pip install lambeq

pip install cython numpy
pip install lambeq[depccg]
depccg_en download

Git installation

This required Git to be installed.

1. Download this repository:

   git clone https://github.com/CQCL/lambeq

2. Enter the repository:

   cd lambeq

3. Make sure pip is up-to-date:

   pip install --upgrade pip wheel

4. (Optional) If installing the optional dependency depccg, the following packages must be installed before installing depccg:

   pip install cython numpy

   Further information can be found on the depccg homepage.

5. Install lambeq from the local repository using pip:

   pip install --use-feature=in-tree-build .

   To include depccg, run instead:

   pip install --use-feature=in-tree-build .[depccg]

   To include all optional dependencies, run instead:

   pip install --use-feature=in-tree-build .[all]

6. If using a pretrained depccg parser, download a pretrained model:

   depccg_en download

Usage

The docs/examples directory contains notebooks demonstrating usage of the various tools in lambeq.

Example – parsing a sentence into a diagram (see docs/examples/ccg2discocat.ipynb):

from lambeq.ccg2discocat import DepCCGParser

depccg_parser = DepCCGParser()
diagram = depccg_parser.sentence2diagram('This is a test sentence')
diagram.draw()

Note: all pre-trained depccg models apart from the basic one are broken, and depccg has not yet been updated to fix this. Therefore, it is recommended to just use the basic parser, as shown here.

Testing

Run all tests with the command:

pytest

Note: if you have installed in a virtual environment, remember to install pytest in the same environment using pip.

Building Documentation

To build the documentation, first install the required dependencies:

pip install -r docs/requirements.txt

then run the commands:

cd docs
make clean
make html

the docs will be under docs/_build.

To rebuild the rst files themselves, run:

sphinx-apidoc --force -o docs lambeq

License

Distributed under the Apache 2.0 license. See LICENSE for more details.

Citation

If you wish to attribute our work, please cite the accompanying paper:

@article{kartsaklis2021lambeq,
   title={lambeq: {A}n {E}fficient {H}igh-{L}evel {P}ython {L}ibrary for {Q}uantum {NLP}},
   author={Dimitri Kartsaklis and Ian Fan and Richie Yeung and Anna Pearson and Robin Lorenz and Alexis Toumi and Giovanni de Felice and Konstantinos Meichanetzidis and Stephen Clark and Bob Coecke},
   year={2021},
   journal={arXiv preprint arXiv:2110.04236},
}

GitHub

View Github