search

IsarNejad / TCAV-for-Text-Classifiers

TCAV for NLP, published at ACL2022
MIT License
6 stars 0 forks source link

Question about EA data set #1

Open jiannan-xu opened 1 year ago

jiannan-xu commented 1 year ago

Hello, I am currently attempting to replicate some results from this paper, but I've hit a block while trying to figure out how to access the EA data set. Could you please provide more detailed instructions on how to find the EA-dev data set? I have already looked up the link provided (https://zenodo.org/record/3816667#.ZEh0IuyZOqV), but unfortunately, I was unable to find any useful information on how to reconstruct this data set. Thank you for your assistance!

IsarNejad commented 1 year ago

The authors of the EA paper have created new versions since I did this work, which is why you are seeing various versions of this data. From the link you referred to "hs_AsianPrejudice_20kdataset_cleaned_anonymized.tsv" should be the one I used. The ids of the dev set can be found at https://github.com/IsarNejad/TCAV-for-Text-Classifiers/tree/main/Data For further communications, please use this email [isar.nejadgholi@nrc.cnrc.gc.ca]