Automatic correction of part-of-speech corpora
Uwe D. Reichel, Lia Saki Bučar Shigemori · Open access LMU (Ludwid Maxmilian's Universitat Munchen) · 2008
In this study a simple method for automatic correction of part-ofspeech corpora is presented, which works as follows: Initially two or more already available part-of-speech taggers are applied on the data. Then a sample of differing outputs is taken to train a classifier to predict for each difference which of the taggers (if any) delivered the correct output. As classifiers we employed instance-based learning, a C4.5 decision tree and a Bayesian classifier. Their performances ranged from 59.1 % to 67.3 %. Training on the automatically corrected data finally lead to significant improvements in tagger performance.