End-to-End Attention-based Large Vocabulary Speech Recognition
This work addresses speech recognition for general applications by proposing an end-to-end method, but it is incremental as it builds on existing HMM-free RNN approaches.
The paper tackled large vocabulary speech recognition by replacing HMMs with an RNN using an attention mechanism for character-level sequence prediction, achieving recognition accuracies similar to other HMM-free RNN-based approaches.
Many of the current state-of-the-art Large Vocabulary Continuous Speech Recognition Systems (LVCSR) are hybrids of neural networks and Hidden Markov Models (HMMs). Most of these systems contain separate components that deal with the acoustic modelling, language modelling and sequence decoding. We investigate a more direct approach in which the HMM is replaced with a Recurrent Neural Network (RNN) that performs sequence prediction directly at the character level. Alignment between the input features and the desired character sequence is learned automatically by an attention mechanism built into the RNN. For each predicted character, the attention mechanism scans the input sequence and chooses relevant frames. We propose two methods to speed up this operation: limiting the scan to a subset of most promising frames and pooling over time the information contained in neighboring frames, thereby reducing source sequence length. Integrating an n-gram language model into the decoding process yields recognition accuracies similar to other HMM-free RNN-based approaches.