A retrieval model for cross-media objects based on semantic consistency

The Internet has become the main information resource, but the retrieval capacity of traditional search engines is poorly text-based or single-modal. In this paper, based on the category structure and conjunction reference analysis of the cross-media objects and their primary features, we intent to solve the problem of the heterogeneous and incompatibility among various categories of media features, in order to accomplish the task of cross-media similarity measurement, which is fundamental for both theoretical and practical methods of cross-media information retrieval.