The Evolution of Scientific Paper Title Networks

Abstract

In spite of enormous previous efforts to model the growth of various networks, there have only been a few works that successfully describe the evolution of latent networks. In a latent network edges do not represent interactions between nodes, but show some proximity values. In this paper we analyze the structure and evolution of a specific type of latent networks over time by looking at a wide range of document similarity networks, in which scientific titles are nodes and their similarities are weighted edges. We use scientific papers as the corpora in order to determine the behavior of authors in choosing words for article titles. The aim of our work is to see whether term selection for titles depends on earlier published titles.