The proposed implementation performs information retrieval on documents starting from a question. The documents are aggregations of the paragraphs from each entry of the SQuAD 1.1 dataset. A solution that matched questions to paragraphs was developed on a subset during the exploratory phase (code snippets in src/testing_illuin.ipynb), yet was hard to scale due to RAM constraints when running on the training set.

1. Run the executable in src/dist/
2. Provide the path to the SQuAD JSON file of your choice (both training and testing work)
3. Wait a few seconds for the dataset to be preprocessed then select a number of questions you want to retrieve documents for
4. Repeat