Unsupervised learning of a rule-based Spanish Part of Speech tagger
Chinatsu Aone, Kevin Hausman · 1996
This paper describes a Spanish Part-of-Speech (POS) tagger which applies and extends Brill's algorithm for unsupervised learning of rule-based taggers (Brill, 1995). First, we discuss our general approach including extensions we made to the algorithm in order to handle unknown words and parameterize learning and tagging options. Next, we report and analyze our experimental results using different parameters. Then, we describe our "hybrid" approach which was necessary in order to overcome a fundamental limitation in Brill's original algorithm. Finally, we compare our tagger with Hidden Markov Model (HMM)-based taggers.