The roBERTa-based target-demographic classifier is finetuned on the CONAN dataset for classifying whether a response's content is about one or multiple of the 8 target demographics, based on the topic classifier cardiffnlp/tweet-topic-21-multi

Currently trained for the following classes: ["MIGRANTS", "POC", "LGBT+", "MUSLIMS", "WOMEN", "JEWS", "other", "DISABLED"]

The model is intended for classifying LM-generated dialogue responses, and evaluating their relevancy to the given input sequence.