imclab / parts-of-speech-tagging Goto Github PK

........................................................................					                                                A Hidden Markov Model Based Part-Of-Speech Tagger for Hindi Language
........................................................................

Author	:		Krishna Chaitanya Chavati(chaithuzz2)

........................................................................

Instructions

........................................................................
 
1. Requires python 2.7

2. To generate the hmm model use the following command. From , Terminal go to the tagger folder and run

		count_freqs.py hindi.train > hindi.counts

3. To tag with unigrams use the following command. (Enter the sentence to be tagged in the same format as the training data 		    without tags in hindi.dev) 

		python tagger.py TAG1 hindi.counts hindi.dev > hindi.output

4. To tag with Trigram and viterbi use the following command

		python tagger.py TAG hindi.counts hindi.dev > hindi.output

5. To tag using Trigram and Rare words use the following command

	python tagger.py TAGCLASS hindi.counts hindi.dev > hindi.output  

........................................................................

Known Bugs and enhancements

........................................................................

1. Works better with language model containing low number of tags and large training data.

2. Scope for including Smoothing techniques. In this case the increase in precision is very low and is almost the same.

3. Can be made Language Independent.