Hybrid techniques for training HMM part-of-speech taggers
Ted Briscoe, Gregory Grefenstette, Lluís Padró, Serail Iskander · LA Referencia (Red Federada de Repositorios Institucionales de Publicaciones Científicas) · 1996
We describe and experimentally evaluate a hybrid technique for training part-of-speech taggers which utilises training from small quantities of unambiguously-tagged material combined with maximum likelihood re-estimation over the target untagged corpus. This approach, unlike previous ones employing re-estimation, does not involve skilled manipulation of the initial parameters of the model or the use of sophisticated models of suffix-tag probabilities derived form unambiguously-tagged material. We conclude that this technique can yield usefully accurate taggers for several languages, but that the conditions required for success are difficult to state precisely.