Attention-Based Models for Speech Recognition


Recurrent sequence generators conditioned on input data through an attention mechanism have recently shown very good performance on a range of tasks including machine translation, handwriting synthesis [1, 2] and image caption generation [3]. We extend the attention-mechanism with features needed for speech recognition. We show that while an adaptation of the model used for machine translation in [2] reaches a competitive 18.7% phoneme error rate (PER) on the TIMIT phoneme recognition task, it can only be applied to utterances which are roughly as long as the ones it was trained on. We offer a qualitative explanation of this failure and propose a novel and generic method of adding location-awareness to the attention mechanism to alleviate this issue. The new method yields a model that is robust to long inputs and achieves 18% PER in single utterances and 20% in 10-times longer (repeated) utterances. Finally, we propose a change to the attention mechanism that prevents it from concentrating too much on single frames, which further reduces PER to 17.6% level.

Extracted Key Phrases

5 Figures and Tables

Citations per Year

245 Citations

Semantic Scholar estimates that this publication has 245 citations based on the available data.

See our FAQ for additional information.

Cite this paper

@inproceedings{Chorowski2015AttentionBasedMF, title={Attention-Based Models for Speech Recognition}, author={Jan Chorowski and Dzmitry Bahdanau and Dmitriy Serdyuk and Kyunghyun Cho and Yoshua Bengio}, booktitle={NIPS}, year={2015} }