Presentation 2010-06-25
Refining Noisy Training Examples Based on Ensemble Learning for Intelligent Domain-Specific WEB Search
Hiroki HIRABAYASHI, Koji IWANUMA, Yoshitaka YAMAMOTO, Hidetomo NABESHIMA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) The Keyword Spices, proposed Oyama et al., is a sort of a query-expansion technology, which adds pre-computed additional words to a given query in order to perform an effective domain-specific WEB search. The Keyword Spice technology can achieve a significant performance, but needs a great deal of high-quality training data for learning a decision tree, from which adequate additional words to a query are generated. In this paper, we study an ensemble learning method, especially so-called a bagging, for decision trees used to refine noisy training data for synthesizing good keyword spice words. Throughout experimental evaluations, we show that a bagging method has a high possibility for stabilizing the effects for refining noisy data.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) WEB search / keyword spice / ensemble learning / decision tree / domain-specific search / refinement
Paper # AI2010-5
Date of Issue

Conference Information
Committee AI
Conference Date 2010/6/18(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Artificial Intelligence and Knowledge-Based Processing (AI)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Refining Noisy Training Examples Based on Ensemble Learning for Intelligent Domain-Specific WEB Search
Sub Title (in English)
Keyword(1) WEB search
Keyword(2) keyword spice
Keyword(3) ensemble learning
Keyword(4) decision tree
Keyword(5) domain-specific search
Keyword(6) refinement
1st Author's Name Hiroki HIRABAYASHI
1st Author's Affiliation Yamanashi University()
2nd Author's Name Koji IWANUMA
2nd Author's Affiliation Yamanashi University
3rd Author's Name Yoshitaka YAMAMOTO
3rd Author's Affiliation Yamanashi University
4th Author's Name Hidetomo NABESHIMA
4th Author's Affiliation Yamanashi University
Date 2010-06-25
Paper # AI2010-5
Volume (vol) vol.110
Number (no) 105
Page pp.pp.-
#Pages 6
Date of Issue