chunked/nltk_chunking/tasks_document_question_answering/chunk_20.txt

updated_dataset = updated_dataset.filter(lambda x: len(x["words"]) + len(x["question"].split()) < 512)

At this point let's also remove the OCR features from this dataset.