A text input transcription of the utterance. It is only applicable for test-sets containing text data.