SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural Network
This approach addresses the challenge of improving speech recognition accuracy across diverse tasks and datasets, showing incremental gains by combining existing data sources.
The paper tackled the problem of speech recognition by training a single large neural network, SpeechStew, on a mix of multiple publicly available datasets without special re-weighting, achieving state-of-the-art or near state-of-the-art results such as 9.0% WER on AMI-IHM and 4.7% WER on Switchboard without an external language model.
We present SpeechStew, a speech recognition model that is trained on a combination of various publicly available speech recognition datasets: AMI, Broadcast News, Common Voice, LibriSpeech, Switchboard/Fisher, Tedlium, and Wall Street Journal. SpeechStew simply mixes all of these datasets together, without any special re-weighting or re-balancing of the datasets. SpeechStew achieves SoTA or near SoTA results across a variety of tasks, without the use of an external language model. Our results include 9.0\% WER on AMI-IHM, 4.7\% WER on Switchboard, 8.3\% WER on CallHome, and 1.3\% on WSJ, which significantly outperforms prior work with strong external language models. We also demonstrate that SpeechStew learns powerful transfer learning representations. We fine-tune SpeechStew on a noisy low resource speech dataset, CHiME-6. We achieve 38.9\% WER without a language model, which compares to 38.6\% WER to a strong HMM baseline with a language model.