모바일 환경에서 연관단어를 활용한 이미지태그 추천 기법

An Image Tag Recommendation Technique Exploiting Associated Words in Mobile Environment

초록

The amount of multimedia content has rapidly increased in social multimedia sites due to the prevalence of digital camera and smartphone. Recently, it is common that users take pictures, upload and share them to the social image site using their own smartphone. In order to understand images better and improve the performance of image retrieval, tags are mainly used in social multimedia sites. However, the use of unaccurate and untrustable tags and the limitations of mobile devices lead to the performance degradation of tag-based image retrievals. In order to handle these problems, various tag recommendation techniques have been proposed but these efforts have not been satisfied yet. In this paper, we propose an image tag recommendation technique exploiting Wikipedia-based associated words so as to recommend accurate and trustable image tag for an input image. The proposed method constructs associated words by exploiting a link-based ranking algorithm after generating a word relationship graph with meaningful words in Wikipedia. Then, the proposed method selects the candidate tags for an input image and recommends image tags by searching the associated words for the candidate tags from the associated words database. The experiment results show that the proposed method can improve the accuracy by up to 10% compared to other works.

키워드

태그 추천; 태그 기반 이미지 검색; 연관 단어; 위키피디아; Tag recommendation; Tag-based image retrieval; Associated words; Wikipedia
제목
모바일 환경에서 연관단어를 활용한 이미지태그 추천 기법
제목 (타언어)
An Image Tag Recommendation Technique Exploiting Associated Words in Mobile Environment
저자
홍현기; 김건우; 이동호
발행일
2016-04
저널명
데이타베이스연구
권
32
호
1
페이지
63 ~ 77