Multi-view discriminative sequential learning

Ulf Brefeld; Christoph Büscher; Tobias Scheffer

doi:10.1007/11564096_11

Multi-view discriminative sequential learning

Ulf Brefeld^*
, Christoph Büscher
, Tobias Scheffer

^*Korrespondierende/r Autor/-in für diese Arbeit

Humboldt-Universität zu Berlin

Publikation: Beiträge in Sammelwerken › Aufsätze in Konferenzbänden › Forschung › Begutachtung

31 Zitate (Scopus)

Abstract

Discriminative learning techniques for sequential data have proven to be more effective than generative models for named entity recognition, information extraction, and other tasks of discrimination. However, semi-supervised learning mechanisms that utilize inexpensive unlabeled sequences in addition to few labeled sequences - such as the Baum-Welch algorithm - are available only for generative models. The multi-view approach is based on the principle of maximizing the consensus among multiple independent hypotheses; we develop this principle into a semi-supervised hidden Markov perceptron, and a semi-supervised hidden Markov support vector learning algorithm. Experiments reveal that the resulting procedures utilize unlabeled data effectively and discriminate more accurately than their purely supervised counterparts

Originalsprache	Englisch
Titel	Machine Learning: ECML 2005 : 16th European Conference on Machine Learning
Seitenumfang	12
Herausgeber (Verlag)	Springer Verlag
Erscheinungsdatum	01.01.2005
Seiten	60-71
ISBN (Print)	978-3-540-29243-2
ISBN (elektronisch)	978-3-540-31692-3
DOIs	https://doi.org/10.1007/11564096_11
Publikationsstatus	Erschienen - 01.01.2005
Extern publiziert	Ja
Veranstaltung	16th European Conference on Machine Learning and 9th European Conference on Principles and Practice of Knowledge Discovery in Databases - ECML PKDD 2005 - Porto, Portugal Dauer: 03.10.2005 → 07.10.2005 Konferenznummer: 16 http://www.inescporto.pt/~jgama/ecmlpkdd05/

Fachgebiete und Schlagwörter

Informatik
Wirtschaftsinformatik

Zugriff auf Dokument

10.1007/11564096_11Lizenz: Urheberrechtlich geschützt

Dieses zitieren

Brefeld, U., Büscher, C., & Scheffer, T. (2005). Multi-view discriminative sequential learning. in Machine Learning: ECML 2005: 16th European Conference on Machine Learning (S. 60-71). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Band 3720 LNAI). Springer Verlag. https://doi.org/10.1007/11564096_11

@inbook{6fcea8700db74e138fa597ba9e24dd2d,

title = "Multi-view discriminative sequential learning",

abstract = "Discriminative learning techniques for sequential data have proven to be more effective than generative models for named entity recognition, information extraction, and other tasks of discrimination. However, semi-supervised learning mechanisms that utilize inexpensive unlabeled sequences in addition to few labeled sequences - such as the Baum-Welch algorithm - are available only for generative models. The multi-view approach is based on the principle of maximizing the consensus among multiple independent hypotheses; we develop this principle into a semi-supervised hidden Markov perceptron, and a semi-supervised hidden Markov support vector learning algorithm. Experiments reveal that the resulting procedures utilize unlabeled data effectively and discriminate more accurately than their purely supervised counterparts.",

keywords = "Informatics, Unlabeled Data, Neural Information Processing System, Name Entity Recognition, Entity Recognition, Word Sense Disambiguation, Business informatics",

author = "Ulf Brefeld and Christoph B{\"u}scher and Tobias Scheffer",

note = "Funding Information: Thanks to NICT for their support, Takayuki Kuribayashi for providing native judgments, and Marcus Dickinson for comments on an early draft.; 16th European Conference on Machine Learning, ECML PKDD 2005 ; Conference date: 03-10-2005 Through 07-10-2005",

year = "2005",

month = jan,

day = "1",

doi = "10.1007/11564096\_11",

language = "English",

isbn = "978-3-540-29243-2",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "60--71",

booktitle = "Machine Learning: ECML 2005",

address = "Germany",

url = "http://www.inescporto.pt/\textasciitilde{}jgama/ecmlpkdd05/",

}

Brefeld, U, Büscher, C & Scheffer, T 2005, Multi-view discriminative sequential learning. in Machine Learning: ECML 2005: 16th European Conference on Machine Learning. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Bd. 3720 LNAI, Springer Verlag, S. 60-71, 16th European Conference on Machine Learning and 9th European Conference on Principles and Practice of Knowledge Discovery in Databases - ECML PKDD 2005, Porto, Portugal, 03.10.05. https://doi.org/10.1007/11564096_11

Multi-view discriminative sequential learning. / Brefeld, Ulf; Büscher, Christoph; Scheffer, Tobias.
Machine Learning: ECML 2005: 16th European Conference on Machine Learning. Springer Verlag, 2005. S. 60-71 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Band 3720 LNAI).

Publikation: Beiträge in Sammelwerken › Aufsätze in Konferenzbänden › Forschung › Begutachtung

TY - CHAP

T1 - Multi-view discriminative sequential learning

AU - Brefeld, Ulf

AU - Büscher, Christoph

AU - Scheffer, Tobias

N1 - Conference code: 16

PY - 2005/1/1

Y1 - 2005/1/1

N2 - Discriminative learning techniques for sequential data have proven to be more effective than generative models for named entity recognition, information extraction, and other tasks of discrimination. However, semi-supervised learning mechanisms that utilize inexpensive unlabeled sequences in addition to few labeled sequences - such as the Baum-Welch algorithm - are available only for generative models. The multi-view approach is based on the principle of maximizing the consensus among multiple independent hypotheses; we develop this principle into a semi-supervised hidden Markov perceptron, and a semi-supervised hidden Markov support vector learning algorithm. Experiments reveal that the resulting procedures utilize unlabeled data effectively and discriminate more accurately than their purely supervised counterparts.

AB - Discriminative learning techniques for sequential data have proven to be more effective than generative models for named entity recognition, information extraction, and other tasks of discrimination. However, semi-supervised learning mechanisms that utilize inexpensive unlabeled sequences in addition to few labeled sequences - such as the Baum-Welch algorithm - are available only for generative models. The multi-view approach is based on the principle of maximizing the consensus among multiple independent hypotheses; we develop this principle into a semi-supervised hidden Markov perceptron, and a semi-supervised hidden Markov support vector learning algorithm. Experiments reveal that the resulting procedures utilize unlabeled data effectively and discriminate more accurately than their purely supervised counterparts.

KW - Informatics

KW - Unlabeled Data

KW - Neural Information Processing System

KW - Name Entity Recognition

KW - Entity Recognition

KW - Word Sense Disambiguation

KW - Business informatics

UR - https://www.scopus.com/pages/publications/33646415916

UR - https://www.mendeley.com/catalogue/f5846d33-6389-3b3a-b23a-bdde947af373/

U2 - 10.1007/11564096_11

DO - 10.1007/11564096_11

M3 - Article in conference proceedings

AN - SCOPUS:33646415916

SN - 978-3-540-29243-2

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 60

EP - 71

BT - Machine Learning: ECML 2005

PB - Springer Verlag

T2 - 16th European Conference on Machine Learning

Y2 - 3 October 2005 through 7 October 2005

ER -

Multi-view discriminative sequential learning

Abstract

Fachgebiete und Schlagwörter

Zugriff auf Dokument

Fingerprint

Dieses zitieren