Romanian Covid Question Answering

This repository contains the a Romanian question answering for Covid-19 related queries. We used as the base model the Romanian DistilBERT.

The model was introduced in this paper.

Usage

Important Note: Currently, the model is not available for usage through the HuggingFace library.

Please install the following package:

pip install git+https://github.com/racai-ai/BERT-SQuAD

Then you can use the model with the following code sample:

from bert import QA

model = QA("racai/e4a-covid-question-answering")

context = "Reacțiile adverse frecvente la tratamentul cu paxlovid sunt diaree, vărsături, mioartralgii, astenie fizică, erupții, inclusiv sindrom Stevens-Johnson."
question = "La ce reacții adverse să mă aștept dacă iau paxlovid?"

out = model.predict(context, question)
print(out)
# {'answer': 'diaree, vărsături, mioartralgii, astenie fizică, erupții, inclusiv sindrom Stevens-Johnson.', 'start': 8, 'end': 16, 'confidence': 0.999903004533245, 'document': [...]}

Citation

@article{ion2022open,
  title={An Open-Domain QA System for e-Governance},
  author={Ion, Radu and Avram, Andrei-Marius and P{\u{a}}i{\c{s}}, Vasile and Mitrofan, Maria and Mititelu, Verginica Barbu and Irimia, Elena and Badea, Valentin},
  journal={arXiv preprint arXiv:2206.08046},
  year={2022}
}