Skip to content
Religious Hate Speech Detection for Arabic Tweets
Branch: master
Clone or download
Latest commit bd1e609 Feb 1, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
AraHate-BNS.csv
AraHate-CHI.csv adding files May 3, 2018
AraHate-PMI.csv
README.md Update README.md Feb 1, 2019
stop_words.csv adding files May 3, 2018
test.csv update files May 3, 2018
train.csv update files May 3, 2018

README.md

Religious Hate Speech Detection for Arabic Tweets

This repo contains both training (train.csv) and testing (test.csv) datasets that are used in our paper. The training dataset contains 5,569 examples, while the testing dataset contains 567 examples. As per Twitter's Developer Policy, we are only allowed to publicaly share tweet ids along with its annotaion (1 = hate , 0 = not hate). Full tweets (except for those that have been deleted or made private) can be obtained by hydrating tweet ids using tools such as twarc.

We also provide three Arabic hate term lexicons. Each term is assigned a positive or a negative real-valued score. positive scores represent positive association with hate class, while negative scores reflect negative assocation with hate class.

  1. AraHate-PMI.csv: Hate scores were assigned based on the Pointwise Mutual Information (PMI) metric.
  2. AraHate-CHI.csv: Hate scores were calculated using the chi-square statistic.
  3. AraHate-BNS.csv: Hate scores were computed using the Bi-Normal Separation (BNS) method.

Finally, we provide a list of 356 Arabic stop words (stop_words.csv) accounting for both Dialectal Arabic and Modern Standard Arabic.

Please cite our paper if you find any of our data helpful for your research:

@inproceedings{Albadi2018are,
  title={Are They Our Brothers? Analysis and Detection of Religious Hate Speech in the Arabic Twittersphere},
  author={Albadi, Nuha and Kurdi, Maram and Mishra, Shivakant},
  booktitle={Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining},
  pages={69--76},
  year={2018},
  organization={ACM}
}
You can’t perform that action at this time.