Breeze Taigi: Benchmarks and Models for Taiwanese Hokkien Speech Recognition and Synthesis

TL;DR

This paper introduces Breeze Taigi, a standardized benchmarking framework for Taiwanese Hokkien speech recognition and synthesis, utilizing parallel Mandarin resources and synthetic data to develop and evaluate models.

Contribution

It presents a reproducible evaluation methodology, curated datasets, and baseline models for Taigi speech technology, facilitating cross-system comparison and future research.

Findings

Achieved 30.13% CER on the benchmark

Fine-tuned Whisper model on 10,000 hours of synthetic data

Provided open datasets and baseline models for Taigi speech tasks

Abstract

Taiwanese Hokkien (Taigi) presents unique opportunities for advancing speech technology methodologies that can generalize to diverse linguistic contexts. We introduce Breeze Taigi, a comprehensive framework centered on standardized benchmarks for evaluating Taigi speech recognition and synthesis systems. Our primary contribution is a reproducible evaluation methodology that leverages parallel Taiwanese Mandarin resources. We provide 30 carefully curated Mandarin-Taigi audio pairs from Taiwan's Executive Yuan public service announcements with normalized ground truth transcriptions. We establish Character Error Rate (CER) as the standard metric and implement normalization procedures to enable fair cross-system comparisons. To demonstrate the benchmark's utility and provide reference implementations, we develop speech recognition and synthesis models through a methodology that leverages existing Taiwanese Mandarin resources and large-scale synthetic data generation. In particular, we fine-tune a Whisper model on approximately 10,000 hours of Taigi synthetic speech data. Our ASR model achieves 30.13% average CER on the benchmark, outperforming existing commercial and research systems. By providing standardized evaluation protocols, diverse training datasets, and open baseline models, we offer a replicable framework with methodologies applicable to various linguistic contexts.