Brief

Researchers have developed a new data augmentation technique called Phonetically-Informed Data Augmentation (PiDA) to improve Vietnamese speech translation. The method addresses error propagation in cascaded speech translation systems by generating ASR-like corruptions based on phonetic confusions. Fine-tuning with PiDA on the FLEURS Vietnamese-English dataset enhanced translation accuracy for erroneous ASR outputs, showing a notable improvement in BLEU scores. AI

Researchers have developed Ouvia, a new framework for evaluating speech translation (ST) systems based on user-perceived usability in real-world scenarios. A study involving over 1,750 interactions between English and Portuguese speakers found that current ST systems are only usable in about half of cases, with notable disparities across demographic groups. The framework highlights the need for situated, user-centered evaluations that go beyond standard quality metrics to better understand how ST technology serves diverse users. AI