Do End-to-End Speech Recognition Models Care About Context?
This work addresses the problem of understanding context sensitivity in speech recognition models for researchers and practitioners, showing incremental improvements by modifying CTC models.
The study tested whether attention-based encoder-decoder (AED) models are more context-sensitive than connectionist temporal classification (CTC) models in end-to-end speech recognition, finding that AED models are indeed more sensitive but this gap can be closed by adding self-attention to CTC models, and both models perform similarly when contextual information is constrained, with CTC models being highly competitive on WSJ and LibriSpeech without an external language model.
The two most common paradigms for end-to-end speech recognition are connectionist temporal classification (CTC) and attention-based encoder-decoder (AED) models. It has been argued that the latter is better suited for learning an implicit language model. We test this hypothesis by measuring temporal context sensitivity and evaluate how the models perform when we constrain the amount of contextual information in the audio input. We find that the AED model is indeed more context sensitive, but that the gap can be closed by adding self-attention to the CTC model. Furthermore, the two models perform similarly when contextual information is constrained. Finally, in contrast to previous research, our results show that the CTC model is highly competitive on WSJ and LibriSpeech without the help of an external language model.