Information extraction from Legal Documents

This packages contains:

• a criminal ontology for criminal appeals to the French Cassation Court (2023)
• a pipeline for fetching pdfs of appeals from the French Cassation Court
• a pipeline for deriving RDF triples from the appeals pdfs (based on GPT-4o mini)
• upload pipeline to Fuseki triple store
• sparql analysis scripts

Downloading court decisions use geckodriver

1. Download gecko driver

wget https://github.com/mozilla/geckodriver/releases/download/v0.32.0/geckodriver-v0.32.0-linux64.tar.gz

2. install xvfb and set parameters, do not run geckodriver as root user

sudo apt install xvfb
Xvfb :99 -screen 0 1024x768x24 &
export DISPLAY=:99

3. run geckodriver (default port is 4444)

nohup geckodriver --port 4444 &

4. scrape

setting up triple store (Fuseki)

cd docker/fuseki
cp .example.env .env

Appeals Ontology