Connectionist temporal classification: Labelling unsegmented sequence data with recurrent neural networks

DMCA

Connectionist temporal classification: Labelling unsegmented sequence data with recurrent neural networks (2006)

Cached

Download Links

Other Repositories/Bibliography

DBLP

by Alex Graves , Santiago Fernández , Faustino Gomez

Venue:	In Proceedings of the International Conference on Machine Learning, ICML 2006
Citations:	60 - 20 self

BibTeX

@INPROCEEDINGS{Graves06connectionisttemporal,
    author = {Alex Graves and Santiago Fernández and Faustino Gomez},
    title = {Connectionist temporal classification: Labelling unsegmented sequence data with recurrent neural networks},
    booktitle = {In Proceedings of the International Conference on Machine Learning, ICML 2006},
    year = {2006},
    pages = {369--376}
}

OpenURL

Abstract

Many real-world sequence learning tasks require the prediction of sequences of labels from noisy, unsegmented input data. In speech recognition, for example, an acoustic signal is transcribed into words or sub-word units. Recurrent neural networks (RNNs) are powerful sequence learners that would seem well suited to such tasks. However, because they require pre-segmented training data, and post-processing to transform their outputs into label sequences, their applicability has so far been limited. This paper presents a novel method for training RNNs to label unsegmented sequences directly, thereby solving both problems. An experiment on the TIMIT speech corpus demonstrates its advantages over both a baseline HMM and a hybrid HMM-RNN. 1.

Keyphrases

recurrent neural network connectionist temporal classification sequence data speech recognition label sequence powerful sequence learner unsegmented input data unsegmented sequence hybrid hmm-rnn acoustic signal sub-word unit timit speech corpus novel method many real-world sequence learning task pre-segmented training data baseline hmm

DMCA