Qiantong Xu, Tatiana Likhomanenko, Jacob Kahn, Awni Hannun, Gabriel Synnaeve, Ronan Collobert
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
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Speech Recognition | LibriSpeech test-clean | Word Error Rate (WER) | 2.1 | Conv + Transformer AM + Iterative Pseudo-Labeling (n-gram LM + Transformer Rescoring) |
| Speech Recognition | LibriSpeech test-other | Word Error Rate (WER) | 3.83 | Conv + Transformer AM + Iterative Pseudo-Labeling (n-gram LM + Transformer Rescoring) |