• MeDAL: Medical Dataset for Abbreviation Disambiguation for Natural Language Understanding
  • A public medical text dataset that is comparable in size with Wikipedia, and has a training task that emphasizes domain knowledge.
  • We empirically observed gain in performance when models are first trained on MeDAL before being applied to downstream medical tasks.
  • Published at EMNLP 2020 Clinical NLP workshop.
Bruce (Zhi) Wen
Master Student in Computer Science