Improving Speech-to-Speech Translation Through Unlabeled Text
This addresses the challenge of limited training data for speech-to-speech translation systems, particularly benefiting low-resource language pairs, though it is incremental as it builds on existing cascaded model approaches.
The paper tackles the problem of data scarcity in direct speech-to-speech translation by leveraging unlabeled text to generate synthetic training data, achieving up to a 2 BLEU improvement in Spanish-English translation and showing gains in low-resource settings.
Direct speech-to-speech translation (S2ST) is among the most challenging problems in the translation paradigm due to the significant scarcity of S2ST data. While effort has been made to increase the data size from unlabeled speech by cascading pretrained speech recognition (ASR), machine translation (MT) and text-to-speech (TTS) models; unlabeled text has remained relatively under-utilized to improve S2ST. We propose an effective way to utilize the massive existing unlabeled text from different languages to create a large amount of S2ST data to improve S2ST performance by applying various acoustic effects to the generated synthetic data. Empirically our method outperforms the state of the art in Spanish-English translation by up to 2 BLEU. Significant gains by the proposed method are demonstrated in extremely low-resource settings for both Spanish-English and Russian-English translations.