Semantic Linking between Video Ads and Web Services with Progressive Search

With the proliferation of online media services, ad video has become an important way to promote various products, services and ideas. Research efforts have been devoted to the contextual advertising whereas comprehensive recommendation of video ads is less exploited. In this paper, we propose to establish a semantic linking between video ads and relevant product/service online in a cross-media manner. First, we extract a representative key frame from the ad video and then conduct a three-step progressive search (i. e., visual search, tag aggregation and textual re-search) to link video ads with relevant Web service. We search visually similar product images, rank the context textual information by tags aggregation, and refine the results by textual re-search. Finally, we collect relevant products for user recommendation. Experiments on some popular E-commerce websites like eBay and Amazon have demonstrated the attractiveness of the semantic linking.