Efficient linearization of tree kernel functions

Daniele Pighin, Alessandro Moschitti

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

7 Citations (Scopus)

Abstract

The combination of Support Vector Machines with very high dimensional kernels, such as string or tree kernels, suffers from two major drawbacks: first, the implicit representation of feature spaces does not allow us to understand which features actually triggered the generalization; second, the resulting computational burden may in some cases render unfeasible to use large data sets for training. We propose an approach based on feature space reverse engineering to tackle both problems. Our experiments with Tree Kernels on a Semantic Role Labeling data set show that the proposed approach can drastically reduce the computational footprint while yielding almost unaffected accuracy.

Original languageEnglish
Title of host publicationCoNLL 2009 - Proceedings of the Thirteenth Conference on Computational Natural Language Learning
Pages30-38
Number of pages9
Publication statusPublished - 1 Dec 2009
Externally publishedYes
Event13th Conference on Computational Natural Language Learning, CoNLL 2009 - Boulder, CO, United States
Duration: 4 Jun 20095 Jun 2009

Other

Other13th Conference on Computational Natural Language Learning, CoNLL 2009
CountryUnited States
CityBoulder, CO
Period4/6/095/6/09

Fingerprint

Aerospace engineering
Reverse engineering
Trees (mathematics)
Linearization
Labeling
Support vector machines
Semantics
Experiments
semantics
engineering
experiment

ASJC Scopus subject areas

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

Cite this

Pighin, D., & Moschitti, A. (2009). Efficient linearization of tree kernel functions. In CoNLL 2009 - Proceedings of the Thirteenth Conference on Computational Natural Language Learning (pp. 30-38)

Efficient linearization of tree kernel functions. / Pighin, Daniele; Moschitti, Alessandro.

CoNLL 2009 - Proceedings of the Thirteenth Conference on Computational Natural Language Learning. 2009. p. 30-38.

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

Pighin, D & Moschitti, A 2009, Efficient linearization of tree kernel functions. in CoNLL 2009 - Proceedings of the Thirteenth Conference on Computational Natural Language Learning. pp. 30-38, 13th Conference on Computational Natural Language Learning, CoNLL 2009, Boulder, CO, United States, 4/6/09.
Pighin D, Moschitti A. Efficient linearization of tree kernel functions. In CoNLL 2009 - Proceedings of the Thirteenth Conference on Computational Natural Language Learning. 2009. p. 30-38
Pighin, Daniele ; Moschitti, Alessandro. / Efficient linearization of tree kernel functions. CoNLL 2009 - Proceedings of the Thirteenth Conference on Computational Natural Language Learning. 2009. pp. 30-38
@inproceedings{48924eeee53640099fb0a4fe5520ccfb,
title = "Efficient linearization of tree kernel functions",
abstract = "The combination of Support Vector Machines with very high dimensional kernels, such as string or tree kernels, suffers from two major drawbacks: first, the implicit representation of feature spaces does not allow us to understand which features actually triggered the generalization; second, the resulting computational burden may in some cases render unfeasible to use large data sets for training. We propose an approach based on feature space reverse engineering to tackle both problems. Our experiments with Tree Kernels on a Semantic Role Labeling data set show that the proposed approach can drastically reduce the computational footprint while yielding almost unaffected accuracy.",
author = "Daniele Pighin and Alessandro Moschitti",
year = "2009",
month = "12",
day = "1",
language = "English",
isbn = "1932432299",
pages = "30--38",
booktitle = "CoNLL 2009 - Proceedings of the Thirteenth Conference on Computational Natural Language Learning",

}

TY - GEN

T1 - Efficient linearization of tree kernel functions

AU - Pighin, Daniele

AU - Moschitti, Alessandro

PY - 2009/12/1

Y1 - 2009/12/1

N2 - The combination of Support Vector Machines with very high dimensional kernels, such as string or tree kernels, suffers from two major drawbacks: first, the implicit representation of feature spaces does not allow us to understand which features actually triggered the generalization; second, the resulting computational burden may in some cases render unfeasible to use large data sets for training. We propose an approach based on feature space reverse engineering to tackle both problems. Our experiments with Tree Kernels on a Semantic Role Labeling data set show that the proposed approach can drastically reduce the computational footprint while yielding almost unaffected accuracy.

AB - The combination of Support Vector Machines with very high dimensional kernels, such as string or tree kernels, suffers from two major drawbacks: first, the implicit representation of feature spaces does not allow us to understand which features actually triggered the generalization; second, the resulting computational burden may in some cases render unfeasible to use large data sets for training. We propose an approach based on feature space reverse engineering to tackle both problems. Our experiments with Tree Kernels on a Semantic Role Labeling data set show that the proposed approach can drastically reduce the computational footprint while yielding almost unaffected accuracy.

UR - http://www.scopus.com/inward/record.url?scp=77958030821&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77958030821&partnerID=8YFLogxK

M3 - Conference contribution

SN - 1932432299

SN - 9781932432299

SP - 30

EP - 38

BT - CoNLL 2009 - Proceedings of the Thirteenth Conference on Computational Natural Language Learning

ER -