Extracting a causal network of news topics

Abstract

Because of the abundance of online news, it is impossible for users to process all the available information. Tools are needed to help process this information. To mitigate this challenge we propose generating a network of causally related news topics to help the user understand and navigate throughout the news. We assume that by providing the causes or effects of a news topics, the user will be able to relate current news to past news topics that the user knows about, or that the user will discover past news topics as currently relevant. Also, the additional context will facilitate the understanding of the current news topic. To generate the causal network, information is extracted from several distributed news sources while maintaining important journalistic features such as source referencing and author attribution. We propose ranking different causes of an event, to provide a more intuitive summary of multiple causal relations. To make the network easily understandable, news topics must be represented in a format that can be causally related therefore, a news topic model is proposed. The model is based on the phrases used by online news sources to describe an event or activities, during a limited time-frame. To maintain usability the results must be provided in a timely manner from streaming sources and in an easy to understand format. © 2012 Springer-Verlag.