Year: 2021
Type: article
Abstract: In this paper we utilize an opportunity to construct ground truths for topics in the field of atomic, molecular and optical physics. Our research questions in this paper focus on (i) how to construct a ground truth for topics and (ii) the suitability of common algorithms applied to bibliometric networks to reconstruct these topics. We use the ground truths to test two data models (direct citation and bibliographic coupling) with two algorithms (the Leiden algorithm and the Infomap algorithm). Our results are discomforting: none of the four combinations leads to a consistent reconstruction of the ground truths. No combination of data model and algorithm simultaneously reconstructs all micro-level topics at any resolution level. Meso-level topics are not reconstructed at all. This suggests (a) that we are currently unable to predict which combination of data model, algorithm and parameter setting will adequately reconstruct which (types of) topics, and (b) that a combination of several data models, algorithms and parameter settings appears to be necessary to reconstruct all or most topics in a set of papers.
Source: Scientometrics
Authors: Matthias Held, Grit Laudel, Jochen Gläser
Institution: Technische Universität Berlin
Language: English
Subfield: Statistics, Probability and Uncertainty
Field: Decision Sciences
Domain: Social Sciences
Open Access status: hybrid
Estimated APC paid: $2,780
Award: G1225137271
OpenAlex ID: W3133819788