Author: Bradford Demarest

Publisher:

Published: 2021

Total Pages: 0

ISBN-13:

There has been a convergence of economic, epistemological, social, and communicative trends over recent decades that shape the ways scholars communicate within their fields, across fields with other academics, and with non-academic entities in industry, government, and the lay public. Consequently, scholars must not only convince those controlling funding that their research is relevant within their field, but also that it is more relevant than research conducted in other fields, and they must spell out its relevance to the wider world. Meanwhile, more interdisciplinary research is encouraged. In order for this research to yield successful results however, researchers from different disciplinary backgrounds need to recognize, and reconcile, differences in social and epistemological assumptions. Although linguistics research has developed and tested frameworks to identify and describe such characteristics in the written discourse of scientific and scholarly communities, scientometrics has heretofore modeled disciplinary and interdisciplinary social and communicative structures through patterns of citation and reference, authorial collaborations, and conceptual proximity, but left aside measuring and describing disciplinary cultures based on written social and epistemic discourse.This dissertation proposes and investigates a new method for evaluating the degree and kind of social and epistemic differences between academic disciplines based on these previous studies of discourse, via three studies. In the discourse epistemetrics method (Demarest & Sugimoto, 2015), texts of a given genre are collected and grouped by the knowledge-oriented community that created the texts, translated into vectors of frequencies of lexical terms commonly used to express social and epistemic stance, and modeled using Support Vector Machines (SVMs), with accuracy rates used as metrics of distance between disciplines and feature weights conveying specific terms indicating one discipline or the other. The first study establishes a proof-of-concept, modeling differences between philosophy, psychology, and physics dissertation abstracts and testing the accuracy and interpretability of five feature sets. The second study expands the set of disciplines to 14 and tests whether a network of disciplines can be derived and interpreted, as well as comparing the resulting network to one derived from discipline-level bibliographic coupling data for the same disciplines. The third and final study compares networks for abstracts and full texts from the same papers from five disciplines. The discourse epistemetrics method is found to distinguish accurately and meaningfully between disciplines, especially as applied to abstracts, and is proposed as a useful tool for search ranking as well as aiding interdisciplinary scholarship.