Parsing Data Generation for WSJ
April 24, 2019 ยท View on GitHub
- Place a copy of the Penn Treebank
(LDC99T42) in
corpora/treebank_3. After doing this,corpora/treebank_3/parsed/mrg/wsjshould have folders named00-24. - Ensure that the active version of Python has
nltkinstalled. Python 2.7 and Python 3 should both work. cd corpora/wsj && ./build_corpus.sh
Processed trees are written to the following files:
corpora/wsj/train.gold.strippedcorpora/wsj/dev.gold.strippedcorpora/wsj/test.gold.stripped