chunked/nltk_chunking/tasks_question_answering/chunk_18.txt

Use the [~tokenizers.Encoding.sequence_ids] method to
   find which part of the offset corresponds to the question and which corresponds to the context.