MeDAL: Medical Abbreviation Disambiguation Dataset for Natural Language Understanding Pretraining
2020-12-27EMNLP (ClinicalNLP) 2020Code Available1· sign in to hype
Zhi Wen, Xing Han Lu, Siva Reddy
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/mcGill-NLP/medalOfficialpytorch★ 285
Abstract
One of the biggest challenges that prohibit the use of many current NLP methods in clinical settings is the availability of public datasets. In this work, we present MeDAL, a large medical text dataset curated for abbreviation disambiguation, designed for natural language understanding pre-training in the medical domain. We pre-trained several models of common architectures on this dataset and empirically showed that such pre-training leads to improved performance and convergence speed when fine-tuning on downstream medical tasks.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| MIMIC-III | ELECTRA (pretrained) | Accuracy | 0.84 | — | Unverified |
| MIMIC-III | ELECTRA (from scratch) | Accuracy | 0.83 | — | Unverified |
| MIMIC-III | LSTM+SA (pretrained) | Accuracy | 0.83 | — | Unverified |
| MIMIC-III | LSTM (pretrained) | Accuracy | 0.83 | — | Unverified |
| MIMIC-III | LSTM+SA (from scratch) | Accuracy | 0.8 | — | Unverified |