Presented at: Incentives For The Semantic Web (INSEMTIVE2008)
by Yukino Baba, Fuyuki Ishikawa, Shinichi Honiden
Folksonomy is a method of classifying content, and it iswidely used in some web services. It allows users to choose tags (keywords or terms assigned to specific content) freely and to search content by referring to the tags. Compared to existing classification methods, folksonomy reflects the users’ intention more directly because of its unlimited vocabulary and multiple tags for one content item. Moreover, it has a useful characteristic where tags represent the description of the content. Although tags are intended to be a rich semantic description of web content, machines cannot understand what the tags mean because they are just keywords. We describe a method to extract the concept related to the tag in a machine-understandable way by focusing on the features of content annotated with each tag. In particular, we target the problem of extracting the temporal and spatial concepts of the tags on Flickr, a popular photo sharing service by looking at the date and location distributions of photos for each tag. We evaluated the concept extracting method on a snapshot of actual Flickr data and show that it can identify a tags’ concept in a manner similar to the way a person can.
Resource URI on the dog food server: http://data.semanticweb.org/workshop/insemtive/2008/paper/main/6
Explore this resource elsewhere: