[Corpora-List] semantic similarity

[Aristomenis Thanopoulos]

Dear Adam and All

That's true of course for English and a few more languages but not at all
for the majority of contemporary languages for which syntactic parsers are
still unreliable or simply non-existent...

Aris Thanopoulos
University of Patras, Greece

-----Original Message-----
From: owner-corpora at lists.uib.no [mailto:owner-corpora at lists.uib.no] On
Behalf Of Adam Kilgarriff
Sent: Thursday, January 20, 2005 10:32 PM
To: 'Jana Diesner'; CORPORA at hd.uib.no
Subject: RE: [Corpora-List] semantic similarity

Jana

> The approach must not require POS tagging

whyever not? Ten years ago there was an excuse for ignoring syntax (no
tools, too slow to run over big corpora, expensive) but I don't think there
is any more.

You get much better results if you respect syntax (see eg thesaurus at
www.sketchengine.co.uk which shallow-parses and uses Dekang Lin's
similiarity measure)


Adam Kilgarriff