chunked/nltk_chunking/model_doc_mms/chunk_3.txt

The main ingredients are a new dataset based on readings of publicly available religious texts and effectively leveraging
self-supervised learning.