95 0

Semantic tag recommendation based on associated words exploiting the interwiki links of Wikipedia

Title
Semantic tag recommendation based on associated words exploiting the interwiki links of Wikipedia
Author
이동호
Keywords
Associated words; image annotation; image retrieval; tag recommendation; Wikipedia
Issue Date
2018-06
Publisher
SAGE PUBLICATIONS LTD
Citation
JOURNAL OF INFORMATION SCIENCE, v. 44, No. 3, Page. 298-313
Abstract
The volumes of multimedia content and users have increased on social multimedia sites due to the prevalence of smart mobile devices and digital cameras. It is common for users to take pictures and upload them to image-sharing websites using their smartphones. However, the tag characteristics deteriorate the quality of tag-based image retrieval and decrease the reliability of social multimedia sites. In this article, we propose a semantic tag recommendation technique exploiting associated words that are semantically similar or related to each other using the interwiki links of Wikipedia. First, we generate a word relationship graph after extracting meaningful words from each article in Wikipedia. The candidate words are then rearranged according to importance by applying a link-based ranking algorithm and then the top-k words are defined as the associated words for the article. When a user uploads an image, we collect visually similar images from a social image database. After propagating the proper tags from the collected images, we recommend associated words related to the candidate tags. Our experimental results show that the proposed method can improve the accuracy by up to 14% compared with other works and that exploiting associated words makes it possible to perform semantic tag recommendation.
URI
https://journals.sagepub.com/doi/abs/10.1177/0165551517693497http://repository.hanyang.ac.kr/handle/20.500.11754/81157
ISSN
0165-5515
DOI
10.1177/0165551517693497
Appears in Collections:
COLLEGE OF COMPUTING[E] > COMPUTER SCIENCE(소프트웨어학부) > Articles
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE