[Corpora-List] WordNet vs Ontology

[Yannick Versley]

An ontology is, intuitively, a collection of facts about a set of entities.
- It may have complex logical formulae (the heavyweight kind of ontology)
- It may have logical statements of a restricted kind, as in description
logics or CycL
  This is the case that would count as prototypical for most researchers in
ontologies
- It may have semantic relations between concepts, as in DBPedia,
ConceptNet or the various "extended" wordnets
  This is the prototypical case for people from the "semantic web" community
- It may have a high coverage of taxonomical relations, with selected
non-taxonomical relations added
  This is the typical case for wordnets (i.e. Princeton WordNet, WOLF,
GermaNet, etc.)

As such, wordnets make rather lightweight ontologies, but they are more
ontology-like than, say,
thesauri or machine-readable dictionaries.

What makes WordNet useful is that you have information attached to each
concept that relates it to things
that are external to the ontology - word lemmas in WordNet, or images in
ImageNet. Without this additional
information, you would just have a very big tree-ish graph, which is why
WordNet would count as trivial and uninteresting
for some ontology researchers. On the other hand, the more complex
ontologies are typically small, and
usually are limited to some domain or an upper-level ontology.

The taxonomic information in wordnets is quite useful, but e.g. for
graph-based WSD you cannot use only
taxonomic links. Hence, people working on graph-based WSD have added
nontaxonomic links, sometimes
without a specific interpretation as a semantic relation (i.e., something
that can be interpreted as a claim
about instances of these concepts), which again would be weird if you only
looked at the ontology part and
not at the whole "words -> concepts -> approximate semantics" story.

So, in summary, WordNet and other wordnets could be adequately described as
machine-readable dictionaries with
an ontological component, but saying they are ontologies would do injustice
to both WordNet and ontologies.

To quote Miller et al (1993: Introduction to WordNet: An On-line Lexical
Database):
`` The initial idea was to provide an aid to use in searching dictionaries
conceptually, rather
than merely alphabetically—it was to be used in close conjunction with an
on-line
dictionary of the conventional type. As the work proceeded, however, it
demanded a
more ambitious formulation of its own principles and goals. ''

Eventually, those wanting to provide conceptual underpinnings for
machine-readable dictionaries
and those who want to link ontologies to texts found out that they pursue
the same long-term goals,
althought maybe from a different starting point, and even now with some
differences in short-term goals.

Best wishes,
Yannick



On Wed, Aug 6, 2014 at 11:57 AM, liling tan <alvations at gmail.com> wrote:

> Dear corpora linguists,
>
> There is recently a discussion on stackoverflow about "wordnet vs
> ontology". I would like your perspective on several issues about wordnet
> and ontology:
>
> - Is wordnet an ontology? If it is not an ontology, what is it?
>
> - What is the definition of an ontology? Is anything
> (words/concept/entities) under a hierarchical structure some sort of
> linguistic ontology?
>
> - Are linguistic onotology / information science ontology subjected to
> only upper and domain ontology?
>
> - Any other comments about ontology and wordnet?
>
> Regards,
> liling
>
>
>
> _______________________________________________
> UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
> Corpora mailing list
> Corpora at uib.no
> http://mailman.uib.no/listinfo/corpora
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/corpora/attachments/20140806/36a9cb88/attachment.htm>
-------------- next part --------------
_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora