Re: Corpora: P-O-S Tagging

Dan Tufis (Dan.Tufis@xrce.xerox.com)
Wed, 26 Nov 1997 09:51:59 +0100

> In view of the recent reports on tagger evaluation I just wanted to
> wave a little flag with `QTAG' written on it: this is the recently
> released language and platform independent tagger developed here at
> Birmingham. There has not yet been a lot of `formal' evaluation,
> but for those of you who would like to hear some rough estimates:
> for Rumanian we achieved 97+% with about 45 tags,

Actually the tagset contained 73 word tags plus 8 punctuation tags.
A newer tagset doing even better contains 77 word tags (+8 for
punctuation).

Dan Tufis

> and for Swedish I have been told by a user that he reached 96.3% correct
> assignments
> with 156 tags.
>
> More information on QTAG is available on our website:
> http://www-clg.bham.ac.uk/QTAG
>
> and a previous version of the tagger can be accessed via email:
> http://www-clg.bham.ac.uk/tagger.html