Recognizing NPs with an Ensemble of Classifiers

Erik F. Tjong Kim Sang (University of Antwerp)

An ensemble of classification algorithms can improve the performance
of its members by avoiding errors which are made by a minority of the
individual classifiers. We apply this idea to noun phrase recognition.
The classifier ensemble is created by processing different
representations of the data with a single machine learning algorithm. 
The results are combined with voting techniques such as described in
(Van Halteren et al. 1998). The combined system improves the best 
performances known to us for standard data sets for base noun phrases 
and arbitrary noun phrases.