Reproduce “eDIAR: Training-Free Open-Speaker Streaming Diarization for Long Meetings and Large Archives via e-means” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Kyudan Jung, Jihwan Kim, Minwoo Lee, Eunji Shin, Jaegul Choo

eDIAR extends fixed-slot streaming diarization by clustering voices that share a slot after capacity is exhausted, while retaining the neural head’s activity and overlap decisions. On a constructed 20-speaker LibriSpeech-OpenSpk streaming task, it reaches 44% DER versus 66% and 81% for the eight-slot Nemotron-3 head and four-slot Streaming Sortformer. These results concern constructed read-speech conversations, not arbitrary real meetings.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.