A Comprehensive Survey on Urdu Hate Speech Detection: Methods, Evaluation, and Challenges

Hate speeches has risen significantly because of ever expanding online communities, easy accessibility of social media platforms, anonymity of the users, rapid spread of information, and increased visibility of hate related incidents. Detection of such hate speech, especially in low-resource language i.e. Urdu, on social media has become active research area for safe, and inclusive online environment. This paper introduces a new taxonomy, and presents a survey of Urdu hate speech detection techniques. It also explores future research directions, and highlights the main challenges which persist in this field such as limited data availability, code-switching, and requirements for adaptability to contextual perspectives. The taxonomy, and the survey form the basis for continued advancement in Urdu hate speech detection that yields improved insights regarding Urdu hate speech patterns, and contributes to better detection system’s development. It serves a comprehensive review of Urdu hate speech detection methods available in the research literature, and compares those methods at an abstract level, discusses their limitations, and provides future research directions for new researches in this domain.