Hugging Face's logo

Datasets:
manueltonneau
/
hateday

Tasks:
Text Classification
Languages:
English
French
Spanish
Size:
100K<n<1M
ArXiv:
Tags:
hate speech
Dataset card Files Community

You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

You agree to not use the dataset to conduct any activity that causes harm to human subjects.

Log in or Sign Up to review the conditions and access this dataset content.

HateDay [in preparation]

This dataset consists of twelve representative sets of Twitter annotated for hate speech detection for eight languages and four countries. Each representative set corresponds to a language or country and consists of 20,000 tweets randomly sampled from all tweets posted on September 21, 2022 in that language or country, for a total of 240K annotated tweets. We cover eight languages (Arabic, English, French, German, Indonesian, Portuguese, Spanish and Turkish) and four countries where English is the main language on Twitter (United States, India, Nigeria, Kenya). Each tweet is labeled as hateful, offensive or neutral by three human annotators and the final label is determined by majority vote. In case the tweet is marked as hateful, the target of hate is also indicated.

The dataset and annotation process are presented in more details in the corresponding paper.

Data access and intended use

Please send an access request detailing how you plan to use the data. The main purpose of this dataset is to evaluate hate speech detection models, as well as study hateful discourse online. This dataset is NOT intended to train generative LLMs to produce hateful content.

Columns

The dataset contains six columns:

  • tweet_id: the ID the of the tweet
  • text: the text of the tweet
  • class: the label of the tweet (2 if hateful, 1 if offensive and 0 if neutral)
  • hate_target: the target of hate in case the tweet is hateful
  • lang_country: the language or country of interest

Preprocessing

We replace all usernames and links by fixed tokens to maximize user privacy.

Citation

Please cite our paper if you use this dataset.

@misc{tonneau2024hatedayinsightsglobalhate,
      title={HateDay: Insights from a Global Hate Speech Dataset Representative of a Day on Twitter}, 
      author={Manuel Tonneau and Diyi Liu and Niyati Malhotra and Scott A. Hale and Samuel P. Fraiberger and Victor Orozco-Olvera and Paul Röttger},
      year={2024},
      eprint={2411.15462},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2411.15462}, 
}
Downloads last month
1
Edit dataset card