Search papers, labs, and topics across Lattice.
The paper introduces PhoneticXEUS, a phone recognition model trained on large-scale multilingual data, achieving state-of-the-art performance on both multilingual and accented English speech. Through controlled ablations across 100+ languages, the authors quantify the impact of self-supervised learning (SSL) representations, data scale, and loss objectives on phone recognition performance. The study provides an empirical recipe for universal phone recognition and analyzes error patterns across diverse linguistic contexts.
Forget hand-tuning for each language: this recipe achieves state-of-the-art phone recognition across 100+ languages, revealing the surprising power of scaling data and SSL representations.
Phone recognition (PR) is a key enabler of multilingual and low-resource speech processing tasks, yet robust performance remains elusive. Highly performant English-focused models do not generalize across languages, while multilingual models underutilize pretrained representations. It also remains unclear how data scale, architecture, and training objective contribute to multilingual PR. We present PhoneticXEUS -- trained on large-scale multilingual data and achieving state-of-the-art performance on both multilingual (17.7% PFER) and accented English speech (10.6% PFER). Through controlled ablations with evaluations across 100+ languages under a unified scheme, we empirically establish our training recipe and quantify the impact of SSL representations, data scale, and loss objectives. In addition, we analyze error patterns across language families, accented speech, and articulatory features. All data and code are released openly.
CMU ML