chunked/nltk_chunking/tasks_text-to-speech/chunk_22.txt

You'll need the tokenizer part of the processor to process the text:

tokenizer = processor.tokenizer

The dataset examples contain raw_text and normalized_text features.