NLP system trained on the Stanford Question Answering Dataset (SQuAD) that, given a paragraph and a question regarding it, provides a single answer, which is obtained selecting a span of text from the paragraph.
In order to install all the necessary libraries it suffices to run
pip install -r requirements.txtWe suggest the use of venv if you are running in local.
Be sure to download the weights of our model from this link and save them in the same directory of the compute_answers.py script.
The predictions.json file can be obtained with the following command
python3 compute_answers.py <dataset.json>