Part-of-Speech Tagging Using the Brill Method

Part-of-speech tagging is the process of associating each word in a text with it’s part-of-speech category and possibly a set of morphosyntactic features. This information is represented by part-of-speech tags. This paper describes an implementation of a part-of-speech tagger for Swedish based on the Brill method. The basic idea is to apply a set of rules to an initial annotation achieved using a simple algorithm. The rules are found using transformation-based learning applied to a manually tagged training corpus. The paper also addresses the problem of tagging unknown words, i.e. words that don’t appear in the training corpus.