Irish text-to-speech improved using pronunciation guides and small data
Pronunciation-Supervised Text-to-Speech for Irish (Gaeilge)
Computation and Language
Summary
Making computers speak Irish well is hard because there is little recorded speech for training. The authors show it helps to use written pronunciation guides from dictionaries to teach a powerful speech model how to pronounce Irish words. They fine-tuned a large pre-existing model with a small amount of Irish speech and improved accuracy by checking pronunciations against dictionary rules. This approach also measures where spoken data is missing, like rare dialects.
What this means in practice
- •For software developers: Build better Irish speech synthesis systems by combining small speech sets with dictionary pronunciation rules.
- •For language technology companies: Create more accurate text-to-speech voices for low-resource languages using existing pronunciation documentation as quality checks.$Commercial implications: Enables improved commercial Irish speech products by leveraging pronunciation data to enhance pretrained models with limited audio.
Authors
William Lane
Abstract
Speech technology for Irish must be built from very little transcribed speech, yet the language's pronunciation is thoroughly documented in dictionaries and dialect descriptions. We ask whether that documentation can supervise the adaptation of a large pretrained text-to-speech model to a language it was never trained on. We adapt Qwen3-TTS to Irish using 12.8 hours of crowdsourced Common Voice speech, correcting two defects in the vendor fine-tuning script that training loss does not reveal (Appendix A), and reach character error rate (CER) 0.225 on held-out cross-corpus text, against 0.397 for an exhaustively tuned from-scratch system and 0.437 for zero-shot cloning. We then compile the Wiktionary pronunciation record into a lexicon and a phoneme critic, and use that critic as the verifier in one round of rejection fine-tuning: CER 0.159 cross-corpus and 0.067 in-domain, with the voice still anonymous across all 202 training speakers. The same critic explains why the voice cannot yet be steered to a minority dialect: 229 of 12,311 training clips lean Ulster against 8,451 Munster. Pronunciation documentation, converted into a verifier, is a practical lever for low-resource speech synthesis, and a way to measure what the available data cannot support.