Hate Speech Detection and Online Public Opinion Regulation Using Support Vector Machine Algorithm: Application and Impact on Social Media (MDPI)

The rarity, sarcasm, and spelling variations of hate speech make it difficult to identify on social media. This work uses SVM, which uses kernel functions to translate text information into high-dimensional space, to increase classification accuracy. Social media APIs and custom crawlers are used to gather data, which is subsequently preprocessed using stop-word removal, denoising, and regular expressions. Contextual semantics are improved using word embeddings from TF-IDF and Word2Vec’s Skip-gram model. BERT and sentiment analysis are used in a multi-level framework to provide sophisticated recognition. High accuracy and efficiency in the experimental results facilitate real-time hate speech identification and help create a safer online environment.

https://www.mdpi.com/2078-2489/16/5/344