<![CDATA[Hate speech is a biased, antagonistic, and discriminatory expression that commonly appears on social media platforms, including TikTok. The high volume of comments and varied language styles make manual detection challenging. This research proposes a hate speech detection model using the Multinomial Naïve Bayes algorithm with smoothing to address zero-probability issues and enhance prediction performance. The dataset is split into 80% training and 20% testing portions. The model achieves an accuracy of 88.41%, with precision, recall, and F1-score showing balanced performance. A user evaluation involving 35 participants and 7,415 TikTok comments records a detection accuracy of 68.6%. The model is further implemented into a Google Chrome extension capable of real-time hate speech detection, displaying prediction probabilities and allowing user validation. This study aims to support healthier digital interactions by improving automated hate speech detection on social media.]]>
Copyrights © 2025
