using automatically derived transcript of audio data