A Discriminative Model for Traditional Mongolian Part of Speech

This paper presents a discriminative model for part of speech tagging of traditional Mongolian.We use Maximum Entropy Model with Morphological features of Mongolian. First, the context feature templates are defined and extracted from the training corpus. Then, the parameters of maximum entropy probability models are calculated. Experimental results show that integration of morphological features of Maximum Entropy Model for Mongolian part of speech tagging outperform HMM since they are flexible enough to capture many correlated non-independent features.