Ouvia study: only about half of real-world speech translation interactions rated usable
MaartenSap · x · 2026-10-25
Ouvia, a paper by Giuseppe Attanasio, Maarten Sap et al. (ICLR poster Oct 26), introduces a user-centered framework for measuring speech translation usability. Across 1,750+ real one-to-one interactions in healthcare and everyday settings, mediated by 4 ST systems, only half of interactions were rated usable, with significant gaps across demographic groups. QA-based evaluation predicted real-world usability far better than standard metrics.
More from Research
- EMNLP 2026 paper exposes hidden representational instability in VLMs — PMinervini · 2026-10-25
- AI fixes 95.8% of failed spinal-cord labels: nnU-Net rescues patient MRI data — bravo_abad · 2026-10-25
- MemOS teases memt1.0: memory-native MoE architecture, ~5% coding gains — CuriousSiopao · 2026-10-25
- Compositionality in Emerging Vocal Communication Can Be Linear, Simultaneous, and Iconic — abenitezburraco · 2026-10-25
- Anthropic co-founder Jack Clark moves self-improving AI prediction to 2027 — The Decoder · 2026-10-25
- GPT-6 Astra Hits 39% on OpenDS-World Live Web Deep Search Benchmark vs 21.1% Human Average — emmanuelvivier · 2026-10-25