chunked/nltk_chunking/tasks_translation/chunk_6.txt

Tokenize the input (English) and target (French) separately because you can't tokenize French text with a tokenizer pretrained on an English vocabulary.