Flamers'Shield

Is it possible to detect and classify toxic comments without the presence of a human agent and act accordingly? This project tries to help by automating this.