Hello!

I've got to know your work at the Cicling 2003, now I see that it has been
continued in a pretty nice way, putting it all together. This is a very
good idea because this shortens the long way between a scientific
something and a practical use and gives other scientists the possibility
to take this as a baseline, thus increasing the overall quality.

I have one comment on the methodology, SVD & co. is only one way to
calculate semantical relations or similarity. Sentence-based co-occurrence
measures might actually provide comparable dimensionality reduction for
many reasons like seamless zooming, not having to calculate huge matrices
(thus better time complexity) and some more. Based on this something
like senseclustering can be done as well. Some experiments, clustering
words by senses using this on two corpora german and english can be seen
on that site:

http://wortschatz.uni-leipzig.de/~sbordag/
(if it works - if not please email me)

Maybe this approach could be added to SenseClusters but it would need
reimplementation since it was only done as a proof-of-concept.

Best regards,
Stefan Bordag

-- 
---------------------------------------------------------------------
- Bordag Stefan, [EMAIL PROTECTED]                  -
- Institut fuer Informatik, Abt. Automatische Sprachverarbeitung    -
- Universitaet Leipzig                                              -
---------------------------------------------------------------------



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
_______________________________________________
senseclusters-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/senseclusters-users

Reply via email to