ASR course slides refreshed: less HMM, more speech tokens, speech LLMs and full-duplex
solyarisoftware · x · 2026-10-31
- Laurent Besacie refreshed his ASR (speech recognition) course deck — under 100 slides, with figures partly made using Claude — and is soliciting feedback.
- The structure mirrors a generational shift: far less HMM coverage, with new focus on speech tokens, speech LLMs, and full-duplex models.
- Quote-tweeter @joshmeyer admits he still has a soft spot for classic GMMs, underscoring the migration from statistical modeling to speech foundation models. A handy primer for anyone catching up on modern ASR.
More from Research
- Prune-then-loop retrofit turns a pretrained VLM recurrent, beating baseline with 23.3% fewer language params — HildeKuehne · 2026-10-31
- Microsoft Research open-sources Agensh: 1,024 coding agents, no boss, 55% on Pandoc rebuild — JafarNajafov · 2026-10-31
- Tsinghua team's ActionCodec2 streams action tokens per step, hitting 46.2% grasp vs FAST's 9.2% — ZeYanjie · 2026-10-31
- OpenAI leak's hidden gem: a Log-Sobolev result completing KLS picture — IgorCarron · 2026-10-31
- "Probably present": A Visual Explainer of How Bloom Filters Work — YashasGunderia · 2026-10-31
- S2COPE discovers interpretable visual concepts from unlabeled images — burkov · 2026-10-31