A novel framework of training hidden Markov support vector machines from lightly-annotated data

Deyu Zhou, Yulan He

Research output: Chapter in Book/Report/Conference proceedingOther chapter contribution

Abstract

Natural language understanding (NLU) aims to map sentences to their semantic mean representations. Statistical approaches to NLU normally require fully-annotated training data where each sentence is paired with its word-level semantic annotations. In this paper, we propose a novel learning framework which trains the Hidden Markov Support Vector Machines (HM-SVMs) without the use of expensive fully-annotated data. In particular, our learning approach takes as input a training set of sentences labeled with abstract semantic annotations encoding underlying embedded structural relations and automatically induces derivation rules that map sentences to their semantic meaning representations. The proposed approach has been tested on the DARPA Communicator Data and achieved 93.18% in F-measure, which outperforms the previously proposed approaches of training the hidden vector state model or conditional random fields from unaligned data, with a relative error reduction rate of 43.3% and 10.6% being achieved.
Original languageEnglish
Title of host publicationProceeding : CIKM '11 proceedings of the 20th ACM international conference on Information and knowledge management
EditorsBettina Berendt, Arjen de Vries, Wenfei Fan, Craig Macdonald, Iadh Ounis, Ian Ruthven
Place of PublicationNew York (US)
PublisherACM
Pages2025-2028
Number of pages4
ISBN (Print)978-1-4503-0717-8
DOIs
Publication statusPublished - 2011
Event20th ACM international conference on Information and knowledge management, CIKM '11 - Glasgow, United Kingdom
Duration: 24 Oct 201128 Oct 2011

Conference

Conference20th ACM international conference on Information and knowledge management, CIKM '11
CountryUnited Kingdom
CityGlasgow
Period24/10/1128/10/11

Fingerprint

Support vector machines
Semantics

Cite this

Zhou, D., & He, Y. (2011). A novel framework of training hidden Markov support vector machines from lightly-annotated data. In B. Berendt, A. de Vries, W. Fan, C. Macdonald, I. Ounis, & I. Ruthven (Eds.), Proceeding : CIKM '11 proceedings of the 20th ACM international conference on Information and knowledge management (pp. 2025-2028). New York (US): ACM. https://doi.org/10.1145/2063576.2063881
Zhou, Deyu ; He, Yulan. / A novel framework of training hidden Markov support vector machines from lightly-annotated data. Proceeding : CIKM '11 proceedings of the 20th ACM international conference on Information and knowledge management . editor / Bettina Berendt ; Arjen de Vries ; Wenfei Fan ; Craig Macdonald ; Iadh Ounis ; Ian Ruthven. New York (US) : ACM, 2011. pp. 2025-2028
@inbook{629f92f2e4aa4eb289f01d85f7eca3d2,
title = "A novel framework of training hidden Markov support vector machines from lightly-annotated data",
abstract = "Natural language understanding (NLU) aims to map sentences to their semantic mean representations. Statistical approaches to NLU normally require fully-annotated training data where each sentence is paired with its word-level semantic annotations. In this paper, we propose a novel learning framework which trains the Hidden Markov Support Vector Machines (HM-SVMs) without the use of expensive fully-annotated data. In particular, our learning approach takes as input a training set of sentences labeled with abstract semantic annotations encoding underlying embedded structural relations and automatically induces derivation rules that map sentences to their semantic meaning representations. The proposed approach has been tested on the DARPA Communicator Data and achieved 93.18{\%} in F-measure, which outperforms the previously proposed approaches of training the hidden vector state model or conditional random fields from unaligned data, with a relative error reduction rate of 43.3{\%} and 10.6{\%} being achieved.",
author = "Deyu Zhou and Yulan He",
year = "2011",
doi = "10.1145/2063576.2063881",
language = "English",
isbn = "978-1-4503-0717-8",
pages = "2025--2028",
editor = "Bettina Berendt and {de Vries}, Arjen and Wenfei Fan and Craig Macdonald and Iadh Ounis and Ian Ruthven",
booktitle = "Proceeding : CIKM '11 proceedings of the 20th ACM international conference on Information and knowledge management",
publisher = "ACM",
address = "United States",

