Renard

Renard (Relationship Extraction from NARrative Documents) is a modular library for creating and using custom character networks extraction pipelines. Renard can extract dynamic as well as static character networks. Renard is modular, in the sense that you can easily create a custom extraction pipeline that fits your needs.

Installation

Currently, Renard supports Python>=3.10,<=3.14. You can install the latest version using pip:

pip install renard-pipeline

By default, this pulls the cuda version of pytorch. If you want a CPU-only version of torch you should do:

pip install torch --index-url https://download.pytorch.org/whl/cpu && pip install renard-pipeline

For an AMD rocm version of torch (ROCm 7.1):

pip install torch --index-url https://download.pytorch.org/whl/rocm6.4 && pip install renard-pipeline

Documentation

Documentation, including installation instructions, can be found at https://compnet.github.io/Renard/

If you need local documentation, it can be generated using Sphinx. From the docs directory, make html should create documentation under docs/_build/html.

Interactive Demo

You can check the interactive demo of Renard at HuggingFace. The demo uses the UI integrated in Renard.

Tutorial

Renard's central concept is the Pipeline.A Pipeline is a list of PipelineStep that are run sequentially in order to extract a character graph from a document. Here is a simple example:

from renard.pipeline import Pipeline
from renard.pipeline.tokenization import NLTKTokenizer
from renard.pipeline.ner import NLTKNamedEntityRecognizer
from renard.pipeline.character_unification import GraphRulesCharacterUnifier
from renard.pipeline.graph_extraction import CoOccurrencesGraphExtractor

with open("./my_doc.txt") as f:
	text = f.read()

pipeline = Pipeline(
	[
		NLTKTokenizer(),
		NLTKNamedEntityRecognizer(),
		GraphRulesCharacterUnifier(min_appearance=10),
		CoOccurrencesGraphExtractor(co_occurrences_dist=25)
	]
)

out = pipeline(text)

For more information, see renard_tutorial.py, which is a tutorial in the jupytext format. You can open it as a notebook in Jupyter Notebook (or export it as a notebook with jupytext --to ipynb renard-tutorial.py).

Contributing

see the "Contributing" section of the documentation.

Running tests

Renard uses pytest for testing. To launch tests, use the following command :

uv run python -m pytest tests

Alternatively, the project Makefile has a test target:

make test

Expensive tests are disabled by default. These can be run by setting the environment variable RENARD_TEST_SLOW to 1.

Renard UI

Since version 0.7, Renard has a web interface powered by gradio. First, install the additional dependencies:

uv sync --group ui

Then, simply run:

make ui

And open your browser at http://127.0.0.1:7860

How to cite

If you use Renard in your research project, please cite it as follows:

@Article{Amalvy2024,
  doi	       = {10.21105/joss.06574},
  year	       = {2024},
  publisher    = {The Open Journal},
  volume       = {9},
  number       = {98},
  pages	       = {6574},
  author       = {Amalvy, A. and Labatut, V. and Dufour, R.},
  title	       = {Renard: A Modular Pipeline for Extracting Character
                  Networks from Narrative Texts},
  journal      = {Journal of Open Source Software},
} 

We would be happy to hear about your usage of Renard, so don't hesitate to reach out!