G22.2591 - Advanced Natural Language Processing - Spring 2004

Assignment #4:  Maximum Entropy Markov Model

Due 8 am Monday March 22

This assignment builds on Assignment #3:  for #3, you (presumably) did tagging deterministically from left to right.

For assignment #4, you will build a Viterbi decoder and compute the globally most likely tagging for a sentence, using the same maximum entropy methods as before.

We will provide (hopefully, by March 8th) a second set of training / test data in similar format for name tagging.

For both the chunking and the name tagging task, compare the performance of the deterministic and Viterbi algorithms.