Learning Label Embeddings for Nearest-Neighbor Multi-class Classification with an Application to Speech Recognition
Author(s)
Singh-Miller, Natasha; Collins, Michael
DownloadCollins_Learning label.pdf (98.36Kb)
OPEN_ACCESS_POLICY
Open Access Policy
Creative Commons Attribution-Noncommercial-Share Alike
Terms of use
Metadata
Show full item recordAbstract
We consider the problem of using nearest neighbor methods to provide a conditional
probability estimate, P(y|a), when the number of labels y is large and the
labels share some underlying structure. We propose a method for learning label
embeddings (similar to error-correcting output codes (ECOCs)) to model the similarity
between labels within a nearest neighbor framework. The learned ECOCs
and nearest neighbor information are used to provide conditional probability estimates.
We apply these estimates to the problem of acoustic modeling for speech
recognition. We demonstrate significant improvements in terms of word error rate
(WER) on a lecture recognition task over a state-of-the-art baseline GMM model.
Date issued
2009-12Department
Massachusetts Institute of Technology. Computer Science and Artificial Intelligence LaboratoryJournal
23rd Annual Conference on Neural Information Processing Systems, 2009
Publisher
Neural Information Processing Systems (NIPS) Foundation
Citation
Singh-Miller, Natasha and Michael Collins. "Learning Label Embeddings for Nearest-Neighbor Multi-class Classification with an Application to Speech Recognition."
Version: Author's final manuscript
ISBN
9781615679119