BERT (Bidirectional Encoder Representations from Transformers)

A transformer-based language model approach that learns representations using context from both directions in text. It supports language-understanding tasks; suitability depends on the data, task and evaluation requirements.