Quick, get me a Dr. BERT: Automatic Grading of Evidence using Transfer Learning
We describe our methods for automatically grading the level of clinical evidence in medical papers, as part of the ALTA 2021 shared task. We use a combination of transfer learning and a hand-crafted, feature-based classifier. Our system (�orangutanV3�) obtained an accuracy score of 0.4918, which placed third in the leaderboard. From our failure analysis, we find that our classification techniques do not appropriately handle cases when the conclusions of across the medical papers are themselves inconclusive. We believe that this shortcoming can be overcome�thus improving the classification accuracy�by incorporating document similarity techniques.
PDF Abstract