Sequence-Level Unsupervised Training in Speech Recognition: A Theoretical Study
Abstract
Unsupervised speech recognition is a task of training a speech recognition model with unpaired data. To determine when and how unsupervised speech recognition can succeed, and how classification error relates to candidate training objectives, we develop a theoretical framework for unsupervised speech recognition grounded in classification error bounds. We introduce two conditions under which unsupervised speech recognition is possible. The necessity of these conditions are also discussed. Under these conditions, we derive a classification error bound for unsupervised speech recognition and validate this bound in simulations. Motivated by this bound, we propose a single-stage sequence-level cross-entropy loss for unsupervised speech recognition.
Cite
@article{arxiv.2603.02285,
title = {Sequence-Level Unsupervised Training in Speech Recognition: A Theoretical Study},
author = {Zijian Yang and Jörg Barkoczi and Ralf Schlüter and Hermann Ney},
journal= {arXiv preprint arXiv:2603.02285},
year = {2026}
}
Comments
accepted to ICASSP 2026