BERT is a popular Masked Language Model. Some words are hidden from the model and trained to predict them. The model is bidirectional, meaning it has access to the words to the left and right, making it a good choice for tasks such as text classification. Training BERT can quickly
