Detecting trending topics using page visitation statistics
Sayandev Mukherjee, Ronald Sujithan, Pero Subašić · 2014
Many applications including realtime recommenders and ad-targeting systems have a need to identify trending concepts to prioritize the information presented to end-users. In this paper, we describe a novel approach that identifies trending concepts using the hourly Wikipedia page visitation statistics freely available for download. We describe a MapReduce framework that analyzes the raw hourly visitation logs and generates a ranked list of trending concepts on a daily basis. We validate this approach by extracting hourly lists of trending news articles, mapping these articles to Wikipedia concepts, and computing the similarity of the two lists according to several commonly used measures.