Skip to main content

Neural coreference resolution

Project description

DeCOFre

Detecting Coreferences for Oral French¹.

This was developed for application on spoken French as part of my PhD thesis, it is relatively easy to apply it to other languages and genres, though.

Installation

  1. Install with pip (Using the --pre flag until allennlp release a stable 0.9.1, at which point we'll also put a release on pypi)

    python -m pip install --pre git+https://github.com/LoicGrobol/decofre
    
  2. Install the additional dependencies

    python -m spacy download fr_core_news_sm
    

Running a pretrained model

Use decofre-infer, e.g.

decofre-infer path/to/detector.model path/to/coref.model path/to/raw_text.txt

Its output is still rather crude and mostly meant for demonstration purpose.

Training a model

Downloading ANCOR

So far the only corpus we officially support (more in preparation, along with an easier preparing procedure).

  • Clone this repo git clone https://github.com/LoicGrobol/decofre && cd decofre
  • Ensure you are in an environment where DeCOFre has been installed (to be sure that all the dependencies are correct)
  • Run the bootstrap script doit run datasets/ancor/ancor.py

Actual training

Use decofre-train, e.g.

decofre-train --config tests/sanity-check.jsonnet --model-config decofre/models/default.jsonnet --out-dir /path/to/an/output/directory

This will put a detector.model and a coref.model files in the selected output directory, that you can then load in decofre-infer.

You probably want to substitute the config files for your own, see also ANCOR config files in datasets/ancor/. The config files are not really documented right now, but you can take inspiration from the provided examples. See also decofre-train --help for other options.

This is by no mean fast, you have been warned.

Citation

@inproceedings{grobol2019NeuralCoreferenceResolution,
  author = {Grobol, Loïc},
  date = {2019-06},
  eventtitle = {Proceedings of the {{Second Workshop}} on {{Computational Models}} of {{Reference}}, {{Anaphora}} and {{Coreference}}},
  pages = {8-14},
  title = {Neural {{Coreference Resolution}} with {{Limited Lexical Context}} and {{Explicit Mention Detection}} for {{Oral French}}},
  url = {https://www.aclweb.org/anthology/papers/W/W19/W19-2802/},
  urldate = {2019-06-24}
}

1. Let me know if you think of a better name.

Licence

Unless otherwise specified (see below), the following licence (the so-called “MIT License”) applies to all the files in this repository. See also LICENCE.md.

Copyright 2020 Loïc Grobol <loic.grobol@gmail.com>

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and
associated documentation files (the "Software"), to deal in the Software without restriction,
including without limitation the rights to use, copy, modify, merge, publish, distribute,
sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or
substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT
NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT
OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Licence exceptions

The files listed here are distributed under different terms. When redistributing or building upon this work, you have to comply with their respective restrictions separately.

ANCOR

CC-BY-NC-SA-4.0 badge

The following files are derived from the ANCOR Corpus and distributed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for decofre, version 0.5.0.dev2
Filename, size File type Python version Upload date Hashes
Filename, size decofre-0.5.0.dev2-py3-none-any.whl (78.4 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size decofre-0.5.0.dev2.tar.gz (66.5 kB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page