Filtering-ranking perceptron learning for partial parsing

Xavier Carreras, Lluis Marques, Jorge Castro

Research output: Contribution to journalArticle

17 Citations (Scopus)

Abstract

This work introduces a general phrase recognition system based on perceptrons, and a global online learning algorithm to train them together. The method applies to complex domains in which some structure has to be recognized. This global problem is broken down into two layers of local subproblems: a filtering layer, which reduces the search space by identifying plausible phrase candidates; and a ranking layer, which builds the optimal phrase structure by discriminating among competing phrases. A recognition-based feedback rule is presented which reflects to each local function its committed errors from a global point of view, and allows to train them together online as perceptrons. As a result, the learned functions automatically behave as filters and rankers, rather than binary classifiers, which we argue to be better for this type of problems. Extensive experimentation on partial parsing tasks gives state-of-the-art results and evinces the advantages of the global training method over optimizing each function locally and independently.

Original languageEnglish
Pages (from-to)41-71
Number of pages31
JournalMachine Learning
Volume60
Issue number1-3
DOIs
Publication statusPublished - 1 Sep 2005
Externally publishedYes

Fingerprint

Neural networks
Learning algorithms
Classifiers
Feedback

Keywords

  • Natural language processing
  • Online learning
  • Partial parsing
  • Perceptron
  • Sequential data
  • Structure recognition

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Artificial Intelligence

Cite this

Filtering-ranking perceptron learning for partial parsing. / Carreras, Xavier; Marques, Lluis; Castro, Jorge.

In: Machine Learning, Vol. 60, No. 1-3, 01.09.2005, p. 41-71.

Research output: Contribution to journalArticle

Carreras, Xavier ; Marques, Lluis ; Castro, Jorge. / Filtering-ranking perceptron learning for partial parsing. In: Machine Learning. 2005 ; Vol. 60, No. 1-3. pp. 41-71.
@article{54b62f915d2445af9c5d3fa8b5262559,
title = "Filtering-ranking perceptron learning for partial parsing",
abstract = "This work introduces a general phrase recognition system based on perceptrons, and a global online learning algorithm to train them together. The method applies to complex domains in which some structure has to be recognized. This global problem is broken down into two layers of local subproblems: a filtering layer, which reduces the search space by identifying plausible phrase candidates; and a ranking layer, which builds the optimal phrase structure by discriminating among competing phrases. A recognition-based feedback rule is presented which reflects to each local function its committed errors from a global point of view, and allows to train them together online as perceptrons. As a result, the learned functions automatically behave as filters and rankers, rather than binary classifiers, which we argue to be better for this type of problems. Extensive experimentation on partial parsing tasks gives state-of-the-art results and evinces the advantages of the global training method over optimizing each function locally and independently.",
keywords = "Natural language processing, Online learning, Partial parsing, Perceptron, Sequential data, Structure recognition",
author = "Xavier Carreras and Lluis Marques and Jorge Castro",
year = "2005",
month = "9",
day = "1",
doi = "10.1007/s10994-005-0917-x",
language = "English",
volume = "60",
pages = "41--71",
journal = "Machine Learning",
issn = "0885-6125",
publisher = "Springer Netherlands",
number = "1-3",

}

TY - JOUR

T1 - Filtering-ranking perceptron learning for partial parsing

AU - Carreras, Xavier

AU - Marques, Lluis

AU - Castro, Jorge

PY - 2005/9/1

Y1 - 2005/9/1

N2 - This work introduces a general phrase recognition system based on perceptrons, and a global online learning algorithm to train them together. The method applies to complex domains in which some structure has to be recognized. This global problem is broken down into two layers of local subproblems: a filtering layer, which reduces the search space by identifying plausible phrase candidates; and a ranking layer, which builds the optimal phrase structure by discriminating among competing phrases. A recognition-based feedback rule is presented which reflects to each local function its committed errors from a global point of view, and allows to train them together online as perceptrons. As a result, the learned functions automatically behave as filters and rankers, rather than binary classifiers, which we argue to be better for this type of problems. Extensive experimentation on partial parsing tasks gives state-of-the-art results and evinces the advantages of the global training method over optimizing each function locally and independently.

AB - This work introduces a general phrase recognition system based on perceptrons, and a global online learning algorithm to train them together. The method applies to complex domains in which some structure has to be recognized. This global problem is broken down into two layers of local subproblems: a filtering layer, which reduces the search space by identifying plausible phrase candidates; and a ranking layer, which builds the optimal phrase structure by discriminating among competing phrases. A recognition-based feedback rule is presented which reflects to each local function its committed errors from a global point of view, and allows to train them together online as perceptrons. As a result, the learned functions automatically behave as filters and rankers, rather than binary classifiers, which we argue to be better for this type of problems. Extensive experimentation on partial parsing tasks gives state-of-the-art results and evinces the advantages of the global training method over optimizing each function locally and independently.

KW - Natural language processing

KW - Online learning

KW - Partial parsing

KW - Perceptron

KW - Sequential data

KW - Structure recognition

UR - http://www.scopus.com/inward/record.url?scp=24044479403&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=24044479403&partnerID=8YFLogxK

U2 - 10.1007/s10994-005-0917-x

DO - 10.1007/s10994-005-0917-x

M3 - Article

VL - 60

SP - 41

EP - 71

JO - Machine Learning

JF - Machine Learning

SN - 0885-6125

IS - 1-3

ER -