This repository is the implementation of our COLING 22 paper A Two Stage Adaptation Framework for Frame Detection via Prompt Learning. Our implementation is based on the project OpenPrompt, many thanks to THU for the open-source resource.
we get datasets available mainly from three resources: (1) news media (articles), (2) social media, (3) debates and statements. Five datasets are listed.
We use mfc for pre-training, and use the others for downstream generalization.
see pre-train directory.
Initialize the components with pre-trained parameters and tune on downstream tasks.
see gene directory.
Our pre-trained model can be download through link