After Babel Fish: The promise of cheap translations at the speed of the Web
3 days ago
- #translation-ethics
- #machine-translation
- #language-technology
- Translation is a complex process requiring imagination, ingenuity, and freedom, not just copying.
- Babel Fish, introduced in 1997, was an early online machine translation tool but had significant limitations and errors.
- Umberto Eco highlighted Babel Fish's shortcomings, showing that translation requires contextual and world knowledge beyond simple synonym-swapping.
- Modern machine translation (MT) has improved with statistical and neural net approaches, but still struggles with idiomatic expressions, grammar, and long texts.
- Current MT tools like Google Translate and Meta's SEAMLESSM4T handle high-resource languages well but fail with low-resource languages due to lack of data.
- David Bellos noted that human translators develop 'automatisms' for recurring issues, but MT is catching up in routine tasks.
- Industry leaders predict fully automated translation, reducing human involvement to fringe roles like fail-safe mechanisms.
- MT still faces challenges like translationese, verbosity, hallucinations, and errors in gender, idioms, and wordplay.
- Ethical concerns arise as MT prioritizes technical metrics over faithfulness, loyalty, and cultural understanding in translation.
- The push for automation risks narrowing linguistic and cultural diversity, reducing translation to a purely technical activity.