New algorithm of hidden Markov model-based part-of-speech tagging for Chinese texts
QU Hui-ya · Journal of Northeast Normal University · 2013
One piece of primary work in hidden Markov model is calculating parameter which is often premised for part-of-speech tagging by using Viterbi algorithm.In this paper,the authors presents a new corpus training algorithm,combined with training data on binary model search using the forward and reverse bi-directional scanning method,in order to complete the expansion of the training corpus,and put forward the improvement of Viterbi algorithm.The authors adopt training corpus of different scale to test,compare and analyze testing corpus of same scale based on the result of the work in training corpus,the results show that the algorithm is feasible.