“…The first step in the proposed method is to identify a small number of key, or "seed," articles on which citation relationships can subsequently be based. Starting with a set of key articles echoes the use of training sets for machine learning techniques and has been proposed for bibliometric tasks in a variety of contexts, such as identifying journal relationships (Medina & van Leeuwen, 2012), for author name disambiguation (Reijnhoudt, Costas, Noyons, Börner, & Scharnhorst, 2014), and information retrieval (Glänzel, 2015). How many key articles are necessary for the proposed method to work is an open question.…”