Konstantinos/BERTaTweetGR

The training dataset of this model consists of 23 million tweets in Greek, of approximately 5000 users in total, spanning from 2008 to 2018 . The model has been trained to support the work for the paper Multimodal Hate Speech Detection in Greek Social Media [link TBD] The training data includes approximately 5000 tweets in total from 2008-2018 . The training is based on the data from the work of a paper on hate speech detection in Greek social media in the form of hate speech in a paper called "Hate Speech Detectives in Social Media" The model is a light-lite RoBERTa fill mask model trained mostly in greek tweets with the help of AutoTokenizer and AutoModel,