TUPA 1.1

Transition-based UCCA Parser

TUPA is a transition-based parser for Universal Conceptual Cognitive Annotation (UCCA).

Requirements

• Python 3.x

• All dependencies for DyNet

Build

Create a virtual environment:

virtualenv --python=/usr/bin/python3 venv
. venv/bin/activate              # on bash
source venv/bin/activate.csh     # on csh

Install:

python setup.py install

Train the parser

Having a directory with UCCA passage files (for example, the Wiki corpus), run:

python tupa/parse.py -t <train_dir> -d <dev_dir> -c <model_type> -m <model_filename>

The possible model types are sparse, mlp and bilstm.

Parse a text file

Run the parser on a text file (here named example.txt) using a trained model:

python tupa/parse.py example.txt -c <model_type> -m <model_filename>

An xml file will be created per passage (separate by blank lines in the text file).

Pre-trained models

To download and extract the pre-trained models, run:

curl --remote-name-all http://www.cs.huji.ac.il/~danielh/ucca/{sparse,mlp,bilstm}.tar.gz
tar xvzf sparse.tar.gz
tar xvzf mlp.tar.gz
tar xvzf bilstm.tar.gz

Run the parser using any of them:

python tupa/parse.py example.txt -c sparse -m models/sparse
python tupa/parse.py example.txt -c mlp -m models/mlp
python tupa/parse.py example.txt -c bilstm -m models/bilstm

Author

• Daniel Hershcovich: [e-mail]

Citation

If you make use of this software, please cite the following paper:

@inproceedings{hershcovich2017a,
  title={A Transition-Based Directed Acyclic Graph Parser for {UCCA}},
  author={Hershcovich, Daniel and Abend, Omri and Rappoport, Ari},
  booktitle={Proc. of ACL},
  year={2017}
}

The version of the parser used in the paper is v1.0. To reproduce the experiments from the paper, run in an empty directory (with a new virtualenv):

pip install "tupa>=1.0,<1.1"
mkdir pickle models
curl -L http://www.cs.huji.ac.il/~danielh/ucca/ucca_corpus_pickle.tgz | tar xz -C pickle
curl --remote-name-all http://www.cs.huji.ac.il/~danielh/ucca/{sparse,mlp,bilstm}.tgz
tar xvzf sparse.tgz
tar xvzf mlp.tgz
tar xvzf bilstm.tgz
python -m spacy download en
python -m scripts.split_corpus pickle -t 4282 -d 454 -l
python -m tupa.parse -c sparse -m models/ucca-sparse -Web pickle/test
python -m tupa.parse -c mlp -m models/ucca-mlp -Web pickle/test
python -m tupa.parse -c bilstm -m models/ucca-bilstm -Web pickle/test

License

This package is licensed under the GPLv3 or later license (see `LICENSE.txt <LICENSE.txt>`__).