Beyond plain toxic: building datasets for detection of flammable topics and inappropriate statements | Publicación