This data is used in the meta-evaluation of metrics for content preservation in cases with style and attribute transfer.
The data consists of 500 samples with a source sentence and a transfer sentence to a specific style.
The data is annotated by 3 humans on two dimension 'style strength' (AnswerS_1,AnswerS_2,AnswerS_3) and 'content preservation' (AnswerC_1,AnswerC_2,AnswerC_3) on a… See the full description on the dataset page:
https://huggingface.co/datasets/APauli/style_eval_content_test.