PolyIPA -- Multilingual Phoneme-to-Grapheme Conversion Model

TL;DR

PolyIPA is a multilingual phoneme-to-grapheme conversion model that improves transliteration accuracy across languages using auxiliary models and beam search, benefiting onomastic research and information retrieval.

Contribution

The paper introduces PolyIPA, a novel multilingual conversion model with auxiliary models for data augmentation, achieving state-of-the-art performance in cross-linguistic phoneme-to-grapheme tasks.

Findings

Achieves a mean CER of 0.055 across multiple languages.

Character-level BLEU score of 0.914 indicating high translation quality.

Top-3 candidates reduce error rate by 52.7%.

Abstract

This paper presents PolyIPA, a novel multilingual phoneme-to-grapheme conversion model designed for multilingual name transliteration, onomastic research, and information retrieval. The model leverages two helper models developed for data augmentation: IPA2vec for finding soundalikes across languages, and similarIPA for handling phonetic notation variations. Evaluated on a test set that spans multiple languages and writing systems, the model achieves a mean Character Error Rate of 0.055 and a character-level BLEU score of 0.914, with particularly strong performance on languages with shallow orthographies. The implementation of beam search further improves practical utility, with top-3 candidates reducing the effective error rate by 52.7\% (to CER: 0.026), demonstrating the model's effectiveness for cross-linguistic applications.