Measuring cognitive distance between publication portfolios

Journal article, 2017

We study the problem of determining the cognitive distance between the publication portfolios of two units. In this article we provide a systematic overview of five different methods (a benchmark Euclidean distance approach, distance between barycenters in two and in three dimensions, distance between similarity-adapted publication vectors, and weighted cosine similarity) to determine cognitive distances using publication records. We present a theoretical comparison as well as a small empirical case study. Results of this case study are not conclusive, but we have, mainly on logical grounds, a small preference for the method based on similarity-adapted publication vectors