Two Step POS Selection for SVM Based Text Categorization

Takeshi MASUYAMA; Hiroshi NAKAGAWA

Two Step POS Selection for SVM Based Text Categorization

Takeshi MASUYAMA, Hiroshi NAKAGAWA

Full Text Views

0

Share
Cite this

Summary :

Although many researchers have verified the superiority of Support Vector Machine (SVM) on text categorization tasks, some recent papers have reported much lower performance of SVM based text categorization methods when focusing on all types of parts of speech (POS) as input words and treating large numbers of training documents. This was caused by the overfitting problem that SVM sometimes selected unsuitable support vectors for each category in the training set. To avoid the overfitting problem, we propose a two step text categorization method with a variable cascaded feature selection (VCFS) using SVM. VCFS method selects a pair of the best number of words and the best POS combination for each category at each step of the cascade. We made use of the difference of words with the highest mutual information for each category on each POS combination. Through the experiments, we confirmed the validation of VCFS method compared with other SVM based text categorization methods, since our results showed that the macro-averaged F₁ measure (64.8%) of VCFS method was significantly better than any reported F₁ measures, though the micro-averaged F₁ measure (85.4%) of VCFS method was similar to them.

Publication: IEICE TRANSACTIONS on Information Vol.E87-D No.2 pp.373-379

Publication Date: 2004/02/01

Publicized

Online ISSN

DOI

Type of Manuscript: Special Section PAPER (Special Section on Information Processing Technology for Web Utilization)

Category

Cite this

Copy

Takeshi MASUYAMA, Hiroshi NAKAGAWA, "Two Step POS Selection for SVM Based Text Categorization" in IEICE TRANSACTIONS on Information, vol. E87-D, no. 2, pp. 373-379, February 2004, doi: .
Abstract: Although many researchers have verified the superiority of Support Vector Machine (SVM) on text categorization tasks, some recent papers have reported much lower performance of SVM based text categorization methods when focusing on all types of parts of speech (POS) as input words and treating large numbers of training documents. This was caused by the overfitting problem that SVM sometimes selected unsuitable support vectors for each category in the training set. To avoid the overfitting problem, we propose a two step text categorization method with a variable cascaded feature selection (VCFS) using SVM. VCFS method selects a pair of the best number of words and the best POS combination for each category at each step of the cascade. We made use of the difference of words with the highest mutual information for each category on each POS combination. Through the experiments, we confirmed the validation of VCFS method compared with other SVM based text categorization methods, since our results showed that the macro-averaged F₁ measure (64.8%) of VCFS method was significantly better than any reported F₁ measures, though the micro-averaged F₁ measure (85.4%) of VCFS method was similar to them.
URL: https://globals.ieice.org/en_transactions/information/10.1587/e87-d_2_373/_p

Copy

@ARTICLE{e87-d_2_373,
author={Takeshi MASUYAMA, Hiroshi NAKAGAWA, },
journal={IEICE TRANSACTIONS on Information},
title={Two Step POS Selection for SVM Based Text Categorization},
year={2004},
volume={E87-D},
number={2},
pages={373-379},
abstract={Although many researchers have verified the superiority of Support Vector Machine (SVM) on text categorization tasks, some recent papers have reported much lower performance of SVM based text categorization methods when focusing on all types of parts of speech (POS) as input words and treating large numbers of training documents. This was caused by the overfitting problem that SVM sometimes selected unsuitable support vectors for each category in the training set. To avoid the overfitting problem, we propose a two step text categorization method with a variable cascaded feature selection (VCFS) using SVM. VCFS method selects a pair of the best number of words and the best POS combination for each category at each step of the cascade. We made use of the difference of words with the highest mutual information for each category on each POS combination. Through the experiments, we confirmed the validation of VCFS method compared with other SVM based text categorization methods, since our results showed that the macro-averaged F₁ measure (64.8%) of VCFS method was significantly better than any reported F₁ measures, though the micro-averaged F₁ measure (85.4%) of VCFS method was similar to them.},
keywords={},
doi={},
ISSN={},
month={February},}

Copy

TY - JOUR
TI - Two Step POS Selection for SVM Based Text Categorization
T2 - IEICE TRANSACTIONS on Information
SP - 373
EP - 379
AU - Takeshi MASUYAMA
AU - Hiroshi NAKAGAWA
PY - 2004
DO -
JO - IEICE TRANSACTIONS on Information
SN -
VL - E87-D
IS - 2
JA - IEICE TRANSACTIONS on Information
Y1 - February 2004
AB - Although many researchers have verified the superiority of Support Vector Machine (SVM) on text categorization tasks, some recent papers have reported much lower performance of SVM based text categorization methods when focusing on all types of parts of speech (POS) as input words and treating large numbers of training documents. This was caused by the overfitting problem that SVM sometimes selected unsuitable support vectors for each category in the training set. To avoid the overfitting problem, we propose a two step text categorization method with a variable cascaded feature selection (VCFS) using SVM. VCFS method selects a pair of the best number of words and the best POS combination for each category at each step of the cascade. We made use of the difference of words with the highest mutual information for each category on each POS combination. Through the experiments, we confirmed the validation of VCFS method compared with other SVM based text categorization methods, since our results showed that the macro-averaged F₁ measure (64.8%) of VCFS method was significantly better than any reported F₁ measures, though the micro-averaged F₁ measure (85.4%) of VCFS method was similar to them.
ER -