Presentation 2002/9/12
Learning Ordering Function from Ranked Examples
Hideto KAZAWA, Tsutomu HIRAO, Eisaku MAEDA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, we study the learning problem of ordering function, where training examples are ranked according to unknown ordering and the learner is to estimate ordering function which maps each sample to a real number which is consistent with the unkown ordering. Recently, a classifier estimation method was applied to this problem and obtained good results with real-world data. Although the method has several advantages over other existing methods in terms of computation time and design simplicity, it lacks theoretical justification. In this paper, we theoretically argue why classifiers can order examples. Additionally, we propose "Ranking SVM" to overcome some shortcomings of classifier-based methods. Ranking SVM and existing learning methods are experimentally compared using artificial data and real "sentense selection" data.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) ordered data / support vector machines / parallel boundaries / parameter variation regularization / sentence selection
Paper # WIT2002-15
Date of Issue

Conference Information
Committee WIT
Conference Date 2002/9/12(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 Well-being Information Technology(WIT)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Learning Ordering Function from Ranked Examples
Sub Title (in English)
Keyword(1) ordered data
Keyword(2) support vector machines
Keyword(3) parallel boundaries
Keyword(4) parameter variation regularization
Keyword(5) sentence selection
1st Author's Name Hideto KAZAWA
1st Author's Affiliation NTT Communication Science Laboratories()
2nd Author's Name Tsutomu HIRAO
2nd Author's Affiliation NTT Communication Science Laboratories
3rd Author's Name Eisaku MAEDA
3rd Author's Affiliation NTT Communication Science Laboratories
Date 2002/9/12
Paper # WIT2002-15
Volume (vol) vol.102
Number (no) 319
Page pp.pp.-
#Pages 6
Date of Issue