Reproduce “End-to-End Neural Speaker Diarization with Self-attention” 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
Yusuke Fujita, Naoyuki Kanda, Shota Horiguchi, Yawen Xue, Kenji Nagamatsu, Shinji Watanabe
The paper presents SA-EEND, an end-to-end speaker diarization system that predicts separate speech activity for each speaker at every frame, allowing overlapping speech. It replaces BLSTM encoding with self-attention and uses permutation-free training. On a two-speaker simulated test set with 34.4% overlap, it achieved 7.91% DER versus 12.28% for BLSTM-EEND; this result is limited to reported two-speaker conditions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.