Social media has revolutionized the way we communicate, but it has also opened the door for the rapid spread of harmful content, including hate speech. Hate speech on social platforms not only affects individuals and communities but can also escalate into violence and societal harm. In recent years, there has been a growing concern about the intersection of hate speech and fake narratives, which has prompted the need for specialized methods to detect and curb their spread.
Briefly, Faux-Hate is the generation of hate speech driven by fake narratives. This task focuses on identifying comments that blend fake information with hateful language to mislead and provoke individuals, exacerbating the impact of hate speech. The goal of this shared task is to explore how fake narratives can contribute to the propagation of hate and to develop models that can detect such instances within code-mixed Hindi-English social media text.
The Faux-Hate shared task is designed to challenge participants to tackle both fake and hate detection in social media comments, with additional emphasis on identifying the target and severity of hateful speech.
Participants will receive a dataset containing text samples, each labeled with:
The objective of this sub-task is to develop a single multi-tasking model that outputs both the fake and hate labels for each text sample.
Participants will receive a dataset containing text samples, each labeled with:
The objective of this sub-task is to develop a single model that generates both the target and severity labels for a given text sample.
The provided dataset is divided into three sections, each set will be available for download on the respective release dates.
| Dataset | Date of Release |
|---|---|
| Train | 28th October |
| Validation | 13th November |
| Test | 18th November |
All test results should be submitted through the Google Form, accessible at (link will be provided later).
Each team must submit a zip file named teamname.zip, containing CSV files for both Task A and Task B. Include one or two prediction files per task, named as teamname_TaskA_runX.csv or teamname_TaskB_runX.csv (replace X with run number) for clarity.
Submissions will be evaluated based on the Macro F1 Score as the primary metric.
The rank list will be generated based on the scores from the evaluation metrics.
To participate, please fill out the registration form. Click here to register.
Upon successful registration, participants will receive the dataset via email. Please ensure your email address is valid and accessible during the competition.
All participants are required to cite the following article in their working notes:
@article{biradar2024faux,
title={Faux Hate: Unravelling the Web of Fake Narratives in Spreading Hateful Stories:
A Multi-Label and Multi-Class Dataset in Cross-Lingual Hindi-English Code-Mixed Text},
author={Biradar, Shankar and Saumya, Sunil and Chauhan, Arun},
journal={Language Resources and Evaluation},
pages={1--32},
year={2024},
publisher={Springer}
}
This citation is essential for acknowledging the foundational work that contributes to the shared task and ensuring proper attribution in your research.
For queries, please reach out to the following:
Shankar Biradar: shankar@iiitdwd.ac.in
We’re here to help and look forward to your participation!