IIIT Hyderabad Publications |
|||||||||
|
Improving statistical POS tagging using Linguistic feature for Hindi and TeluguAuthors: Phani Gadde,Meher Vijay Yeleti Conference: ICON-2008: International Conference on Natural Language Processing (ICON-2008 2008) Date: 2008-12-20 Report no: IIIT/TR/2008/189 AbstractIn this paper we describe some strategies for improving statistical POS tagging us-ing Hidden Markov Models (HMM) for Hindi and Telugu. We describe how add-ing features to HMM improves its accu-racy. We also describe a method for ef-fective handling of compound words in Hindi. Experiments show that GNP1 and category information of a word are cru-cial in achieving better results. The max-imum accuracy achieved with HMM based approach is 92.36% for Hindi and 91.23% for Telugu. We achieved an im-provement of 1.85% in Hindi and 0.72% in Telugu over the previous methods. Full paper: pdf Centre for Language Technologies Research Centre |
||||||||
Copyright © 2009 - IIIT Hyderabad. All Rights Reserved. |