Importance of Supra-Segmental Information and Self-Supervised Framework for Spoken Language Diarization Task