Fast, Not Fancy: Rethinking G2P with Rich Data and Rule-Based Models
Mahta Fetrat Qharabagh, Zahra Dehghanian, Hamid R. Rabiee
2025-05-20
Summary
This paper talks about a new approach to converting written words into how they sound, especially for languages that don't have a lot of digital resources, by using a combination of rules and data.
What's the problem?
The problem is that for many less common languages, it's hard to accurately figure out how to pronounce written words, especially when the same spelling can have different meanings and sounds depending on the context.
What's the solution?
To solve this, the researchers created a semi-automated system that uses both detailed language rules and rich data to quickly and accurately decide how to pronounce tricky words, making the process much faster and improving accuracy by a large margin.
Why it matters?
This matters because it helps people who speak less common languages use technology like speech recognition and text-to-speech, making digital tools more inclusive and useful for everyone.
Abstract
A semi-automated pipeline and a fast rule-based system enhance homograph disambiguation in grapheme-to-phoneme conversion for low-resource languages, improving accuracy by 30%.