Protecting Children from Harmful Audio Content: Automated Profanity Detection From English Audio in Songs and Social-Media

A novel approach for the automated detection of profanity in English audio songs using machine learning techniques. One primary drawbacks existing systems is only confined to textual data. The proposed method utilizes a combination feature extraction techniques and algorithms identify songs. Specifically, employs popular Term frequency–inverse document frequency (TF-IDF), Bidirectional Encoder Representations from Transformers (BERT) Doc2vec extract relevant features TF-IDF used capture importance each word song, while BERT utilized contextualized representations words that can more nuanced meanings. To semantic meaning songs, also explored use Doc2Vec model, which neural network-based study Open Whisper, an open-source library, develop implement approach. dataset was evaluate performance method. results showed both models outperformed model terms accuracy identifying has potential applications various forms content, including clips, social media, reels, shorts.