Using Uncertain Graphs to Automatically Generate Event Flows from News Stories

Capturing the branching flow of events described in text aids a host of tasks, from summarization to narrative generation to classification and prediction of events at points along the flow. In this paper, we present a framework for the automatic generation of an uncertain, temporally directed event graph from online sources such as news stories or social media posts. The vertices are generated using Natural Language Processing techniques on the source documents and the probabilities associated with edges, indicating the degree of certainty those connections exist, are derived based on shared entities among events. Graph edges are directed based on temporal information on events. Furthermore, we apply uncertain graph clustering in order to reduce noise and focus on higher-level event flows. Preliminary results indicate the uncertain event graph produces a coherent navigation through events described in a corpus.