEntry <-> LexicalEntry) would not be inappropriate if we think the
ontology as the starting point. In this sense, we (in my humble opinion)
shouldn't care if there's "something about the meaning of a word that we
cannot represent with an ontology model" (I don't delve into the matter of
whether this is true or false) as from our mission, we are defining an
ontology-lexicon interface, and not a linguistic model per se. The ontology
(to be better "linguistically represented") is the cause for our work, and
not the effect. Thus if there's a conjecture about some "dark matter" in the
coverage of ontologies wrt some "word sense", it is out of scope.

I would instead add a practical/social motivation: if we are thinking about
the LOD, we would think about lot of RDFized lexical resources (e.g. I would
like to see all multilingual wordnets publicly released as LOD, but that's
not the case at the moment. and also there are many more resources that are
or will be released as LOD: Framenet, Verbnet, Senso Comune etc..). They
provide lot of common sense which is not committed to any particular
ontology. All of these resources expose, to some extent, some sort of
"semantic layer" (I use this vague name by purpose, as I want to include all
potential different theories and models, and just want to state that they
are not just collections of terms/labels). Now the point is: if we want to
reuse explicitly these resources, we need to address their semantic layer.

The entities in such a non-ontology-committed semantic layer (e.g. synsets
in wordnet) would be a potentially useful lingua franca in many
mapping/matching scenarios, because they are still not a (domain) ontology
resource (thus with very easy to express, non-committing, relations between
ontology and lexicon), though they have the power to be much less ambiguous
than pure language. Their high shareability would act as a informal glue
between ontologies, with high potentials for a new generation of matchers
which are trained to expect this "formally stated informal content".

On the other side, having the possibility to state connections, for
instance, between semantic roles of a frame and elements in an ontology
under a shared vocabulary, would allow the creation of computational objects
and greatly support their distribution as LOD resources themselves and their
reuse in heterogeneous applications, as these will based on: a shared
ontological resource, a shared linguistic resource (expressed in a standard
RDF vocabulary) and the links between them (again expressed in a shared
vocabulary).

I would like to see patterns like the one depicted in:

http://art.uniroma2.it/publications/docs/2008_SWAP2008_LinguisticWatermark3.
0.pdf (upper part of pag. 8)

and the idea of an abstract vocabulary of "extension points" where different
lexical resources can easily fit.

After all, we too have a use case, that is: LLD (lexical linked data). So,
by simple modus ponens, I would say that if the LLD use case is valid, then
we necessarily need that Sense element, and in case, we should find space
for connecting it to ontology entities!

 

Concerning the name, I vote for "sense", because sememes, acceptations, and
others, are either very technical for the layman, or even wrong, as Philipp
reminds us about the original notion of sememe. The only real alternative
would be "meaning", but I'd rather keep that term for the top-level class of
a meaning taxonomy, as I suggest in the following.

 

Naming of elements in the PATH

I don't like "Sense" that much, for two reasons:

1)      To me (but I've not a strong background in linguistics) sense evokes
the idea of a relationship. I feel like I cannot speak of a sense without a
word. E.g. In wordnet, we may say that the first sense of "runner" is
"someone who imports or exports without paying duties" and that the second
is "someone who travels on foot by running" and this would be correct. But
these senses, as entities, are called synsets in wordnet and these, and only
these, are the semantic units adopted in wordnet. A synset plays the role of
a "sense" for a given word (or set of). Or better, the sense of a word is a
link to a synset (which, practically, in Wordnet is more than just a set of
words, and more close to a referent of the world).

2)      +1 on the idea of a Taxonomy, I did the same on previous works on
the matter, and was proposing the same to do here. In this sense (erm,
pardon the abuse of the word!), I think Sense is too much constraining what
we could put here.

I liked Sememe, as it seemed quite abstract to fit, but by reading the
definition, it seems not appropriate (it is atomic).

Was thinking about Meaning, though this world is not clearly used in
literature (e.g. think about its translation used alternatively in different
translations for both terms in the "Sinn and Bedeutung" dichotomy, but this
is not the only case.).

 

I'm more oriented towards a very abstract particle, such as
SemanticUnit/Index/Entry. I'm not really sure about Semantic Field or
Semantic Class, which have been already defined in linguistics. If they fit
this level of abstraction (e.g. would synset be a semantic field (/class?),
where the shared semantic property is dictated by the Leibniz's principle of
meaning invariance wrt lexical substitution?), then they may go even better.

 

Also, I think that we have to really state what we expect (intension and
extension, through examples) what we would like to see "hosted" by this
class. Though I suggested to come as soon as possible to some terminology,
to have a more agile talking, I think we may coin a temporary terminology
(for the said reason of ease-of-speak), check what we want to define with
it, and then verify if it still covers what we need. Rem tene.verba
sequentur (it is rumoured that someone heard Cato murmuring "hopefully" :-D
).

 

Concerning the property names, I'm ok with both LexicalEntry - meaning ->
Sense, and with Sense - representedBy -> OntologyEntity. 

Maybe we could get rid of multiple related uses of the "mean" notion, which
can be somehow disturbing: Meaning as a class, meaning as a property between
lexical entries and senses, means as a property between lexical entries and
ontology entities . it may look like we are playing with words . what about
following the conventional naming patterns that employs the name of the
property range? E.g. LexicalEntry - sense -> Sense ; LexicalEntry -
meansOntologyEntity -> OntologyEntity. The advantage of using this
apparently redundant naming is that at the instance level, the triple become
very clear, e.g. Saxophone - sense -> wordsense-saxophone-1 ; Saxophone -
hasOntologyEntity -> music:Saxophone.

I also prefer "representedBy" to "characterizes", because the second is very
generic and not attested in any related literature. 

 

Lexical Entry.is it not Lexical Unit more appropriate? An entry exists
somewhere, and an object may appear as an entry in more "wheres". A unit
exists per se. Again, pardon my ignorance if I touched some already existing
definition which is not appropriate.

I really like the 

 

 

---Property chaining over senses---

 

I cut it short: totally agree on the property chain, and with perfect
overlap with all of the motivations provided by Aldo (thanks for saving me
more words to write, and for expressing them much better than I would have
done :-) ).

 

Cheers,

 

Armando