A name tagger built using a MEMM classifier
We will be using the corpus from 2003 CoNLL (Conference on Natural Language Learning), which uses texts from the Reuters news service.
The training data has four items per line (tab separated): the token, the part-of-speech, the BIO chunk tag, and the name tag
Ch 8.5 in Speech and Language Processing by Jurafsky and Martin discusses the Maximum Entropy Markov Model (MEMM): https://web.stanford.edu/~jurafsky/slp3/8.pdf