Skip to Main Content
Recently, video sharing sites that allow visitors to attach comments to video have attracted a lot of attention. Video also provides a means of communication on social network sites. This kind of communication is expected to become even more popular in the future. One problem is that discussion jumps from one topic to another as users add their viewpoints. Also, conventional video sharing sites show, unsorted, all of a particular user's attached comments. It is often difficult in such systems to identify and retrieve comments relevant to one's own interest. We have developed a method for determining comment relationship types on the basis of temporal duration and the pointing region. Our method can also be used for extracting comments related to other comments made by the user. In this paper we describe how our method classifies comment relationships and discuss our prototype system which uses the method.