Deep Biaffine Parser implementation as in https://arxiv.org/abs/1611.01734 Since I have no access to CoNLL09 data or the Penn Treebank data, I trained it on the publicly available CoNLL17 dataset. Here I uploaded a csv version converted using https://github.com/interrogator/conll-df
Here's how to properly set up the project:
- To correctly clone the project install git lfs first https://git-lfs.github.com/
- Install pytorch 0.4 https://pytorch.org/
- Install requirements.txt by running
pip install -r requirements.txt
Run the training launching
python -m parser.train
At the moment Hyperparameters information is in the code.