Find a story
Search Spins
Search titles, summaries, and missing voices across published articles — press releases, announcements, and media coverage.
6 results for “speech recognition”
Measuring benchmark optimization in speech recognition
Hugging Face published a blog post analyzing how speech recognition models are optimized for benchmark performance, highlighting methodological concerns in evaluation practices.
Aug 21, 2026
DonorRank: Donor Language Selection for Low-Resource Cross-Lingual Speech Recognition
Researchers introduced DonorRank, a learning-to-rank framework to improve donor language selection for zero-shot cross-lingual ASR in low-resource languages, validated on Indic and African speech corpora.
Aug 13, 2026
Speech Recognition and TTS in less than 500kb
A forum thread on Hacker News titled 'Speech Recognition and TTS in less than 500kb' surfaces community discussion around a compact AI model for speech tasks, but contains no substantive reporting, technical details, or verifiable claims about the model’s performance, origin, or deployment.
Published Jul 14, 2026 · Analyzed Jul 19, 2026
COALA: Robust Contextualized Speech-augmented Language Modeling for ASR via Contrastive Regularizer and Biasing Score Estimation
COALA is a new research framework for improving automatic speech recognition in multi-entity scenarios by introducing contrastive regularization and biasing score estimation to better match audio segments with domain-specific entities.
Jul 10, 2026
Revisiting the Relation Between Language Model Perplexity and ASR Word Error Rate for Modern End-to-End Speech Recognition
A new arXiv preprint questions the long-standing assumption that language model perplexity (PPL) reliably predicts ASR word error rate (WER) in modern end-to-end systems, showing the relationship breaks down due to internal language modeling, encoder context limits, and LLM integration.
Jul 9, 2026
S-DiverSe: Spanish Diverse Speech
Researchers released S-DiverSe, a 3.2-hour Spanish speech corpus of 22 speakers with neurological conditions (ALS, Parkinson’s, stroke), containing 444 manually transcribed segments and metadata, to address the lack of in-the-wild evaluation benchmarks for neurologically affected ASR.
Jul 8, 2026