Using machine learning for toxic text detection on Android

Harmful content is gradually emerging under the joint promotion of social media and instant messaging applications, such as hate speeches and fake news, which have a huge impact on social public opinion. Therefore, it is necessary to detect and filter harmful content through effective detection methods to prevent harmful content from causing harm to users. This detection technology is very important for social media, instant messaging applications, and content platforms. To this end, we mainly developed a harmful content detection tool for the Android platform, fine-tuned the BERT model to the Android platform, and finally got a robust scheme to detect harmful content, and got a 98.25% F1 score.