Iterative Pseudo-Labeling for Speech Recognition
This addresses the problem of enhancing speech recognition accuracy for applications in semi-supervised and low-resource scenarios, representing an incremental improvement over existing pseudo-labeling methods.
The paper tackles improving automatic speech recognition by proposing Iterative Pseudo-Labeling, which achieves state-of-the-art word-error rates on Librispeech test sets, including in low-resource settings.
Pseudo-labeling has recently shown promise in end-to-end automatic speech recognition (ASR). We study Iterative Pseudo-Labeling (IPL), a semi-supervised algorithm which efficiently performs multiple iterations of pseudo-labeling on unlabeled data as the acoustic model evolves. In particular, IPL fine-tunes an existing model at each iteration using both labeled data and a subset of unlabeled data. We study the main components of IPL: decoding with a language model and data augmentation. We then demonstrate the effectiveness of IPL by achieving state-of-the-art word-error rate on the Librispeech test sets in both standard and low-resource setting. We also study the effect of language models trained on different corpora to show IPL can effectively utilize additional text. Finally, we release a new large in-domain text corpus which does not overlap with the Librispeech training transcriptions to foster research in low-resource, semi-supervised ASR