Building competitive direct acoustics-to-word models for English conversational speech recognition
This work improves A2W models for speech recognition, making them competitive with state-of-the-art methods, which is incremental but addresses a known bottleneck in simplifying training and decoding.
The paper tackles the performance gap between direct acoustics-to-word (A2W) models and conventional sub-word models in English conversational speech recognition, achieving a word error rate of 8.8%/13.9% on Hub5-2000 Switchboard/CallHome test sets without a decoder or language model.
Direct acoustics-to-word (A2W) models in the end-to-end paradigm have received increasing attention compared to conventional sub-word based automatic speech recognition models using phones, characters, or context-dependent hidden Markov model states. This is because A2W models recognize words from speech without any decoder, pronunciation lexicon, or externally-trained language model, making training and decoding with such models simple. Prior work has shown that A2W models require orders of magnitude more training data in order to perform comparably to conventional models. Our work also showed this accuracy gap when using the English Switchboard-Fisher data set. This paper describes a recipe to train an A2W model that closes this gap and is at-par with state-of-the-art sub-word based models. We achieve a word error rate of 8.8%/13.9% on the Hub5-2000 Switchboard/CallHome test sets without any decoder or language model. We find that model initialization, training data order, and regularization have the most impact on the A2W model performance. Next, we present a joint word-character A2W model that learns to first spell the word and then recognize it. This model provides a rich output to the user instead of simple word hypotheses, making it especially useful in the case of words unseen or rarely-seen during training.