Researchers create AI tool for detecting hate speech in Southeast Asian languages

SGHateCheck addresses biases seen in traditional evaluation methods by offering over 11,000 culturally relevant test cases

Image credit: Image Creator from Microsoft Designer/DALL.E and Canva

SGHateCheck is an AI-powered hate speech detection tool tailored for Southeast Asia, particularly Singapore.

Current hate speech models often fail to capture the region's unique linguistic and cultural nuances since they are primarily based on Western contexts.

SGHateCheck builds on existing frameworks like HateCheck and Multilingual HateCheck (MHC) but introduces functional tests that reflect the linguistic diversity of Singapore, including English, Mandarin, Tamil, and Malay. The model incorporates large language models (LLMs) and native annotators to ensure accuracy.

The tool addresses biases seen in traditional evaluation methods by offering over 11,000 culturally relevant test cases. It effectively captures regional expressions, like Singlish, to provide a more precise evaluation of hate speech in Southeast Asia. The research highlights the need for culturally sensitive AI models and plans to expand SGHateCheck to include other Southeast Asian languages.