[ontolex] yet another "final" proposal for Vartrans

Dear all,

After some internal discussions we are proposing (Elena, Lupe, and myself)
the following to accommodate the new notion of a generic lexico-semantic
relation in Lemon-Ontolex:

1) To add a generic "Lexico-semantic Relation"  to the CORE module (the
name can be another one of course) for reifying relations whenever it is
necessary, and having the "source" and "target" properties that we have
discussed in the previous telcos. It is specialised in "Lexical Relation"
whenever the relation is at the level of forms or lexical entries, and
"Sense Relation" whenever the relation is between senses. The motivation of
placing these in the core is to support ANY module that wants to reuse the
same reification mechanisms.

[image: Imágenes integradas 1]

2)  Based on those classes, any type of specific relation could be
supported. Now it is our choice whether to include them as a lemon module
or as something external to the model. For instance, new relations could be
defined: "Morphological derivation" (e.g., happy -> unhappy) at the lexical
level or "Antonymy" at the sense level, for instance.


[image: Imágenes integradas 2]

3-A) Vartrans module: the Variant relation is proposed as a specialisation
of "Lexico-semantic Relation". Roughly speaking, we understand "Variant" as
a relation between two entities (lexical entries, forms, senses, ...) that
are interchangeable under certain conditions, still keeping similar
meanings (e.g., "finger" -> "hand" are not variants, but "color" ->
"colour" or "bank"@en -> "banco"@es are variants). There are several
subclasses of Variant:  "Lexical Variant" and "Translatable" are variants
that are lexical relations as well, and "Terminological Variant" and
"Translation" are variants that are also sense relations. See the figure
for some examples...

[image: Imágenes integradas 3]

3-B) However one could say that any possible variant is covered already by
the above referred subclasses "Lexical Variant", "Translation",
"Terminological Variant", ... So another scheme without the "Variant" type
is also possible, although loosing the notion of a relation between
entities that are exchangeable under certain circumstances (which is given
by "Variant").

  [image: Imágenes integradas 4]


I hope you find this reflections useful. Sorry we are not available today
for discussing it, but any feedback or comment by email will be
appreciated, and we can continue the discussion in the following telco.


Best regards,

Jorge, Elena, Lupe


-- 
Jorge Gracia, PhD
Ontology Engineering Group
Artificial Intelligence Department
Universidad Politécnica de Madrid
http://jogracia.url.ph/web/