Real-Time Story Detection and Video Retrieval from Social Media Streams

This chapter introduces two key tools for journalists. Before being able to initiate the process of verification of an online video, they need to be able to determine the news story that is the subject of online video, and they need to be able to find candidate online videos around that story. To do this, we have assessed prior research in the area of topic detection and developed a keyword graph-based method for news story discovery out of Twitter streams. Then we have developed a technique for selection of online videos which are candidates for news stories by using the detected stories to form a query against social networks. This enables relevant information retrieval at Web scale for news story-associated videos. We present these techniques and results of their evaluations by observation of the detected stories and of the news videos which are presented for those stories, demonstrating state-of-the-art precision and recall for journalists to quickly identify videos for verification and re-use.