But it’s analyzed by a language model? But are you putting free text in there as well?
j previous speech k next speech
