Mindswap Weblog

WWW2006 - Experiences and Episodes: Day 1(Through Taowei’s Looking Glass)

by Taowei Wang

The first two days of the conference are scheduled only for
workshops.  Both ROW (Reasoning on the Web) and EON (Evaluation of
Ontologies for the Web), where Chris and I are presenting
respectively, are held today.  I attended the first two and half sessions
of EON, giving my talk just before lunch break.  Then I went to
ROW to see Chris’s talk and the panel discussions.  Here I give the
highlights of my day.  Edinburgh is cloudy and rainy.

The room for EON was tightly packed for 30 people, as the room was very small.
As I walked into the room, Denny Vrandecic
(one of the organizers) came over
and said “David, great paper!”
I was a bit embarrassed and overwhelmed at
the comment.  I might have muttered something like “Thank you, I hope the
audience would like my talk, too.”  Yep, still rainy outside.

There were 6 presenters before me.  Most of the researchers in this
workshop come from UK, though groups from UGA, U. of Madrid, and Slovenia
do have their representatives.  Deviating from previous year’s of EON, where
the focus has been on the evaluation of ontology tools, this year’s EON
was concentrating on the actual evaluations themselves.  My survey of
ontologies was therefore a good fit.  However, my paper was unique in that
it goes into the more nitty-gritty characteristics of ontologies.

The other presentations can be characterized in a few major categories.

1) Ranking of concepts and/or ontologies
2) Ontology Evaluation principles
3) Approaches for evaluation of ontologies (machine-learning, clustering, or
statistical)

The first section tries to answer the question: “If I were to look up for
an ontology with the concept ‘Student’ from my favorite search engine, and
I got 41 hits, how would I choose which one is most suitable for me?”
Though suitability is subjective, what is assumed in this question is that
users want to find ontologies that are ‘adequate’ for their modeling purposes.
The task, then, is to find ontologies that give sufficient weight onto
the search terms.  This weight is determined, for example, by the links
a concept has to the other concepts/properties.  An potential application
is to reorder the ontologies results swoogle gives.

The second section developes of reuses an existing evaluation principle out
side of the semantics of the ontology in question.  Examples include the
well-known OntoClean methodology.  One paper presented CleanOnto, which is a
new system that uses WordNet to evaluate the taxonomic structure.

There are also a few papers using different machine learning-like approaches
as well as statistical approaches to evaluate certain parts of the ontology.
One memorable talk hailed from a group in Slovenia.  The presenter presented
a method that evaluates ontologies using a hypothetically existent golden
standard.  The method relies on a statistical model which involves constants
alpha and beta.  He explained that “alpha and beta are sort of magical, but
we have an intuitio of what they should be”.  The model seemed pretty clean,
but the tune-able parameters alpha and beta made the model a little suspect.

My talk went well.  Made it through in 15 minutes, even though presenters before me
took up all my time.  Perhaps everyone was hungry at the time, no one asked any
questions.  One other organizer came to have a brief chat with me after my talk.
He seemed to like the work :)   By this time, it’s almost noon, but one can
scarcely see the sunlight through the dark clouds.

After lunch and a couple more talks in EON, I walked with Chris to ROW and
heard two talks (including his), and a panel discussion of the logicians.  I am
sure Chris has detailed accounts on the panelists’ interesting positions, so I
won’t repeat it here.  Chris’s talk went well.  We chatted with a Southhampton
delegate along our walk across the town back.  And rain it did.  Onto tomorrow!

[pictures to follow]

3 Responses to “WWW2006 - Experiences and Episodes: Day 1(Through Taowei’s Looking Glass)”

  1. enrique Says:

    Two thoughts:

    1) Where can I access web copies of the talks you went to hear?

    2) Is there a SWOOP newsgroup or user’s mail list that we can subscribe to to help answer usability questions?

    -enrique

  2. Amy Alford Says:

    This page has the list of all the mindswap mailing lists. The swoop@lists.mindswap.org is probably what you’re looking for.

  3. David Wang Says:

    Enrique, the talks are held in WWW 2006 and each workshop may or may not contain the slides of these talks (depending on the organizers). However, you can access the list of workshops here http://www2006.org/workshops/ — which also contains the accepted papers.

Leave a Reply

MINDSWAP is a W3C member