BioELECTRA: An Efficient Approach to Biomedical Question Answering

Biomedical Question Answering system, developed as part of my Masters' Thesis

Project Supervised by Dr Noura Al-Moubayed

The repository is organised as follows:

• Code:
     |-- checkpoints: saved model checkpoints
     |-- datasets: pretraining and finetuning datasets
     |-- src: source code
  
• ShellScripts: slurm scripts for running experiments on NCC
• Papers:
     |-- LiteratureSurvey.pdf
     |-- ProjectLog.pdf
     |-- ProjectPaper.pdf
  

Note:

• ProjectLog.pdf is a log of project progress including notes from supervisor meetings.
• requirements.txt contains the python packages required to run this project.

Datasets are not provided due to copyright, however, they can be acquired from the following locations:

• Pubmed Pretraining Corpus: https://ftp.ncbi.nlm.nih.gov/pubmed/baseline/
• BoolQ: https://github.com/google-research-datasets/boolean-questions
• SQuAD: https://rajpurkar.github.io/SQuAD-explorer/
• BioASQ: http://participants-area.bioasq.org/datasets/