See my paper in the EACL SIGDAT workshop in 1995. You can get it from
the cmp-lg server as 9504002. I tried tagging English, French and
Swedish, and graually simplifying the tagset by combining similar sets
of tags. The general conclusion was that the often made assertion that
smaller tagset give higher accuracy does not hold, and that French
behaves very differently from English and Swedish.
-- David
_______________________________________________________________________
David Elworthy <dahe@cre.canon.co.uk>
Canon Research Centre Europe Ltd., Guildford, Surrey, UK
URL: http://www.cre.canon.co.uk/
Phone: +44 1483 448844; Fax: +44 1483 448845