Multi-task transfer learning for the prediction of entity modifiers in clinical text: application to opioid use disorder case detection.

We show that learned weights from our shared model can be effectively transferred to a new partially matched data set, validating the use of transfer learning for clinical text modifiers.