PRiSM: Benchmarking Phone Realization in Speech Models

TL;DR

PRiSM is a comprehensive benchmark for evaluating phonetic perception in speech models, revealing key factors like language diversity and model stability that influence performance across various applications.

Contribution

It introduces PRiSM, the first open-source benchmark for phonetic perception, combining intrinsic and extrinsic evaluations to identify blind spots in current PR systems.

Findings

Diverse language exposure improves PR performance.

Encoder-CTC models are the most stable.

Specialized PR models outperform Large Audio Language Models.

Abstract

Phone recognition (PR) serves as the atomic interface for language-agnostic modeling for cross-lingual speech processing and phonetic analysis. Despite prolonged efforts in developing PR systems, current evaluations only measure surface-level transcription accuracy. We introduce PRiSM, the first open-source benchmark designed to expose blind spots in phonetic perception through intrinsic and extrinsic evaluation of PR systems. PRiSM standardizes transcription-based evaluation and assesses downstream utility in clinical, educational, and multilingual settings with transcription and representation probes. We find that diverse language exposure during training is key to PR performance, encoder-CTC models are the most stable, and specialized PR models still outperform Large Audio Language Models. PRiSM releases code, recipes, and datasets to move the field toward multilingual speech models with robust phonetic ability: https://github.com/changelinglab/prism.