Implementation of a POS tagger for tweets as an excercise for CS590-NLP at Purdue University.
The goal is to compare the performance of local training (HMM and MEMM) and global training (Structured Perceptron) models on this task. I used the experimental set up of Gimpel et. al for testing.