MTDiag: An Effective Multi-Task Framework for Automatic Diagnosis
Published in AAAI Conference on Artificial Intelligence (AAAI), 2023
MTDiag reformulates symptom checking as a multi-label classification task with multi-task learning and contrastive learning to better distinguish symptoms between diseases. It achieves state-of-the-art performance on four public datasets and is deployed in an online medical consultant system.
Recommended citation: Hou, Z., Cen, Y., Liu, Z., Wu, D., Wang, B., Li, X., Hong, L., and Tang, J. (2023). "MTDiag: An Effective Multi-Task Framework for Automatic Diagnosis." AAAI.
Download Paper