In this project, we propose to deal with the data scarcity problem in a specific NLP task by harnessing existing annotated datasets from related tasks. Our approach involves training a multi-head a…