Code accompanying my paper Boosting Pre-trained Language Models with Task Specific Metadata and Cost Sensitive Learning, at SemEval 2022 colocated with NAACL 2022. This was the solution from PiCkLe for Task 4, Patronising and Condescending Language Detection.
- Using Pre-trained Language Models
- Utilising post metadata
- Using cost sensitive learning to tackle class imbalance.
Link to video presentation: PiCkLe at SemEval 2022
Fun fact: This research is not about pickles (or cucumbers). (Reference to emails I received since this paper was published) 😅