Skip to content
main
Switch branches/tags
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

lambeq

lambeq logo

Build status License PyPI version PyPI downloads arXiv

About

lambeq is a toolkit for quantum natural language processing (QNLP).


Note: Please do not try to read the documentation directly from the preview provided in the repository, since some of the pages will not be rendered properly.


Getting started

Prerequisites

  • Python 3.7+

Installation

Direct pip install

The base lambeq can be installed with the command:

pip install lambeq

This does not include optional dependencies such as depccg and PyTorch, which have to be installed separately. In particular, depccg is required for lambeq.ccg2discocat.DepCCGParser.


Warning: depccg is available only on MacOS and Linux. If you are using Windows, please install the base lambeq. This means that the DepCCGParser class will not be available on Windows, but you can still use all other compositional models from the reader module. Support for parsing on Windows will be added in a future version.


To install lambeq with depccg, run instead:

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

See below for further options.

Automatic installation (recommended)

This runs an interactive installer to help pick the installation destination and configuration.

  1. Run:
    sh <(curl 'https://cqcl.github.io/lambeq/install.sh')

Git installation

This requires 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 depccg dependency, the following packages must be installed before 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},
}

About

A high-level Python library for Quantum Natural Language Processing

Resources

License

Packages

No packages published