A subword tokenization method that breaks words into meaningful pieces, commonly used in BERT-style models.