Optimizing Hate Speech Detection in the Indonesian Language using FastText and LSTM Algorithms

The term hate speech encompasses communication that conveys hostility or prejudice toward individuals or groups based on specific attributes, including race, ethnicity, gender, or religion. Given the prevalence of such harmful content on major social media platforms like Twitter, it is imperative to prioritize ongoing research into automated detection methods. This study employs advanced artificial intelligence and machine learning techniques, involving pivotal stages such as data collection, preprocessing, oversampling, fastText word embedding, and classification using Long Short-Term Memory (LSTM) networks to achieve the objective. The amalgamation of fastText embeddings with LSTM classification yielded noteworthy results, with the model achieving 94% accuracy, 92% precision, 94% recall, and a 93% F1 score. These outcomes underscore the effectiveness of sophisticated machine learning methodologies in accurately discerning and addressing the proliferation of hate speech online, particularly in the context of Indonesian-language content.