Grammatical category disambiguation based on second order hidden Markov model

Grammatical category disambiguation is an important field because of its basis in many applications, for example, parsing, machine translation, phrase recognition and so on. We put forward an improved second-order hidden Markov model that can capture more context information and develop one part-of-speech tagging system based on the model. In order to reduce the number of model parameters, word equivalence classes are used. The parameters of model are achieved by the Baum-Welch algorithm using untagged text. Results show that it improves the accuracy of tagging.