Overlapping speech in meetings: Two-pass stack cuts 29-31% vs single pass

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Two-pass diarization cuts meeting error by 30% vs single-pass. See why overlap deletes speech, how top systems hit 11% DER on VoxConverse and 22% on AMI. Independent technical note mirroring the canonical version: https://transcribeall.io/blog/overlapping-speech-in-meetings-two-pass-stack-cuts-29-31-vs-single-pass.php