Text this: Unsupervised methods for speaker diarization