A 15.3M-parameter end-to-end diarization model with conformer decoders and an attractor-based deep clustering loss reports 4.99% DER on CALLHOME, beating published EEND baselines.
Early methods often combined Gaussian Mixture Models (GMM) or i-vector
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.SD 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
End-to-End Diarization utilizing Attractor Deep Clustering
A 15.3M-parameter end-to-end diarization model with conformer decoders and an attractor-based deep clustering loss reports 4.99% DER on CALLHOME, beating published EEND baselines.