}

Zhou, D & He, Y 2011, A novel framework of training hidden Markov support vector machines from lightly-annotated data. in B Berendt, A de Vries, W Fan, C Macdonald, I Ounis & I Ruthven (eds), Proceeding : CIKM '11 proceedings of the 20th ACM international conference on Information and knowledge management . ACM, New York (US), pp. 2025-2028, 20th ACM international conference on Information and knowledge management, CIKM '11, Glasgow, United Kingdom, 24/10/11. https://doi.org/10.1145/2063576.2063881

A novel framework of training hidden Markov support vector machines from lightly-annotated data. / Zhou, Deyu; He, Yulan.

Proceeding : CIKM '11 proceedings of the 20th ACM international conference on Information and knowledge management . ed. / Bettina Berendt; Arjen de Vries; Wenfei Fan; Craig Macdonald; Iadh Ounis; Ian Ruthven. New York (US) : ACM, 2011. p. 2025-2028.

Research output: Chapter in Book/Report/Conference proceedingOther chapter contribution

TY - CHAP

T1 - A novel framework of training hidden Markov support vector machines from lightly-annotated data

AU - Zhou, Deyu

AU - He, Yulan

PY - 2011

Y1 - 2011

N2 - Natural language understanding (NLU) aims to map sentences to their semantic mean representations. Statistical approaches to NLU normally require fully-annotated training data where each sentence is paired with its word-level semantic annotations. In this paper, we propose a novel learning framework which trains the Hidden Markov Support Vector Machines (HM-SVMs) without the use of expensive fully-annotated data. In particular, our learning approach takes as input a training set of sentences labeled with abstract semantic annotations encoding underlying embedded structural relations and automatically induces derivation rules that map sentences to their semantic meaning representations. The proposed approach has been tested on the DARPA Communicator Data and achieved 93.18% in F-measure, which outperforms the previously proposed approaches of training the hidden vector state model or conditional random fields from unaligned data, with a relative error reduction rate of 43.3% and 10.6% being achieved.

AB - Natural language understanding (NLU) aims to map sentences to their semantic mean representations. Statistical approaches to NLU normally require fully-annotated training data where each sentence is paired with its word-level semantic annotations. In this paper, we propose a novel learning framework which trains the Hidden Markov Support Vector Machines (HM-SVMs) without the use of expensive fully-annotated data. In particular, our learning approach takes as input a training set of sentences labeled with abstract semantic annotations encoding underlying embedded structural relations and automatically induces derivation rules that map sentences to their semantic meaning representations. The proposed approach has been tested on the DARPA Communicator Data and achieved 93.18% in F-measure, which outperforms the previously proposed approaches of training the hidden vector state model or conditional random fields from unaligned data, with a relative error reduction rate of 43.3% and 10.6% being achieved.

UR - http://www.scopus.com/inward/record.url?scp=83055165967&partnerID=8YFLogxK

UR - http://dl.acm.org/citation.cfm?id=2063576.2063881&coll=DL&dl=ACM&CFID=173576560&CFTOKEN=50339695

U2 - 10.1145/2063576.2063881

DO - 10.1145/2063576.2063881

M3 - Other chapter contribution

AN - SCOPUS:83055165967

SN - 978-1-4503-0717-8

SP - 2025

EP - 2028

BT - Proceeding : CIKM '11 proceedings of the 20th ACM international conference on Information and knowledge management

A2 - Berendt, Bettina

A2 - de Vries, Arjen

A2 - Fan, Wenfei

A2 - Macdonald, Craig

A2 - Ounis, Iadh

A2 - Ruthven, Ian

PB - ACM

CY - New York (US)

ER -

Zhou D, He Y. A novel framework of training hidden Markov support vector machines from lightly-annotated data. In Berendt B, de Vries A, Fan W, Macdonald C, Ounis I, Ruthven I, editors, Proceeding : CIKM '11 proceedings of the 20th ACM international conference on Information and knowledge management . New York (US): ACM. 2011. p. 2025-2028 https://doi.org/10.1145/2063576.2063881