Hummingbird

Hummingbird is a dataset to examine stylistic lexical cues from human perception and BERT used to characterize their discrepancy. In HUMMINGBIRD crowd-workers relabeled benchmarking datasets for style classification tasks.