Assessing the Quality of Socially Created Metadata to Image Indexing

摘要

Deploying photograph collections in existing social taggingcommunities (e.g., Flickr) to collect large sets of sociallycreated metadata to extend existing controlled vocabulariesis a recent activity in building Knowledge OrganizationSystems. This study evaluates the reuse value of socialmetadata for image indexing by using a controlled end-userexperimental design to elicit a set of quality criteria forindex terms from users. The findings indicate that a numberof quality criteria of image indexing are different fromthose of textual document indexing, such as moods,importance, generality, level of detail, and ease of use. Thissuggests that specific criteria need to be taken into accountwhen integrating socially created metadata with imageindexing, and that the two processes can be complementary.