chunked/nltk_chunking/model_doc_align/chunk_12.txt

We show that the scale of our corpus can make up for its noise and leads to state-of-the-art representations even with such a simple learning scheme.