This paper describes an experimental comparison between two standard supervised learning methods, namely Naive Bayes and Exemplar--based classification, on the Word Sense Disambiguation (WSD) problem.
In this paper we propose the type of Bayesian networks that we call the hierarchical Bayesian network (HBN) classifiers. We present algorithms for the construction of the HBN classifiers and test them on the Reuters text categorization test collection
In practical classification, there is often a mix of learnable and unlearnable classes and only a classifier above a minimum performance threshold can be deployed.