Their full paper is here http://cepr.org/sites/default/files/news/CEPR_FreeDP_180416.pdf http://cepr.org/sites/default/files/news/CEPR_FreeDP_180416.... . If I

Their full paper is here http://cepr.org/sites/default/files/news/CEPR_FreeDP_180416.pdf http://cepr.org/sites/default/files/news/CEPR_FreeDP_180416.... . If I followed it correctly they measure "novelty" by: - bucketing citations into journals - building a network of journals with edge weights defined by the number of papers that cite both journals - Using cosine similarity between journals (cf https://en.wikipedia.org/wiki/Similarity_(network_science)#Cosine_similarity https://en.wikipedia.org/wiki/Similarity_(network_science)#C... ) in this network to define how banal it is to cite them together - For a new paper, they recompute this matrix and define novelty by summing 1-banality over the new edges I don't know if their method reduces to a standard thing from network science (eg. betweenness, centrality, ?) but I would not be surprised if it did. A closely-related paper with a totally-different writing style did something like this a few years back: "Citing for High Impact" https://cs.stanford.edu/people/jure/pubs/citations-jcdl10.pdf https://cs.stanford.edu/people/jure/pubs/citations-jcdl10.pd...