Environment: This project assumes Python 2.7 in a Unix or similar environment.

See prep/README.md

This is launched by pipeline.py. Required arguments are per-sentence JSON files generated by the preprocessing pipeline. Several examples are in the examples directory. E.g.:

$ pipeline.py examples/wsj_0001.*.json examples/wsj_0002.0.json

See the bottom of pipeline.py for options (-e for error tolerance, -v for verbose printing, etc.).