Author: Md Kamrul Hasan Email: [email protected] Date: 9/7/2017
================================================================================================== Description: Implementation of the HMM decoder for Parts of Speech Tagging
I did it as a part of homework problem in the Statistical Speech and Language Processing class taught by Prof Daniel Gildea (https://www.cs.rochester.edu/~gildea/) in Fall 2017.
================================================================================================== Instruction to run:
python viterbi_final.py train test
It takes around 8 minutes to complete
==================================================================================================
Preprocessing: I have preprocessed all the transition probability, emission probability and intitial state distribution in preprocessing step. I used log technique to avoid floating precision problem. I also used smoothing tehcnique to handle unseen tag or word in test file.
Accuray: 94 %