Generating Pictorial Storylines Via Minimum-Weight Connected Dominating Set Approximation in Multi-View Graphs

This paper introduces a novel framework for generating pictorial storylines given topics from text and image data on the Internet. Unlike traditional summarization timeline generation systems, proposed combines analysis delivers storyline containing textual, pictorial, structural information to provide sketch of topic evolution. A key idea in is use an approximate solution dominating set problem. Given collection topic-related objects consisting images their descriptions, weighted multi-view graph first constructed capture contextual temporal relationships among these objects. Then are selected by solving minimum-weighted connected problem defined this graph. Comprehensive experiments real-world sets demonstrate effectiveness framework.