Word stress in self-supervised speech models: A cross-linguistic comparison

TL;DR

This study examines how self-supervised speech models, specifically Wav2vec 2.0, learn and represent word stress across five languages, revealing language-specific stress patterns and high accuracy in stress classification.

Contribution

It provides a cross-linguistic analysis of stress representations in self-supervised speech models, highlighting language-specific stress encoding.

Findings

Stress classifiers distinguish stressed and unstressed syllables with high accuracy.

Word stress representations are language-specific.

Greater differences observed between variable and fixed stress languages.

Abstract

In this paper we study word stress representations learned by self-supervised speech models (S3M), specifically the Wav2vec 2.0 model. We investigate the S3M representations of word stress for five different languages: Three languages with variable or lexical stress (Dutch, English and German) and two languages with fixed or demarcative stress (Hungarian and Polish). We train diagnostic stress classifiers on S3M embeddings and show that they can distinguish between stressed and unstressed syllables in read-aloud short sentences with high accuracy. We also tested language-specificity effects of S3M word stress. The results indicate that the word stress representations are language-specific, with a greater difference between the set of variable versus the set of fixed stressed languages.