Recomputation of class relevance scores for improving text classification

Sang Bum Kim, Hae-Chang Rim

Research output: Contribution to journalArticle

Abstract

In the text classification task, bag-of-word representation causes a critical problem when the prediction powers for a few words are estimated terribly inaccurately because of the lack of the training documents. In this paper, we propose recomputation of class relenvace scores based on the similarities among the classes for improving text classification. Through the experiments using two different baseline classifiers and two different test data, we prove that our proposed method consistently outperforms the traditional text classification strategy.

Original languageEnglish
Pages (from-to)580-583
Number of pages4
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2945
Publication statusPublished - 2004 Dec 1

Fingerprint

Text Classification
Baseline
Classifiers
Classifier
Prediction
Experiment
Class
Relevance
Experiments

ASJC Scopus subject areas

  • Computer Science(all)
  • Biochemistry, Genetics and Molecular Biology(all)
  • Theoretical Computer Science

Cite this

@article{aa7ae00cd043456d86ca1e14b626e1c8,
title = "Recomputation of class relevance scores for improving text classification",
abstract = "In the text classification task, bag-of-word representation causes a critical problem when the prediction powers for a few words are estimated terribly inaccurately because of the lack of the training documents. In this paper, we propose recomputation of class relenvace scores based on the similarities among the classes for improving text classification. Through the experiments using two different baseline classifiers and two different test data, we prove that our proposed method consistently outperforms the traditional text classification strategy.",
author = "Kim, {Sang Bum} and Hae-Chang Rim",
year = "2004",
month = "12",
day = "1",
language = "English",
volume = "2945",
pages = "580--583",
journal = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
issn = "0302-9743",
publisher = "Springer Verlag",

}

TY - JOUR

T1 - Recomputation of class relevance scores for improving text classification

AU - Kim, Sang Bum

AU - Rim, Hae-Chang

PY - 2004/12/1

Y1 - 2004/12/1

N2 - In the text classification task, bag-of-word representation causes a critical problem when the prediction powers for a few words are estimated terribly inaccurately because of the lack of the training documents. In this paper, we propose recomputation of class relenvace scores based on the similarities among the classes for improving text classification. Through the experiments using two different baseline classifiers and two different test data, we prove that our proposed method consistently outperforms the traditional text classification strategy.

AB - In the text classification task, bag-of-word representation causes a critical problem when the prediction powers for a few words are estimated terribly inaccurately because of the lack of the training documents. In this paper, we propose recomputation of class relenvace scores based on the similarities among the classes for improving text classification. Through the experiments using two different baseline classifiers and two different test data, we prove that our proposed method consistently outperforms the traditional text classification strategy.

UR - http://www.scopus.com/inward/record.url?scp=35048853276&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=35048853276&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:35048853276

VL - 2945

SP - 580

EP - 583

JO - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

JF - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SN - 0302-9743

ER -