[Next] [Up] [Previous]
Next: Principles of P-O-S Tagging Up: Probabilistic Tagging in Previous: Probabilistic Tagging in

Introduction

This document describes a tagging experiment where a probabilistic tagger developed for English has been made language independent in order to work on Romanian data. The results of the tagging are very encouraging, with an error rate below 3%. In order to make those statistics more meaningful, a metric is proposed which allows more accurate comparisons between tagging results. The software described in this article is freely available.



Multext-East