Unknown word processing in HMM-based POS tagging

Ambiguity of POS (part of speech) is a very important ambiguous phenomenon in natural language processing, which needs to be resolved urgently. Especially, it is a very difficult task to disambiguate the POS ambiguity of the unknown words. In this paper, through converting the job of POS tagging the unknown words to computation of the emission probability of the unknown words, a new POS tagging approach based on HMM (Hidden Markov Model) is proposed. This method preferably solved the challenge of processing unknown words in POS tagging. Compared with other HMM-based POS tagging approach, our approach improved the average accuracy of POS tagging by 1%, and gained an accuracy of 97% in close test and an accuracy of 95% in open test. The positive test result confirmed the method's validity.