I would even say that OCR can rеаd the sеntеnсе ϲоrrесtlу, while a tokenizer can't.