Semantic tree kernels to classify predicate argument structures

Alessandro Moschitti, Bonaventura Coppola, Daniele Pighin, Roberto Basili

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

Recent work on Semantic Role Labeling (SRL) has shown that syntactic information is critical to detect and extract predicate argument structures. As syntax is expressed by means of structured data, i.e. parse trees, its encoding in learning algorithms is rather complex. In this paper, we apply tree kernels to encode the whole predicate argument structure in Support Vector Machines (SVMs). We extract from the sentence syntactic parse the subtrees that span potential argument structures of the target predicate and classify them in incorrect or correct structures by means of tree kernel based SVMs. Experiments on the PropBank collection show that the classification accuracy of correct/incorrect structures is remarkably high and helps to improve the accuracy of the SRL task. This is a piece of evidence that tree kernels provide a powerful mechanism to learn the complex relation between syntax and semantics.

Original languageEnglish
Title of host publicationFrontiers in Artificial Intelligence and Applications
Pages568-572
Number of pages5
Volume141
Publication statusPublished - 1 Dec 2006
Externally publishedYes

Publication series

NameFrontiers in Artificial Intelligence and Applications
Volume141
ISSN (Print)09226389

Fingerprint

Semantics
Syntactics
Labeling
Support vector machines
Learning algorithms
Experiments

ASJC Scopus subject areas

  • Artificial Intelligence

Cite this

Moschitti, A., Coppola, B., Pighin, D., & Basili, R. (2006). Semantic tree kernels to classify predicate argument structures. In Frontiers in Artificial Intelligence and Applications (Vol. 141, pp. 568-572). (Frontiers in Artificial Intelligence and Applications; Vol. 141).

Semantic tree kernels to classify predicate argument structures. / Moschitti, Alessandro; Coppola, Bonaventura; Pighin, Daniele; Basili, Roberto.

Frontiers in Artificial Intelligence and Applications. Vol. 141 2006. p. 568-572 (Frontiers in Artificial Intelligence and Applications; Vol. 141).

Research output: Chapter in Book/Report/Conference proceedingChapter

Moschitti, A, Coppola, B, Pighin, D & Basili, R 2006, Semantic tree kernels to classify predicate argument structures. in Frontiers in Artificial Intelligence and Applications. vol. 141, Frontiers in Artificial Intelligence and Applications, vol. 141, pp. 568-572.
Moschitti A, Coppola B, Pighin D, Basili R. Semantic tree kernels to classify predicate argument structures. In Frontiers in Artificial Intelligence and Applications. Vol. 141. 2006. p. 568-572. (Frontiers in Artificial Intelligence and Applications).
Moschitti, Alessandro ; Coppola, Bonaventura ; Pighin, Daniele ; Basili, Roberto. / Semantic tree kernels to classify predicate argument structures. Frontiers in Artificial Intelligence and Applications. Vol. 141 2006. pp. 568-572 (Frontiers in Artificial Intelligence and Applications).
@inbook{9b6be35d5fcb43c1893a46b741143ee2,
title = "Semantic tree kernels to classify predicate argument structures",
abstract = "Recent work on Semantic Role Labeling (SRL) has shown that syntactic information is critical to detect and extract predicate argument structures. As syntax is expressed by means of structured data, i.e. parse trees, its encoding in learning algorithms is rather complex. In this paper, we apply tree kernels to encode the whole predicate argument structure in Support Vector Machines (SVMs). We extract from the sentence syntactic parse the subtrees that span potential argument structures of the target predicate and classify them in incorrect or correct structures by means of tree kernel based SVMs. Experiments on the PropBank collection show that the classification accuracy of correct/incorrect structures is remarkably high and helps to improve the accuracy of the SRL task. This is a piece of evidence that tree kernels provide a powerful mechanism to learn the complex relation between syntax and semantics.",
author = "Alessandro Moschitti and Bonaventura Coppola and Daniele Pighin and Roberto Basili",
year = "2006",
month = "12",
day = "1",
language = "English",
isbn = "9781586036423",
volume = "141",
series = "Frontiers in Artificial Intelligence and Applications",
pages = "568--572",
booktitle = "Frontiers in Artificial Intelligence and Applications",

}

TY - CHAP

T1 - Semantic tree kernels to classify predicate argument structures

AU - Moschitti, Alessandro

AU - Coppola, Bonaventura

AU - Pighin, Daniele

AU - Basili, Roberto

PY - 2006/12/1

Y1 - 2006/12/1

N2 - Recent work on Semantic Role Labeling (SRL) has shown that syntactic information is critical to detect and extract predicate argument structures. As syntax is expressed by means of structured data, i.e. parse trees, its encoding in learning algorithms is rather complex. In this paper, we apply tree kernels to encode the whole predicate argument structure in Support Vector Machines (SVMs). We extract from the sentence syntactic parse the subtrees that span potential argument structures of the target predicate and classify them in incorrect or correct structures by means of tree kernel based SVMs. Experiments on the PropBank collection show that the classification accuracy of correct/incorrect structures is remarkably high and helps to improve the accuracy of the SRL task. This is a piece of evidence that tree kernels provide a powerful mechanism to learn the complex relation between syntax and semantics.

AB - Recent work on Semantic Role Labeling (SRL) has shown that syntactic information is critical to detect and extract predicate argument structures. As syntax is expressed by means of structured data, i.e. parse trees, its encoding in learning algorithms is rather complex. In this paper, we apply tree kernels to encode the whole predicate argument structure in Support Vector Machines (SVMs). We extract from the sentence syntactic parse the subtrees that span potential argument structures of the target predicate and classify them in incorrect or correct structures by means of tree kernel based SVMs. Experiments on the PropBank collection show that the classification accuracy of correct/incorrect structures is remarkably high and helps to improve the accuracy of the SRL task. This is a piece of evidence that tree kernels provide a powerful mechanism to learn the complex relation between syntax and semantics.

UR - http://www.scopus.com/inward/record.url?scp=84885993013&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84885993013&partnerID=8YFLogxK

M3 - Chapter

SN - 9781586036423

VL - 141

T3 - Frontiers in Artificial Intelligence and Applications

SP - 568

EP - 572

BT - Frontiers in Artificial Intelligence and Applications

ER -