Automatic Discovery Of Harsh Messages In Social Media

Cyberbullying has arisen as a widespread issue for teenagers, particularly children and young adults, due to the rise in popularity of social media. Automatic identification of bullying messages in social media becomes possible by machine learning methods, and this will allow for the construction of a secure and stable social media community. The topic of reliable and discriminative numerical representation learning of text messages is a crucial concern in this important research field. We've developed a new learning approach for this project, as described in this article. The method smSDA, which was named after the method SmSDA, which was derived from the common deep learning model stacked denoising autoencoder, is constructed using semantic extension of that model. The semantic extension includes semantic dropout noise, a methodology that utilizes domain awareness, and sparse constraints, which is implemented using term embedding techniques. Our method can leverage the secret text function structure and learn a comprehensive and discriminatory representation of bullying content. The tests are done on two publicly available online corpora of cyberbullying (Twitter and MySpace) and the findings indicate that our suggested baseline techniques outperform other models