Work

Enhancing Automated Hate Speech Detection: Addressing Islamophobia and Freedom of Speech in Online Discussions

Year: 2023
Type: conference-paper
Abstract: This paper emphasizes the necessity of a precise definition of Islamophobia within the realm of social media platforms. The current broad understanding often leads to misclassification and poses challenges to the principles of freedom of speech. Differentiating between Islamophobia and legitimate criticism presents a complex task for automated hate speech detection models, particularly in the presence of offensive language and emotionally charged tones. Furthermore, the paper highlights the inadvertent discriminatory consequences that can arise from misusing Islamophobia detection models against atheists, feminists, ex-Muslims, and others, underscoring the importance of safeguarding their rights. Our study introduces a refined definition and employs advanced deep learning models. It demonstrates a reduction in the number of Islamophobic comments in the dataset while maintaining the accurate identification of genuine instances of Islamophobia. This distinction is made without compromising discussions related to religion and criticism. The results show promise in improving the precision of Islamophobia identification, all while upholding principles of free expression and open dialogue.
Language: English
Open Access status: hybrid
OpenAlex ID: W4392845902
FWCI: 0.686
Cites: 19
Cited by: 11
Related to: 10