EM can find pretty good HMM POS-Taggers (When given a good start)

Yoav Goldberg; Meni Adler; Michael Elhadad

EM can find pretty good HMM POS-Taggers (When given a good start)

Yoav Goldberg
, Meni Adler
, Michael Elhadad

Ben-Gurion University of the Negev

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

49 Scopus citations

Abstract

We address the task of unsupervised POS tagging. We demonstrate that good results can be obtained using the robust EM-HMM learner when provided with good initial conditions, even with incomplete dictionaries. We present a family of algorithms to compute effective initial estimations p(t|w). We test the method on the task of full morphological disambiguation in Hebrew achieving an error reduction of 25% over a strong uniform distribution baseline. We also test the same method on the standard WSJ unsupervised POS tagging task and obtain results competitive with recent state-ofthe- art methods, while using simple and efficient learning methods.

Original language	English
Title of host publication	ACL-08
Subtitle of host publication	HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference
Pages	746-754
Number of pages	9
State	Published - 2008
Externally published	Yes
Event	46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, ACL-08: HLT - Columbus, OH, United States Duration: 15 Jun 2008 → 20 Jun 2008

Publication series

Name	ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference

Conference

Conference	46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, ACL-08: HLT
Country/Territory	United States
City	Columbus, OH
Period	15/06/08 → 20/06/08

Bibliographical note

Funding Information:
Funding.　Funding has been provided in the form of a grant for clinical cancer research from the Ministry of Health, Labour and Welfare, Japan.

Cite this

Goldberg, Y., Adler, M., & Elhadad, M. (2008). EM can find pretty good HMM POS-Taggers (When given a good start). In ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference (pp. 746-754). (ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference).

Goldberg, Yoav ; Adler, Meni ; Elhadad, Michael. / EM can find pretty good HMM POS-Taggers (When given a good start). ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference. 2008. pp. 746-754 (ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference).

@inproceedings{ee1028b52a07473eb281a27c29591178,

title = "EM can find pretty good HMM POS-Taggers (When given a good start)",

abstract = "We address the task of unsupervised POS tagging. We demonstrate that good results can be obtained using the robust EM-HMM learner when provided with good initial conditions, even with incomplete dictionaries. We present a family of algorithms to compute effective initial estimations p(t|w). We test the method on the task of full morphological disambiguation in Hebrew achieving an error reduction of 25\% over a strong uniform distribution baseline. We also test the same method on the standard WSJ unsupervised POS tagging task and obtain results competitive with recent state-ofthe- art methods, while using simple and efficient learning methods.",

author = "Yoav Goldberg and Meni Adler and Michael Elhadad",

note = "Funding Information: Funding.　Funding has been provided in the form of a grant for clinical cancer research from the Ministry of Health, Labour and Welfare, Japan.; 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, ACL-08: HLT ; Conference date: 15-06-2008 Through 20-06-2008",

year = "2008",

language = "אנגלית",

isbn = "9781932432046",

series = "ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference",

pages = "746--754",

booktitle = "ACL-08",

}

Goldberg, Y, Adler, M & Elhadad, M 2008, EM can find pretty good HMM POS-Taggers (When given a good start). in ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference. ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference, pp. 746-754, 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, ACL-08: HLT, Columbus, OH, United States, 15/06/08.

EM can find pretty good HMM POS-Taggers (When given a good start). / Goldberg, Yoav; Adler, Meni; Elhadad, Michael.
ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference. 2008. p. 746-754 (ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference).

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

TY - GEN

T1 - EM can find pretty good HMM POS-Taggers (When given a good start)

AU - Goldberg, Yoav

AU - Adler, Meni

AU - Elhadad, Michael

N1 - Funding Information: Funding.　Funding has been provided in the form of a grant for clinical cancer research from the Ministry of Health, Labour and Welfare, Japan.

PY - 2008

Y1 - 2008

N2 - We address the task of unsupervised POS tagging. We demonstrate that good results can be obtained using the robust EM-HMM learner when provided with good initial conditions, even with incomplete dictionaries. We present a family of algorithms to compute effective initial estimations p(t|w). We test the method on the task of full morphological disambiguation in Hebrew achieving an error reduction of 25% over a strong uniform distribution baseline. We also test the same method on the standard WSJ unsupervised POS tagging task and obtain results competitive with recent state-ofthe- art methods, while using simple and efficient learning methods.

AB - We address the task of unsupervised POS tagging. We demonstrate that good results can be obtained using the robust EM-HMM learner when provided with good initial conditions, even with incomplete dictionaries. We present a family of algorithms to compute effective initial estimations p(t|w). We test the method on the task of full morphological disambiguation in Hebrew achieving an error reduction of 25% over a strong uniform distribution baseline. We also test the same method on the standard WSJ unsupervised POS tagging task and obtain results competitive with recent state-ofthe- art methods, while using simple and efficient learning methods.

UR - https://www.scopus.com/pages/publications/84859881696

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:84859881696

SN - 9781932432046

T3 - ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference

SP - 746

EP - 754

BT - ACL-08

T2 - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, ACL-08: HLT

Y2 - 15 June 2008 through 20 June 2008

ER -

Goldberg Y, Adler M, Elhadad M. EM can find pretty good HMM POS-Taggers (When given a good start). In ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference. 2008. p. 746-754. (ACL-08: HLT - 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference).

EM can find pretty good HMM POS-Taggers (When given a good start)

Abstract

Publication series

Conference

Bibliographical note

Other files and links

Fingerprint

Cite this