Dedicated Nominal Featurization of Portuguese

A widespread assumption about the analysis of inflection features is that this task is to be performed by a tagger with an extended tagset. This typically leads to a POS precision drop due to the data-sparseness problem. In this paper we tackle this problem by addressing inflection tagging as a dedicated task, separated from that of POS tagging. More specifically, this paper describes and evaluates a rulebased approach to the tagging of Gender, Number and Degree inflection of open nominal morphosyntactic categories. This approach achieves a better F-measure than the typical approach of inflection analysis via stochastic state-of-the-art tagging.