In recent years, the growth of the internet and social media platforms has led to a surge in online content creation. Unfortunately, with this growth, there has also been an increase in inappropriate content such as hate speech. Hate speech, which targets individuals based on their ethnicity, religion, sexual orientation, and other characteristics, can have harmful effects on society. To combat this issue, hate speech detection models have been developed as computational systems that can identify and classify online comments as hate speech. These models play a crucial role in moderating online content and reducing the spread of harmful speech, particularly on social media platforms.
One of the challenges faced in evaluating hate speech detection models is the bias present in the datasets used for testing. Traditional evaluation methods, such as using held-out test sets, often fail to provide a comprehensive assessment of the model’s performance. To address this limitation, researchers have introduced functional tests like HateCheck and Multilingual HateCheck (MHC). These tests aim to capture the complexity and diversity of hate speech by simulating real-world scenarios, providing a more robust evaluation of the models.
In their research paper titled “SGHateCheck: Functional tests for detecting hate speech in low-resource languages of Singapore,” Assistant Professor Roy Lee and his team from the Singapore University of Technology and Design (SUTD) present a new approach to hate speech detection. Building on the frameworks of HateCheck and MHC, the team developed SGHateCheck, an artificial intelligence (AI)-powered tool specifically designed for Southeast Asia. This tool aims to distinguish between hateful and non-hateful comments in the context of Singapore and the surrounding regions, addressing the linguistic and cultural specificities of the area.
Current hate speech detection models and datasets are primarily based on Western contexts, which may not accurately represent the social dynamics and issues in Southeast Asia. SGHateCheck fills this gap by providing functional tests tailored to the region’s linguistic and cultural needs. Unlike its predecessors, SGHateCheck utilizes large language models (LLMs) to translate and paraphrase test cases into Singapore’s main languages, ensuring cultural relevance and accuracy. With over 11,000 meticulously annotated test cases, SGHateCheck offers a nuanced platform for evaluating hate speech detection models in the region.
The team behind SGHateCheck found that LLMs trained on monolingual datasets tend to be biased towards non-hateful classifications. In contrast, LLMs trained on multilingual datasets demonstrate a more balanced performance and better accuracy in detecting hate speech across different languages. This highlights the importance of including culturally diverse and multilingual training data for applications in multilingual regions like Southeast Asia. SGHateCheck, with its focus on regional linguistic features and expert guidance, ensures that hate speech detection tests are relevant and effective in the context of the region.
SGHateCheck is poised to make a significant impact on the detection and moderation of hate speech in online environments in Southeast Asia. By enhancing the accuracy and cultural sensitivity of hate speech detection models, SGHateCheck aims to foster a more respectful and inclusive online space. The implementation of SGHateCheck in social media platforms, online forums, news websites, and other online spaces will be valuable in combating hate speech. Asst. Prof. Lee’s plans to develop a new content moderation application using SGHateCheck and expand its capabilities to include additional Southeast Asian languages demonstrate the potential for real-world applications of this technology.
The development of SGHateCheck represents a significant step towards enhancing hate speech detection in Southeast Asia. By focusing on the region’s linguistic and cultural specificities, the tool provides a more accurate and culturally sensitive approach to identifying hate speech online. Through the integration of cutting-edge technological advancements and thoughtful design principles, SGHateCheck exemplifies the importance of a human-centered approach in technological research and development. As online platforms continue to grow, tools like SGHateCheck will play a crucial role in creating a safer and more inclusive online environment for all users.