Usage

Environment

• Python 2.7
• Java

#train
python parser.py train <train_file_name>

#predict
python parser.py predict <test_file_name>

#select features using MI
python parser.py feature

The output predict_pdtb.json is the predict result.

If you want to train from new data, delete all things under /tmp. If you just want to predict on new data, just delete all files whose name doesn't contains train or model under /tmp.

2016.05.29

• generate predict result in file, for later evaluation using scorer.py
• add command line arguments

2016.05.26

• use Lemma for word-pair feature
• add first-last-firstlast3 feature
• F1 measure up to 42.13%

2016.05.24

• parserhelper

2016.05.21

• feature extraction
• training
• predicting