The reference implementation for the papers
Read, Tag, and Parse All at Once, or Fully-neural Dependency Parsing Jan Chorowski, Michal Zapotoczny, Pawel Rychlikowski (arxiv draft)
and
On Multilingual Training of Neural Dependency Parsers Michal Zapotoczny, Pawel Rychlikowski, Jan Chorowski (arxiv draft, accepted into the TSD 2017).
- install all the dependencies (see the list below)
- set your environment variables by calling
source env.sh
- download pre-trained models
- run one of the models e.g.
python bin/parse.py universal/baseline/universal_pl.zip
will run an interactive parser for polish model trained on Universal Dependencies v1.3 data.
For more details please proceed to exp/dependency
.
An online demo is available at http://zapotoczny.pl/parser
- Python packages: pykwalify, toposort, pyyaml, numpy, pandas, progressbar, picklable-itertools, segtok
MIT
This code is based on heavily modified End-to-End Attention-Based Large Vocabulary Speech Recognition