Evolving Text Classifiers with Genetic Programming

Abstract

We describe a method for using Genetic Programming (GP) to evolve document classifiers. GP's create regular expression type specifications consisting of particular sequences and patterns of N-Grams (character strings) and acquire fitness by producing expressions, which match documents in a particular category but do not match documents in any other category. Libraries of N-Gram patterns have been evolved against sets of pre-categorised training documents and are used to discriminate between new texts. We describe a basic set of functions and terminals and provide results from a categorisation task using the 20 Newsgroup data. © Springer-Verlag 2004.