Tagging with Disambiguation Rules - A New Evolutionary Approach to the Part-of-Speech Tagging Problem

Abstract

In this paper we present an evolutionary approach to the part-of-speech tagging problem. The goal of part-ofspeech tagging is to assign to each word of a text its part-of-speech. The task is not straightforward, because a large percentage of words has more than one possible part-of-speech, and the right choice is determined by the surrounding word's part-of-speeches. This means that to solve this problem we need a method to disambiguate a word's possible tags set. Traditionally there are two groups of methods used to tackle this task. The first group is based on statistical data concerning the different context's possibilities for a word, while the second group is based on rules, normally designed by human experts, that capture the language properties. In this work we present a solution that tries to incorporate both these approaches. The proposed system is divided in two components. First, we use an evolutionary algorithm that for each part-of-speech tag of the training corpus, evolves a set of disambiguation rules. We then use a second evolutionary algorithm, guided by the rules found earlier, to solve the tagging problem. The results obtained on two different corpora are amongst the best ones published for those corpora.