RDFizing relation extraction datasets and benchmarking relations and sentences.
The detail documentaion of RELD are available on our homepage
The script can be used to generate the RDF of a single dataset individually or at once for all datasets (The process will take a few hours to complete for all the datasets at once)
The required packages for running the script will be installed by running the following command:
pip install -r requirements.txt
To convert all the datasets at once, you need to run the following command:
python data_loader.py
Note: The datasets files must be in the respective folders inside the data folder, otherwise, you need to set the path variable inside each script.
| Dataset | Download |
|---|---|
| Wikipedia_Wikidata | Download |
| SemEval 2010 Task 8 | Download |
| WEBNLG | Download |
| Google RE | Download |
| FewRel | Download |
| NYT-FB | Download |
| DocRed | Download |
| T-REx | Download |
The generated dumps in .ttl format are available online here
The dumps are also available in JSON-LD format for non-semantic web community here
The endpoint for RELD is live here
A linux based local Virtuoso endpoint is avaliable here that is easily configurable.
Download the local Virtuoso from from the above mentioned link and unzip it; then run the following commands to setup a local instance:
cd bin
sh start_virtuoso.sh
After running the above commands the local instance will be availble on http://localhost:8890/sparql
- Manzoor Ali (DICE, Paderborn University)
- Muhammad Saleem (DICE, Paderborn University)
- Diego Moussallem (DICE, Paderborn University)
- Mohamed Ahmed Sherif (DICE, Paderborn University)
- Axel-Cyrille Ngonga Ngomo (DICE, Paderborn University)
The source code of this repo is published under the GNU General Public License v3.0