Training algorithms for linear text classifiers

David D. Lewis; Robert E. Schapire; James P. Callan; Ron Papka

doi:10.1145/243199.243277

Training algorithms for linear text classifiers

David D. Lewis, Robert E. Schapire, James P. Callan, Ron Papka

Research output: Contribution to journal › Conference article › peer-review

370 Scopus citations

Abstract

Systems for text retrieval, routing, categorization and other IR tasks rely heavily on linear classifiers. We propose that two machine learning algorithms, the Widrow-Hoff and EG algorithms, be used in training linear text classifiers. In contrast to most IR methods, theoretical analysis provides performance guarantees and guidance on parameter settings for these algorithms. Experimental data is presented showing Widrow-Hoff and EG to be more effective than the widely used Rocchio algorithm on several categorization and routing tasks.

Original language	English (US)
Pages (from-to)	298-306
Number of pages	9
Journal	SIGIR Forum (ACM Special Interest Group on Information Retrieval)
DOIs	https://doi.org/10.1145/243199.243277
State	Published - 1996
Externally published	Yes
Event	Proceedings of the 1996 19th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 96 - Zurich, Switz Duration: Aug 18 1996 → Aug 22 1996

All Science Journal Classification (ASJC) codes

Management Information Systems
Hardware and Architecture

Access to Document

10.1145/243199.243277

Cite this

@article{67c2d00fe64d4448a56d3dd0ebc6d44d,

title = "Training algorithms for linear text classifiers",

abstract = "Systems for text retrieval, routing, categorization and other IR tasks rely heavily on linear classifiers. We propose that two machine learning algorithms, the Widrow-Hoff and EG algorithms, be used in training linear text classifiers. In contrast to most IR methods, theoretical analysis provides performance guarantees and guidance on parameter settings for these algorithms. Experimental data is presented showing Widrow-Hoff and EG to be more effective than the widely used Rocchio algorithm on several categorization and routing tasks.",

author = "Lewis, {David D.} and Schapire, {Robert E.} and Callan, {James P.} and Ron Papka",

year = "1996",

doi = "10.1145/243199.243277",

language = "English (US)",

pages = "298--306",

journal = "SIGIR Forum (ACM Special Interest Group on Information Retrieval)",

issn = "0163-5840",

publisher = "Association for Computing Machinery (ACM)",

note = "Proceedings of the 1996 19th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 96 ; Conference date: 18-08-1996 Through 22-08-1996",

}

TY - JOUR

T1 - Training algorithms for linear text classifiers

AU - Lewis, David D.

AU - Schapire, Robert E.

AU - Callan, James P.

AU - Papka, Ron

PY - 1996

Y1 - 1996

N2 - Systems for text retrieval, routing, categorization and other IR tasks rely heavily on linear classifiers. We propose that two machine learning algorithms, the Widrow-Hoff and EG algorithms, be used in training linear text classifiers. In contrast to most IR methods, theoretical analysis provides performance guarantees and guidance on parameter settings for these algorithms. Experimental data is presented showing Widrow-Hoff and EG to be more effective than the widely used Rocchio algorithm on several categorization and routing tasks.

AB - Systems for text retrieval, routing, categorization and other IR tasks rely heavily on linear classifiers. We propose that two machine learning algorithms, the Widrow-Hoff and EG algorithms, be used in training linear text classifiers. In contrast to most IR methods, theoretical analysis provides performance guarantees and guidance on parameter settings for these algorithms. Experimental data is presented showing Widrow-Hoff and EG to be more effective than the widely used Rocchio algorithm on several categorization and routing tasks.

UR - http://www.scopus.com/inward/record.url?scp=0030405371&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0030405371&partnerID=8YFLogxK

U2 - 10.1145/243199.243277

DO - 10.1145/243199.243277

M3 - Conference article

AN - SCOPUS:0030405371

SN - 0163-5840

SP - 298

EP - 306

JO - SIGIR Forum (ACM Special Interest Group on Information Retrieval)

JF - SIGIR Forum (ACM Special Interest Group on Information Retrieval)

T2 - Proceedings of the 1996 19th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 96

Y2 - 18 August 1996 through 22 August 1996

ER -

Training algorithms for linear text classifiers

Abstract

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this