A semantic model for flash retrieval using co-occurrence analysis

Dawei Ding, Qing Li, Bo Feng, Liu Wenyin

Abstract: Flash is experiencing a breathtaking growth and has become one of the prevailing media formats on the Web. Our goal is to exploit the enormous Flash resources by developing a model of content-based Flash retrieval. Towards this end, we introduce a novel approach for discovering semantic relationships among the co-occurrence patterns of elements in Flash movies. The proposed approach includes a three-layered structure to index the Flash movie, a query expansion procedure to improve the recall performance, and a relevance ranking procedure utilizing link analysis to improve the precision performance of Flash retrieval. Experiments show the potential of leveraging co-occurrence analysis of elements in the context of scenes for improving the performance of Flash retrieval.