Text mining in multimedia

Abstract

A large amount of multimedia data (e.g., image and video) is now available on the Web. A multimedia entity does not appear in isolation, but is accompanied by various forms of metadata, such as surrounding text, user tags, ratings, and comments etc. Mining these textual metadata has been found to be effective in facilitating multimedia information processing and management. A wealth of research efforts has been dedicated to text mining in multimedia. This chapter provides a comprehensive survey of recent research efforts. Specifically, the survey focuses on four aspects: (a) surrounding text mining; (b) tag mining; (c) joint text and visual content mining; and (d) cross text and visual content mining. Furthermore, open research issues are identified based on the current research efforts.