Unsupervised Automatic Speech Recognition: A review

Automatic Speech Recognition (ASR) systems can be trained to achieve remarkable performance given large amounts of manually transcribed speech, but labeled data sets difficult or expensive acquire for all languages interest. In this paper, we review the research literature identify models and ideas that could lead fully unsupervised ASR, including sub-word word modeling, segmentation speech signal, mapping from segments text. The objective study is limitations what learned alone understand minimum requirements recognition. Identifying these would help optimize resources efforts in ASR development low-resource languages.