DiscoPhon: Benchmarking the Unsupervised Discovery of Phoneme Inventories With Discrete Speech Units

TL;DR

DiscoPhon is a multilingual benchmark designed to evaluate unsupervised phoneme discovery from speech units, assessing model performance across diverse languages with limited training data.

Contribution

This paper introduces DiscoPhon, a new benchmark for evaluating unsupervised phoneme discovery in multiple languages using discrete speech units.

Findings

Current models' units correlate well with phonemes across languages

Phonemic information is sufficiently present in pretrained models

Variations in performance exist across different languages

Abstract

We introduce DiscoPhon, a multilingual benchmark for evaluating unsupervised phoneme discovery from discrete speech units. DiscoPhon covers 6 dev and 6 test languages, chosen to span a wide range of phonemic contrasts. Given only 10 hours of speech in a previously unseen language, systems must produce discrete units that are mapped to a predefined phoneme inventory, through either a many-to-one or a one-to-one assignment. The resulting sequences are evaluated for unit quality, recognition and segmentation. We provide four pretrained multilingual HuBERT and SpidR baselines, and show that phonemic information is available enough in current models for derived units to correlate well with phonemes, though with variations across languages.