A concept-based domain-tuned speech translation system is more robust to noise and disfluency than general-purpose commercial systems

A concept-based, domain-tuned speech translation system is more robust to background noise, speech disfluency, and rapid speech — in both speech-recognition word error rate and translation accuracy — than general-purpose commercial speech translation systems.

Narrative synthesis #