AkshatSh / DPD

CSE 481N: NLP Capstone (Data Programming by Demonstration)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

MultiTask Learning

AkshatSh opened this issue · comments

Train model as multi task, different head for probabilistic labels vs true labels.

Freeze-train

Train on probabilistic labels, freeze lstm after training probabilistic, and refine CRF for gold.