PURPOSE:
performs cross-doc co-ref resolution for entities and events
HOW TO RUN:
within src/ there are 2 bash scripts. runLSTM_allDirs.sh is the one i invoke on a grid w/ GPUs, which submits several jobs (unique, complete runs of our program, just with different parameters per run).
to invoke the program under any other environment, just:
(1) pass all variables (including 'path') to runLSTM_1b.sh just like how we are doing within runLSTM_allDirs.sh
NOTE: 'path' should point to the base directory (i.e., wherever PredArgAlignment/ resides)
OUTPUT:
the program will output the test set's F1 performance of EVENT coref after every 5 iterations of training.
NOTE: for way more detailed information of what is going on, simply change the 'isVerbose' flag manually which resides at the top in:
- Test.py (the main entry point of the program)
- multilayer_perceptron.py (the FFNN classifier)