Finding and Explaining Similarities in Linked Data.
Catherine Olsson, Plamen V. Petrov, Jeffrey W. Sherman, Andrew Perez-Lopez · 2011
Abstract—Today’s computer users and system designers face increasingly vast amounts of data, yet lack good tools to find pertinent information within those datasets. Linked data technologies add invaluable structure to data, but challenges remain in helping users understand and exploit that structure. One important question users might ask about their data is “What entities are similar to this one, and why? ” or “How similar are these two entities to one another, and why?”. Our work focuses on using the semantic content of linked data not only to facilitate the process of finding similar entities, but also to produce automatically-generated and human-understandable explanations of what makes those entities similar. In this paper, we formulate a definition of an “explanation ” of similarity, we describe a system that can produce such explanations efficiently, and we present a methodology to allow the user to tailor how “obvious ” or “obscure ” the provided explanations are. I.