The code implements a simple structured perceptron tagger.

Train a model:

python simpletagger.py --train TRAINFILE --test TESTFILE

Expects CoNLL input (sentences delimited by a newline, words and tag delimited by tab).

• Add your own features (see get_features method).
• Not memory efficient.
• See all options by running python simpletagger.py

Please let me know if you plan to extend/improve this code basis.