@inproceedings{wang-etal-2023-theory,
    title = "A Theory of Unsupervised Speech Recognition",
    author = "Wang, Liming  and
      Hasegawa-Johnson, Mark  and
      Yoo, Chang",
    editor = "Rogers, Anna  and
      Boyd-Graber, Jordan  and
      Okazaki, Naoaki",
    booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2023",
    address = "Toronto, Canada",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/Author-page-Marten-During-lu/2023.acl-long.67/",
    doi = "10.18653/v1/2023.acl-long.67",
    pages = "1192--1215",
    abstract = "Unsupervised speech recognition ({\{}pasted macro {\textquoteleft}ASRU'{\}}/) is the problem of learning automatic speech recognition (ASR) systems from \textit{unpaired} speech-only and text-only corpora. While various algorithms exist to solve this problem, a theoretical framework is missing to study their properties and address such issues as sensitivity to hyperparameters and training instability. In this paper, we proposed a general theoretical framework to study the properties of {\{}pasted macro {\textquoteleft}ASRU'{\}}/ systems based on random matrix theory and the theory of neural tangent kernels. Such a framework allows us to prove various learnability conditions and sample complexity bounds of {\{}pasted macro {\textquoteleft}ASRU'{\}}/. Extensive {\{}pasted macro {\textquoteleft}ASRU'{\}}/ experiments on synthetic languages with three classes of transition graphs provide strong empirical evidence for our theory (code available at \url{https://github.com/cactuswiththoughts/UnsupASRTheory.gitcactuswiththoughts/UnsupASRTheory.git})."
}