Sometimes I have very long meetings and cannot remember who said what. Identifying speakers from text alone is difficult, but I usually recognize their voices. If I could listen to a voice sample from each unidentified speaker, I could tag them with 99% accuracy.