Discovering, ranking and annotating cross-document relationships between concepts

This paper presents CDRMiner, a system for automatically discovering, ranking and annotating cross-document links between concepts. Specifically, we focus on detecting hidden associations between two concepts and further generating annotations for each discovered hypothesis. We interpret such a relationship query as finding the most meaningful concept chains and evidence trails across multiple documents that potentially connect them. These functionalities are implemented using an interactive visualization paradigm which assists users for a better understanding and interpretation of discovered hypotheses, and matching their domain knowledge with the algorithmic power of text mining techniques.