Self-Reflective and Introspective Feature Model for Hate Content Detection in Sinhala YouTube Videos

摘要

YouTube is considered one of the most popular social media platforms, which provides users with the ability to interact with each other by sharing videos, commenting or liking or disliking. Its free nature has enabled the spread of offensive and hateful content within this environment, resulting in violence and discrimination within society. Therefore, identifying hate content is crucial to mitigating the spread of hatred. This study describes a system to detect hate in Sinhala content associated in YouTube videos by natural language processing techniques. The categorizations are done based on user comments, thumbnail text, and Meta-data, which includes the title, description and tags. Here, the features were derived through self-reflective and introspective data associated with the YouTube video. This system is capable of detecting hate expressions in Sinhala language YouTube videos with nearly 90 percent accuracy.