Cross-Media Annotation Based on Semantic Network

Abstract

The traditional information annotation technology is usually based on text description methods. With the ever increasing multimedia resources and the rapid progress of cross-media technology, the annotation between different modalities of media information is becoming possible. This paper presents a study of a cross-media annotation technology. In this technology, some multimedia examples expressed by structured information segments are submitted to a cross-media meta-search engine, to which the results in return, as the examples of other multimedia, are submitted again. The correlation between those examples and results is calculated per-time so that a cross-media semantic network is constructed by the iterative process. Thus the content of a Web page could be explained by different audio-visual perceptive cross-media information using the semantic network. This technology is proved to be feasible and effectual in the archetype system CMA (Cross Media Annotation).