Investigating Rumor News Using Agreement-Aware Search.

In recent years, rumor news has been generated by humans as well as robots inorder to attract readership, influence opinions, and increase internet clickrevenue. Its detrimental effects have become a worldwide phenomenon, leading toconfusion over facts and causing mistrust about media reports. However,evaluating the veracity of news stories can be a complex and cumbersome task,even for experts. One of the challenging problems in this context is toautomatically understand different points of view, i.e., whether other newsarticles reporting on the same problem agree or disagree with the referencestory. This can then lead to the identification of news articles that propagatefalse rumors.

Here, we propose a novel agreement-aware search framework, Maester, fordealing with the problem of rumor detection. Given an investigative questionsummarizing some news story or topic, it will retrieve related articles to thatquestion, assign and display top articles from agree, disagree, and discusscategories to users, and thus provide a more holistic view. Our work makes twoobservations. First, relatedness can commonly be determined by keywords andentities occurred in both questions and articles. Second, the level ofagreement between the investigative question and the related news article canoften be decided by a few key sentences. Accordingly, we design our approachfor relatedness detection to focus on keyword/entity matching using gradientboosting trees, while leveraging recurrent neural networks and posingattentions to key sentences to infer the agreement level. Our evaluation isbased on the dataset from the Fake News Challenge (FNC). Extensive experimentsdemonstrate up to an order of magnitude improvement of Maester over allbaseline methods for agreement-aware search as well as slightly improvedaccuracy based on the same metrics used in FNC.