EMNLP paper analyzes 79,286 posts: MT users and AI devs sharply disagree on what matters
EhudReiter · x · 2026-09-25
Yujun Wang (student of Ehud Reiter) will give an EMNLP oral on "Beyond Accuracy: Community Perspectives on Machine Translation" — the first large-scale study of how non-AI communities view MT.
- Dataset of 79,286 posts/comments from Reddit, Facebook, Bluesky, and Mastodon (2019–2025)
- Four stakeholder communities analyzed: AI developers, professional translators, language learners, and language service providers
- Communities strongly conflict on translation quality, efficiency, and reliability: AI folks frame these as technical/computational problems, while user communities care about quality nuances, time savings, trust, and broader social issues
- The authors argue research effort should be steered toward problems communities actually care about rather than benchmark performance alone
More from Research
- Universal Dimensions Track Human and Monkey Brains, Alignment Boosts Universality — martin_hebart · 2026-09-25
- Universality as an Index of AI-to-Human Alignment, Says Hebart Team — martin_hebart · 2026-09-25
- Architecture, Data, and Scale Don't Explain Universality in Vision Models — martin_hebart · 2026-09-25
- More Universal Dimensions in Vision Models Are More Semantic and Conceptual — martin_hebart · 2026-09-25
- Human Ratings Show Interpretability Is an Emergent Property of Universality — martin_hebart · 2026-09-25
- 162 Vision Models Compared: NeurIPS Paper Finds Universal Representations Align With Human and Monkey Brains — martin_hebart · 2026-09-25