On reverse feature engineering of syntactic tree kernels

Daniele Pighin, Alessandro Moschitti

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

11 Citations (Scopus)

Abstract

In this paper, we provide a theoretical framework for feature selection in tree kernel spaces based on gradient-vector components of kernel-based machines. We show that a huge number of features can be discarded without a significant decrease in accuracy. Our selection algorithm is as accurate as and much more efficient than those proposed in previous work. Comparative experiments on three interesting and very diverse classification tasks, i.e. Question Classification, Relation Extraction and Semantic Role Labeling, support our theoretical findings and demonstrate the algorithm performance.

Original languageEnglish
Title of host publicationCoNLL 2010 - Fourteenth Conference on Computational Natural Language Learning, Proceedings of the Conference
Pages223-233
Number of pages11
Publication statusPublished - 1 Dec 2010
Externally publishedYes
Event14th Conference on Computational Natural Language Learning, CoNLL 2010 - Uppsala, Sweden
Duration: 15 Jul 201016 Jul 2010

Other

Other14th Conference on Computational Natural Language Learning, CoNLL 2010
CountrySweden
CityUppsala
Period15/7/1016/7/10

Fingerprint

Syntactics
engineering
Labeling
Feature extraction
Semantics
semantics
experiment
performance
Experiments

ASJC Scopus subject areas

  • Artificial Intelligence
  • Human-Computer Interaction
  • Linguistics and Language

Cite this

Pighin, D., & Moschitti, A. (2010). On reverse feature engineering of syntactic tree kernels. In CoNLL 2010 - Fourteenth Conference on Computational Natural Language Learning, Proceedings of the Conference (pp. 223-233)

On reverse feature engineering of syntactic tree kernels. / Pighin, Daniele; Moschitti, Alessandro.

CoNLL 2010 - Fourteenth Conference on Computational Natural Language Learning, Proceedings of the Conference. 2010. p. 223-233.

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

Pighin, D & Moschitti, A 2010, On reverse feature engineering of syntactic tree kernels. in CoNLL 2010 - Fourteenth Conference on Computational Natural Language Learning, Proceedings of the Conference. pp. 223-233, 14th Conference on Computational Natural Language Learning, CoNLL 2010, Uppsala, Sweden, 15/7/10.
Pighin D, Moschitti A. On reverse feature engineering of syntactic tree kernels. In CoNLL 2010 - Fourteenth Conference on Computational Natural Language Learning, Proceedings of the Conference. 2010. p. 223-233
Pighin, Daniele ; Moschitti, Alessandro. / On reverse feature engineering of syntactic tree kernels. CoNLL 2010 - Fourteenth Conference on Computational Natural Language Learning, Proceedings of the Conference. 2010. pp. 223-233
@inproceedings{4f2f3b818ad3402d9f8a135a30e27b3f,
title = "On reverse feature engineering of syntactic tree kernels",
abstract = "In this paper, we provide a theoretical framework for feature selection in tree kernel spaces based on gradient-vector components of kernel-based machines. We show that a huge number of features can be discarded without a significant decrease in accuracy. Our selection algorithm is as accurate as and much more efficient than those proposed in previous work. Comparative experiments on three interesting and very diverse classification tasks, i.e. Question Classification, Relation Extraction and Semantic Role Labeling, support our theoretical findings and demonstrate the algorithm performance.",
author = "Daniele Pighin and Alessandro Moschitti",
year = "2010",
month = "12",
day = "1",
language = "English",
isbn = "9781932432831",
pages = "223--233",
booktitle = "CoNLL 2010 - Fourteenth Conference on Computational Natural Language Learning, Proceedings of the Conference",

}

TY - GEN

T1 - On reverse feature engineering of syntactic tree kernels

AU - Pighin, Daniele

AU - Moschitti, Alessandro

PY - 2010/12/1

Y1 - 2010/12/1

N2 - In this paper, we provide a theoretical framework for feature selection in tree kernel spaces based on gradient-vector components of kernel-based machines. We show that a huge number of features can be discarded without a significant decrease in accuracy. Our selection algorithm is as accurate as and much more efficient than those proposed in previous work. Comparative experiments on three interesting and very diverse classification tasks, i.e. Question Classification, Relation Extraction and Semantic Role Labeling, support our theoretical findings and demonstrate the algorithm performance.

AB - In this paper, we provide a theoretical framework for feature selection in tree kernel spaces based on gradient-vector components of kernel-based machines. We show that a huge number of features can be discarded without a significant decrease in accuracy. Our selection algorithm is as accurate as and much more efficient than those proposed in previous work. Comparative experiments on three interesting and very diverse classification tasks, i.e. Question Classification, Relation Extraction and Semantic Role Labeling, support our theoretical findings and demonstrate the algorithm performance.

UR - http://www.scopus.com/inward/record.url?scp=79959237063&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79959237063&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9781932432831

SP - 223

EP - 233

BT - CoNLL 2010 - Fourteenth Conference on Computational Natural Language Learning, Proceedings of the Conference

ER -