chunked/nltk_chunking/model_doc_bert/chunk_2.txt

The abstract from the paper is the following:
We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations
from Transformers.