chunked/nltk_chunking/model_doc_splinter/chunk_14.txt

Also, its default behavior is to use this token when two sequences are given (for
  example, in the run_qa.py script).