WikiCheck API
Repository with the implementation of WikiCheck API. The project is done in cooperation with Wikimedia Foundation and Ukrainian Catholic University.
The work was accepted to CIKM2021 applied track.
The publication can be found here: WikiCheck: An end-to-end open source Automatic Fact-Checking API based on Wikipedia (https://arxiv.org/abs/2109.00835)
The link to API: https://nli.wmcloud.org
The structure of the project:
The project consists of modules directory with the implementation of modules used for inference along with the script for NLI models training.
The configs directory includes configuration files for training and inference.
The notebooks directory (not added yet) includes .ipynb notebooks with experiments done during the research.
We use DVC with Google drive remote for efficient model version control.
If you want to get access to our fine-tuned models, you can load them from here.
Also, you can train your model by running the modules/model_trainer.py
script.
API setup and run
- Clone the official WikiCheck repo and cd into it
git clone https://github.com/trokhymovych/WikiCheck.git
cd WikiCheck
- Create and activate virtualenv:
virtualenv -p python venv
source venv/bin/activate
- Install requirements from requirements.txt:
pip install -r requirements.txt
-
Load pretrained models. There are two options:
- Loading models with DVC (preferred):
dvc pull
- Loading models from here
-
Run the API:
python start.py --config configs/inference/sentence_bert_config.json
ToDo:
- Refactor code for inference
- Write README.md
- Refactor training script
- Add training configs
- Add notebooks with experiments
- Add aggregation endpoint to API