High-order low-rank tensors for semantic role labeling

Tao Lei, Yuan Zhang, Lluis Marques, Alessandro Moschitti, Regina Barzilay

Research output: Chapter in Book/Report/Conference proceedingConference contribution

19 Citations (Scopus)

Abstract

This paper introduces a tensor-based approach to semantic role labeling (SRL). The motivation behind the approach is to automatically induce a compact feature representation for words and their relations, tailoring them to the task. In this sense, our dimensionality reduction method provides a clear alternative to the traditional feature engineering approach used in SRL. To capture meaningful interactions between the argument, predicate, their syntactic path and the corresponding role label, we compress each feature representation first to a lower dimensional space prior to assessing their interactions. This corresponds to using an overall cross-product feature representation and maintaining associated parameters as a four-way low-rank tensor. The tensor parameters are optimized for the SRL performance using standard online algorithms. Our tensor-based approach rivals the best performing system on the CoNLL-2009 shared task. In addition, we demonstrate that adding the representation tensor to a competitive tensorfree model yields 2% absolute increase in Fscore.

Original languageEnglish
Title of host publicationNAACL HLT 2015 - 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference
PublisherAssociation for Computational Linguistics (ACL)
Pages1150-1160
Number of pages11
ISBN (Print)9781941643495
Publication statusPublished - 2015
EventConference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL HLT 2015 - Denver, United States
Duration: 31 May 20155 Jun 2015

Other

OtherConference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL HLT 2015
CountryUnited States
CityDenver
Period31/5/155/6/15

Fingerprint

Labeling
Tensors
Semantics
semantics
performance standard
interaction
Syntactics
Labels
engineering
Semantic Roles
Interaction

ASJC Scopus subject areas

  • Computer Science Applications
  • Language and Linguistics
  • Linguistics and Language

Cite this

Lei, T., Zhang, Y., Marques, L., Moschitti, A., & Barzilay, R. (2015). High-order low-rank tensors for semantic role labeling. In NAACL HLT 2015 - 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference (pp. 1150-1160). Association for Computational Linguistics (ACL).

High-order low-rank tensors for semantic role labeling. / Lei, Tao; Zhang, Yuan; Marques, Lluis; Moschitti, Alessandro; Barzilay, Regina.

NAACL HLT 2015 - 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference. Association for Computational Linguistics (ACL), 2015. p. 1150-1160.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Lei, T, Zhang, Y, Marques, L, Moschitti, A & Barzilay, R 2015, High-order low-rank tensors for semantic role labeling. in NAACL HLT 2015 - 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference. Association for Computational Linguistics (ACL), pp. 1150-1160, Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL HLT 2015, Denver, United States, 31/5/15.
Lei T, Zhang Y, Marques L, Moschitti A, Barzilay R. High-order low-rank tensors for semantic role labeling. In NAACL HLT 2015 - 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference. Association for Computational Linguistics (ACL). 2015. p. 1150-1160
Lei, Tao ; Zhang, Yuan ; Marques, Lluis ; Moschitti, Alessandro ; Barzilay, Regina. / High-order low-rank tensors for semantic role labeling. NAACL HLT 2015 - 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference. Association for Computational Linguistics (ACL), 2015. pp. 1150-1160
@inproceedings{e137c2cea6fa4e56aaaf9bf40e67ba3c,
title = "High-order low-rank tensors for semantic role labeling",
abstract = "This paper introduces a tensor-based approach to semantic role labeling (SRL). The motivation behind the approach is to automatically induce a compact feature representation for words and their relations, tailoring them to the task. In this sense, our dimensionality reduction method provides a clear alternative to the traditional feature engineering approach used in SRL. To capture meaningful interactions between the argument, predicate, their syntactic path and the corresponding role label, we compress each feature representation first to a lower dimensional space prior to assessing their interactions. This corresponds to using an overall cross-product feature representation and maintaining associated parameters as a four-way low-rank tensor. The tensor parameters are optimized for the SRL performance using standard online algorithms. Our tensor-based approach rivals the best performing system on the CoNLL-2009 shared task. In addition, we demonstrate that adding the representation tensor to a competitive tensorfree model yields 2{\%} absolute increase in Fscore.",
author = "Tao Lei and Yuan Zhang and Lluis Marques and Alessandro Moschitti and Regina Barzilay",
year = "2015",
language = "English",
isbn = "9781941643495",
pages = "1150--1160",
booktitle = "NAACL HLT 2015 - 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference",
publisher = "Association for Computational Linguistics (ACL)",

}

TY - GEN

T1 - High-order low-rank tensors for semantic role labeling

AU - Lei, Tao

AU - Zhang, Yuan

AU - Marques, Lluis

AU - Moschitti, Alessandro

AU - Barzilay, Regina

PY - 2015

Y1 - 2015

N2 - This paper introduces a tensor-based approach to semantic role labeling (SRL). The motivation behind the approach is to automatically induce a compact feature representation for words and their relations, tailoring them to the task. In this sense, our dimensionality reduction method provides a clear alternative to the traditional feature engineering approach used in SRL. To capture meaningful interactions between the argument, predicate, their syntactic path and the corresponding role label, we compress each feature representation first to a lower dimensional space prior to assessing their interactions. This corresponds to using an overall cross-product feature representation and maintaining associated parameters as a four-way low-rank tensor. The tensor parameters are optimized for the SRL performance using standard online algorithms. Our tensor-based approach rivals the best performing system on the CoNLL-2009 shared task. In addition, we demonstrate that adding the representation tensor to a competitive tensorfree model yields 2% absolute increase in Fscore.

AB - This paper introduces a tensor-based approach to semantic role labeling (SRL). The motivation behind the approach is to automatically induce a compact feature representation for words and their relations, tailoring them to the task. In this sense, our dimensionality reduction method provides a clear alternative to the traditional feature engineering approach used in SRL. To capture meaningful interactions between the argument, predicate, their syntactic path and the corresponding role label, we compress each feature representation first to a lower dimensional space prior to assessing their interactions. This corresponds to using an overall cross-product feature representation and maintaining associated parameters as a four-way low-rank tensor. The tensor parameters are optimized for the SRL performance using standard online algorithms. Our tensor-based approach rivals the best performing system on the CoNLL-2009 shared task. In addition, we demonstrate that adding the representation tensor to a competitive tensorfree model yields 2% absolute increase in Fscore.

UR - http://www.scopus.com/inward/record.url?scp=84960155837&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84960155837&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84960155837

SN - 9781941643495

SP - 1150

EP - 1160

BT - NAACL HLT 2015 - 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference

PB - Association for Computational Linguistics (ACL)

ER -