This is a personal convenience copy of the binary Hate Speech (HS) dataset used in the T-Miner paper on defending against trojan attacks on text classifiers: https://arxiv.org/pdf/2103.04264.pdf
The dataset is sourced from the original paper's Github repository: https://github.com/reza321/T-Miner
Label mapping:
- 0 = hate speech
- 1 = normal speech
If you use this dataset please cite the T-Miner paper (see bibtex below), and the two original papers from which T-Miner constructed the dataset (see paper for references):
title={T-Miner: A Generative Approach to Defend Against Trojan Attacks on DNN-based Text Classification},
author={Azizi, Ahmadreza and Tahmid, Ibrahim and Waheed, Asim and Mangaokar, Neal amd Pu, Jiameng and Javed, Mobin and Reddy, Chandan K. and Viswanath, Bimal},
booktitle={Proc. of USENIX Security},
year={2021}}