cardiffnlp/twitter-roberta-base-irony

This is a roBERTa-base model trained on ~58M tweets and finetuned for irony detection with the TweetEval benchmark . The model was trained on tweets from 58M tweets . The results were based on the findings of EMNLP 2020 . The data was downloaded from the official repository of Twitter's Twitter-roberta-base-base task . The code is used to train the model and test the accuracy of the tweetEval test . It is based on tweets sent by users with their username and link placeholders . For example: "Great, it broke the first day. Great, it was great, it wasn't good, it didn't break the first,