Neural Embeddings of Scholarly Periodicals Reveal Complex Disciplinary Organizations
share
Understanding the structure of knowledge domains has been one of the foundational challenges in the science of science. Although there have been a series of studies on the self-organized structure of knowledge domains and their relationships, creating rich, coherent, and quantitative representation frameworks is still an open challenge. Meanwhile, neural embedding methods, which learn continuous vector representations of entities by using neural networks and contextual information, are emerging as a powerful representation framework that can encode nuanced semantic relationships into geometric ones. Here, we propose a neural embedding technique that leverages the information contained in the paper citation network to obtain continuous representations of scientific periodicals. We demonstrate that our embeddings encode nuanced relationships between periodicals as well as the complex disciplinary structure of science, even allowing us to make cross-disciplinary analogies between periodicals. Furthermore, we show that the embeddings capture meaningful "axes" that encompass all disciplines in the knowledge domains, such as an axis from "soft" to "hard" sciences or from "social" to "biological" sciences, which allow us to quantitatively ground a periodical on a given spectrum. Using this new capacity, we test the hypothesis of the hierarchy of the sciences, showing that, in most disciplines such as Social Sciences and Life Sciences, most widely cited papers tend to appear in "harder" periodicals. Our framework may offer novel quantification methods in science of science, which may in turn facilitate the study of how knowledge is created and organized.
READ FULL TEXT