A Scientometric Analysis of Potentially Harmful Speech Detection in Muslim Communities: Trends, Challenges, and Future Directions

With the proliferation of the internet and social media, harmful speeches targeting specific groups such as Muslims have become a prevalent form of online bullying and prejudice. Recent advances in deep learning and natural language processing (NLP) methods now make their detection both feasible and necessary. This study analyzed 217 English-language articles from the Web of Science Core Collection (WoSCC) database (2015–2024) focusing on potentially harmful speech detection methods relevant to Muslim communities. Through bibliometric analysis, the study explores research trends, geographic patterns, and cultural nuances in harmful speech detection, complemented by a SWOT analysis to identify challenges and opportunities. Findings show that research on this topic has intensified significantly since 2018. Saudi Arabia is becoming a hub for international academic cooperation with countries on multiple continents including Asia, Europe, and North America. Researchers are currently leveraging advanced neural architectures and ensemble approaches to achieve higher detection accuracy. However, progress remains uneven; many regions still lack representative studies, annotated datasets are scarce, and the intertwined religious and cultural norms of Muslim communities pose unique challenges. Future research should prioritize creating multilingual and multidialectal models capable of addressing emerging harmful language and behavior patterns on social media platforms. Furthermore, this study examines the trade-offs between detection methods and freedom of expression and proposes feasible solutions from both technological and implementation standpoints.