Generating semantic snapshots of newscasts using entity expansion

TV newscasts report about the latest event-related facts occurring in the world. But relying exclusively on them is therefore insucient to fully grasp the context of the fact being reported. In this paper we propose an approach that retrieves and analyzes related documents in the Web to automatically generate semantic annotations that provide viewers and experts comprehensive information about the news. Named entities detected in the retrieved documents further disclose relevant concepts that were not explicitly mentioned in the original newscast. A ranking algorithm based on entity frequency, popularity peak analysis, and domain experts rules sorts those annotations to generate the Semantic Snapshot of the considered Newscast (NSS). We benchmark this method against a gold standard generated by domain experts and assessed via a user survey over ve BBC newscasts. Results of the experiments show the robustness of the approach holding an Average Normalized Discounted Cumulative Gain of 66.6%.