Spectral Learning of Refinement HMMs

K. Stratos; A. M. Rush; S. B. Cohen; M. Collins

Spectral Learning of Refinement HMMs

K. Stratos, A. M. Rush, S. B. Cohen, M. Collins

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

We derive a spectral algorithm for learning the parameters of a refinement HMM. This method is simple, efficient, and can be applied to a wide range of supervised sequence labeling tasks. Like other spectral methods, it avoids the problem of local optima and provides a consistent estimate of the parameters. Our experiments on a phoneme recognition task show that when equipped with informative feature functions, it performs significantly better than a supervised HMM and competitively with EM.

Original language	English
Title of host publication	Proceedings of CoNLL
Pages	56-64
Number of pages	9
Publication status	Published - 2013

Access to Document

W13-3507Final published version, 357 KB

Fingerprint Dive into the research topics of 'Spectral Learning of Refinement HMMs'. Together they form a unique fingerprint.

Cite this

@inproceedings{d6a8b6a7493a4ad3868942d46890cebc,

title = "Spectral Learning of Refinement HMMs",

abstract = "We derive a spectral algorithm for learning the parameters of a refinement HMM. This method is simple, efficient, and can be applied to a wide range of supervised sequence labeling tasks. Like other spectral methods, it avoids the problem of local optima and provides a consistent estimate of the parameters. Our experiments on a phoneme recognition task show that when equipped with informative feature functions, it performs significantly better than a supervised HMM and competitively with EM. ",

author = "K. Stratos and Rush, {A. M.} and Cohen, {S. B.} and M. Collins",

year = "2013",

language = "English",

pages = "56--64",

booktitle = "Proceedings of CoNLL",

}

TY - GEN

T1 - Spectral Learning of Refinement HMMs

AU - Stratos, K.

AU - Rush, A. M.

AU - Cohen, S. B.

AU - Collins, M.

PY - 2013

Y1 - 2013

N2 - We derive a spectral algorithm for learning the parameters of a refinement HMM. This method is simple, efficient, and can be applied to a wide range of supervised sequence labeling tasks. Like other spectral methods, it avoids the problem of local optima and provides a consistent estimate of the parameters. Our experiments on a phoneme recognition task show that when equipped with informative feature functions, it performs significantly better than a supervised HMM and competitively with EM.

AB - We derive a spectral algorithm for learning the parameters of a refinement HMM. This method is simple, efficient, and can be applied to a wide range of supervised sequence labeling tasks. Like other spectral methods, it avoids the problem of local optima and provides a consistent estimate of the parameters. Our experiments on a phoneme recognition task show that when equipped with informative feature functions, it performs significantly better than a supervised HMM and competitively with EM.

M3 - Conference contribution

SP - 56

EP - 64

BT - Proceedings of CoNLL

ER -