Strategies in Transfer Learning for Low-Resource Speech Synthesis: Phone Mapping, Features Input, and Source Language Selection
Abstract
We compare using a PHOIBLE-based phone mapping method and using phonological features input in transfer learning for TTS in low-resource languages. We use diverse source languages (English, Finnish, Hindi, Japanese, and Russian) and target languages (Bulgarian, Georgian, Kazakh, Swahili, Urdu, and Uzbek) to test the language-independence of the methods and enhance the findings’ applicability. We use Character Error Rates from automatic speech recognition and predicted Mean Opinion Scores for evaluation. Results show that both phone mapping and features input improve the output quality and the latter performs better, but these effects also depend on the specific language combination. We also compare the recently-proposed Angular Similarity of Phone Frequencies (ASPF) with a family tree-based distance measure as a criterion to select source languages in transfer learning. ASPF proves effective if label-based phone input is used, while the language distance does not have expected effects.
Type
Publication
arXiv preprint arXiv:2306.12040
This paper was accepted at the Speech Synthesis Workshop 2023. It compares different approaches for transfer learning in text-to-speech systems for low-resource languages, including phone mapping methods and the use of phonological features as input. The study also evaluates criteria for selecting source languages in transfer learning scenarios.
