Evaluating Solutions for the Rapid Development of State-of-the-Art POS Taggers for Portuguese

Proceedings of the Fourth International Conference on Language Resources and Evaluation (LREC 2004)

Abstract

We report on solutions we adopted for the specific issues that arise when developing new automatic taggers for Portuguese, solutions whose design is general enough, we believe, to be further reused to develop other new taggers for this language, even when using different training data than those we used in our experiments. We report also on the evaluation of tools that make use of such solutions and show that the latter permit to develop POS taggers for Portuguese whose performance matches or surpasses state-of-the-art results obtained for other languages when using the same technology.