This is a simple hate speech classifier that uses a free hate speech data set from https://data.world/thomasrdavidson/hate-speech-and-offensive-language
The model itself uses embedding, 1DConv and LSTM layers, with ~93% classification accuracy. For the purposes of this exercise rows that are labelled to contain offensive language, but are not hateful, are dropped.