PolyNorm: Few-Shot LLM-Based Text Normalization for Text-to-Speech

TL;DR

PolyNorm introduces a prompt-based, multilingual text normalization method using LLMs that reduces manual effort and improves accuracy across diverse languages for TTS systems.

Contribution

It presents a novel LLM-based, language-agnostic approach to text normalization with an automatic data pipeline, enabling scalable, low-resource language coverage.

Findings

Consistent WER reductions across eight languages

Effective in low-resource language settings

Provides a multilingual benchmark dataset

Abstract

Text Normalization (TN) is a key preprocessing step in Text-to-Speech (TTS) systems, converting written forms into their canonical spoken equivalents. Traditional TN systems can exhibit high accuracy, but involve substantial engineering effort, are difficult to scale, and pose challenges to language coverage, particularly in low-resource settings. We propose PolyNorm, a prompt-based approach to TN using Large Language Models (LLMs), aiming to reduce the reliance on manually crafted rules and enable broader linguistic applicability with minimal human intervention. Additionally, we present a language-agnostic pipeline for automatic data curation and evaluation, designed to facilitate scalable experimentation across diverse languages. Experiments across eight languages show consistent reductions in the word error rate (WER) compared to a production-grade-based system. To support further research, we release PolyNorm-Benchmark, a multilingual data set covering a diverse range of text normalization phenomena.