Improving LSTM-CTC based ASR performance in domains with limited training data
This incremental improvement addresses ASR performance for domains with limited training data.
The paper tackled the performance gap between LSTM-CTC and hybrid DNN-CE ASR systems in data-limited domains, showing that with data augmentation and regularization, LSTM-CTC can exceed a strong baseline on the Librispeech 100hr corpus.
This paper addresses the observed performance gap between automatic speech recognition (ASR) systems based on Long Short Term Memory (LSTM) neural networks trained with the connectionist temporal classification (CTC) loss function and systems based on hybrid Deep Neural Networks (DNNs) trained with the cross entropy (CE) loss function on domains with limited data. We step through a number of experiments that show incremental improvements on a baseline EESEN toolkit based LSTM-CTC ASR system trained on the Librispeech 100hr (train-clean-100) corpus. Our results show that with effective combination of data augmentation and regularization, a LSTM-CTC based system can exceed the performance of a strong Kaldi based baseline trained on the same data.