Abstract
In many scientific disciplines, each new ‘product’ of research (method, finding, artifact, etc.) is often built upon previous findings–leading to extension and branching of scientific concepts over time. We aim to understand the evolution of scientific concepts by placing them in phylogenetic hierarchies where scientific keyphrases from a large, longitudinal academic corpora are used as a proxy of scientific concepts. These hierarchies exhibit various important properties, including power-law degree distribution, power-law component size distribution, existence of a giant component and less probability of extending an older concept. We present a generative model based on preferential attachment to simulate the graphical and temporal properties of these hierarchies which helps us understand the underlying process behind scientific concept evolution and may be useful in simulating and predicting scientific evolution.
Highlights
In numerous domains, science work produces specific ‘artifacts,’ including models, methods, theories, algorithms, systems
We focus on four properties: degree distribution, component size distribution, tree depth and year difference between first appearances of parent and child concepts
As we have finite data and we are interested in properties of the graph other than degree distribution, we resort to computer simulation
Summary
Science work produces specific ‘artifacts,’ including models, methods, theories, algorithms, systems. (We distinguish, though only nominally, between produced (i.e., through construction) and discovered. A discovery may include the identification of new natural ‘entities’ (e.g., organisms, genes, planets).) In most situations, produced artifacts–or concepts when they are named– connect to others: one model may build on another; a new theory may refine or compete with an existing one, and one algorithm may extend another. We can model this ‘progression’ as a dynamic graph, where nodes, representing artifacts are added over time. Taken together these form three nodes in our broader graph
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
