TUPA 1.2.4

Transition-based UCCA Parser

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

Requirements

• Python 3.x

• All dependencies for DyNet

Install

Create a Python virtual environment:

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

Install the latest release:

pip install tupa

Alternatively, install the latest code from GitHub (may be unstable):

git clone https://github.com/danielhers/tupa
cd tupa
python setup.py install

Train the parser

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

python -m tupa.parse -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 -m tupa.parse 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 models pre-trained on the Wiki corpus, run:

curl --remote-name-all http://www.cs.huji.ac.il/~danielh/ucca/{sparse,mlp,bilstm}-1.2.tar.gz
tar xvzf sparse-1.2.tar.gz
tar xvzf mlp-1.2.tar.gz
tar xvzf bilstm-1.2.tar.gz

Run the parser using any of them:

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

Other languages

To get a French/German model pre-trained on the *20K Leagues* corpus, run:

curl -O http://www.cs.huji.ac.il/~danielh/ucca/sparse-1.2-{fr,de}.tar.gz
tar xvzf sparse-1.2-fr.tar.gz
tar xvzf sparse-1.2-de.tar.gz

Run the parser on a French/German text file, using the French/German spaCy models too:

export SPACY_MODEL=fr_depvec_web_lg
python -m tupa.parse exemple.txt -c sparse -m models/sparse-fr

export SPACY_MODEL=de_core_news_md
python -m tupa.parse beispiel.txt -c sparse -m models/sparse-de

Author

• Daniel Hershcovich: [e-mail]

Citation

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

@InProceedings{hershcovich2017a,
  author    = {Hershcovich, Daniel  and  Abend, Omri  and  Rappoport, Ari},
  title     = {A Transition-Based Directed Acyclic Graph Parser for UCCA},
  booktitle = {Proc. of ACL},
  year      = {2017},
  pages     = {1127--1138},
  url       = {http://aclweb.org/anthology/P17-1104}
}

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>`__).