Date: Tue, 7 Dec 1999 16:59:00 GMT
From: Yorick Wilks <yorick>
To: yorick@dcs.shef.ac.uk
Subject: Re: Corpora: NEW SENSES SUMMARY - REVISED
Thanks, but I wasnt trying to tell you
Id done a lot of work on sense distinctions in many
etc. but to point out some specific early work on NEW SENSES (not sense distinctions) and not just my own!
PLEASE dont recirculate any redescriptions of my mail--you'll find it irritates people to do that--thanks
YW
----- End Included Message -----
----- Begin Included Message -----
Date: Tue, 07 Dec 1999 17:33:47 +0100
From: "Dimitris K." <svedk@svenska.gu.se>
X-Accept-Language: en
MIME-Version: 1.0
To: corpora@hd.uib.no, svedk@svenska.gu.se
Subject: Corpora: NEW SENSES SUMMARY - REVISED
Content-Transfer-Encoding: 7bit
I have got tenths of mails regarding a complete summary
on "New Senses" - there is HUGE interest out there!
However, there are some reasons why I DID NOT posted a complete summary
on the "New Senses" query earlier, people are overloaded by requests
and they have *explicitly* stated that I do NOT post their answers
further,
projects are confidential or ongoing, etc.
Nevertheless, here comes a LIMITED list of the answers I received-
please contact these people YOURSELF if you need more information.
*******************************************************
Paul Buitelaar <paulb@dfki.de>
.... did some work on that for his PHD, besides the
CoreLex work. "In fact what he did was, collect statistical models on
CoreLex classes (from training corpora: Brown and Wall Street Journal in
this case) and classify words as they occur in corpora according to
these
models. Words will then fall into classes ('senses') that they
originally
belong to (acccording to CoreLex, or whatever classification you use),
but
also sometimes in new classes (again 'senses') because their usage in
this
particualr corpus is rather different from that of the training corpus.
This may be either a sense that was not represented in the training
corpus, or a genuinly new sense that this word did not have in the
CoreLex
database."
-------------------------------------------
Adam.Kilgarriff <Adam.Kilgarriff@itri.brighton.ac.uk>
Adams gave a tip to look at the work by Lin
http://www.cs.umanitoba.ca/~lindek/
Dekang Lin' address - he had a relevant paper on ACL a couple of years
ago
-------------------------------------------
***Benjamin Kjeldsen <bek.eng@cbs.dk>
there is some work towards that direction within the SENSUS project
more info ....is confidential
-------------------------------------------
Bob Krovetz <krovetz@research.nj.nec.com>
wrote a paper about that entitled "Learning to Augment a
Machine-Readable
Dictionary". "he recognized new senses (relative to the Longman
dictionary)
based on comparisons with two test collections used in information
retrieval.
You can get a copy of the paper from:
external.nj.nec.com/homepages/krovetz/publications.html"
-------------------------------------------
Ken Litkowski <ken@clres.com>
replied on the list
-------------------------------------------
Ted Pedersen <tpederse@d.umn.edu>
Have done a bit of work like this - check out
the following paper, available on his web site (follow the
research papers link)
Distinguishing Word Senses in Untagged Text (Pedersen & Bruce) - Appears
in the Proceedings of the Second Conference on Empirical Methods in
Natural Language Processing (EMNLP-2), August 1-2, 1997, Providence, RI
(Also available from CMP-LG E-Print Archive as #9706008)
-------------------------------------------
Sylvain Surcin <surcin@lcr.thomson-csf.com>
wrote a PhD thesis on that in French
-------------------------------------------
Yorick Wilks <y.wilks@dcs.shef.ac.uk>
has done a lot of work on sense distinctions described in many
papers
-------------------------------------------
Some of the work in the ECRAN project has also relevance
http://www.dcs.shef.ac.uk/research/ilash/Ecran/deliverables.html
-------------------------------------------
Hope this makes any 'sense' (as Paul said)
best
Dimitris
----- End Included Message -----