I would appreciate any pointers to work on probabilistic POS/feature tagging
where different morphological decompositions of an ambiguous lexical item
have possibly different number of tokens/morphemes which may result in paths
from the beginning of the sentence to the end, with different lengths.
Please respond directly to me and if there is interest I will summarize
any info I get.
Sincerely
Kemal Oflazer e-mail: ko@cs.bilkent.edu.tr
Bilkent University : ko@hattusas.cs.bilkent.edu.tr
Computer Engineering Department : ko@trbilun.bitnet (BITNET/EARN)
Bilkent, ANKARA, 06533 TURKIYE www:http://www.cs.bilkent.edu.tr/~ko/ko.html
Tel: (90-312) 266-4133 fax: (90-312) 266-4126