An API that predicts the nature of toxicity of a sentence with multi-label prediction
The nature of toxicity that can be detected are: Toxic, Severe Toxic, Obscene, Threat, Insult, Identity Hate. The API can return more than a single label (multi-labeled) and the word Not-Toxic if the text isn't toxic.
Feature to be Added: User can select if they want the labels as a list or all classes with with probablity and decide a threshold for themselves.
Built with: Python, FastAPI, Tensorflow