Presentation 2014-01-23
Improvement of a Distance between Concepts Using a Weighted Combination of Image Instances with Their Text Tags
Nozomi MASUHIRO, Kazuaki NAKAMURA, Noboru BABAGUCHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In recent years, methods for calculating semantic distance between various concepts from a set of tagged images have been actively studied due to their applicability to image retrieval, clustering, annotation, and so on. Most of the existing distance calculation methods can be divided into two categories: the methods based on tag co-occurrence and those based on visual features extracted from the images. The former has a drawback that two similar concepts do not necessarily co-occur as text tags even if they are concurrently presented in an image. On the other hand, the latter has a drawback that visual features do not correspond directly to semantic meaning of concepts due to the semantic gap, especially in the case with abstract concepts. To overcome these drawbacks, in this paper, we propose to combine the above two types of methods by adaptively weighting each of them according to the level of abstraction for each concept. In our experiments, our proposed method outperformed the methods using only either tag co-occurrence or visual features.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) distance between concepts / visual features / tag co-occurrence / level of abstraction / weighted sum
Paper # PRMU2013-103,MVE2013-44
Date of Issue

Conference Information
Committee MVE
Conference Date 2014/1/16(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 Media Experience and Virtual Environment (MVE)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Improvement of a Distance between Concepts Using a Weighted Combination of Image Instances with Their Text Tags
Sub Title (in English)
Keyword(1) distance between concepts
Keyword(2) visual features
Keyword(3) tag co-occurrence
Keyword(4) level of abstraction
Keyword(5) weighted sum
1st Author's Name Nozomi MASUHIRO
1st Author's Affiliation School of Engineering, Osaka University()
2nd Author's Name Kazuaki NAKAMURA
2nd Author's Affiliation Graduate School of Engineering, Osaka University
3rd Author's Name Noboru BABAGUCHI
3rd Author's Affiliation Graduate School of Engineering, Osaka University
Date 2014-01-23
Paper # PRMU2013-103,MVE2013-44
Volume (vol) vol.113
Number (no) 403
Page pp.pp.-
#Pages 6
Date of Issue