This folder contains a subset of the human label annotations carried out for our paper. There is one .tsv file for each model we used in our experiments. The columns are to be read as such:

REFFERRING TO ORIGINAL GOLD DATA:
index
gold text	
label	
target_race	
target_religion	
target_origin	
target_gender	
target_sexuality	
target_age	
target_disability	

REFERRING TO SYNTHETIC DATA:
synth_text	
filter-yes (based on Roberta Large)
fuzzy_match	
Prompt failure?	
hate speech?	
grammar correct?	
world knowledge correct?	
target_origin-s	
target_race-s	
target_religion-s	
target_gender-s	
target_sexuality-s	
target_age-s	
target_disability-s

