Creative Knowledge Discovery by Literature Outlier Detection

This paper investigates the role of outliers in literature-based knowledge discovery. It shows that detecting interesting outliers that appear in the literature about a given phenomenon can help generate novel plausible scientific hypotheses. The underlying assumption is that whereas the majority of domain literatures describe matters related to common understanding of the domain, some particular observations that appear rarely in the literature can indicate a promising direction towards novel discoveries. This rarity principle is used in our method called RaJoLink to guide the knowledge discovery process. The presented method focuses on the role of outliers in the closed discovery process as implemented in the RaJoLink literature mining methodology.