UBC ARBERT and MARBERT Deep Bidirectional Transformers for Arabic

We utilized a pre-trained model to classify Arabic text. After conducting extensive research, we found that MarBERT was the best model for classifying Arabic offensive tweets. It focuses on dialectal Arabic (DA) and Modern Standard Arabic (MSA). The competition involves two shared sub-tasks: detecting whether a tweet is offensive or not; and det…

We compare the performance of multiple BERT-based models for the task of Emotion recognition in Arabic Tweets.