A systematic comparison of feature-rich probabilistic classifiers for NER tasks

Benjamin Rosenfeld*, Moshe Fresko, Ronen Feldman

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

2 Scopus citations

Abstract

In the CoNLL 2003 NER shared task, more than two thirds of the submitted systems used the feature-rich representation of the task. Most of them used maximum entropy to combine the features together. Others used linear classifiers, such as SVM and RRM. Among all systems presented there, one of the MEMM-based classifiers took the second place, losing only to a committee of four different classifiers, one of which was ME-based and another RRM-based. The lone RRM was fourth, and CRF came in the middle of the pack. In this paper we shall demonstrate, by running the three algorithms upon the same tasks under exactly the same conditions that this ranking is due to feature selection and other causes and not due to the inherent qualities of the algorithms, which should be ranked otherwise.

Original languageAmerican English
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages217-227
Number of pages11
DOIs
StatePublished - 2005
Externally publishedYes
Event9th European Conference on Principles and Practice of Knowledge Discovery in Databases, PKDD 2005 - Porto, Portugal
Duration: 3 Oct 20057 Oct 2005

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume3721 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference9th European Conference on Principles and Practice of Knowledge Discovery in Databases, PKDD 2005
Country/TerritoryPortugal
CityPorto
Period3/10/057/10/05

Fingerprint

Dive into the research topics of 'A systematic comparison of feature-rich probabilistic classifiers for NER tasks'. Together they form a unique fingerprint.

Cite this