This would be useful when chunking PDFs scanned with OCR. I've done that before and paragraph breaks were detected pretty inconsistently.