figshare
Browse

NWB2023_Interdisciplinary research classification based on a combined conceptual-empirical framework

Download (1.01 MB)
poster
posted on 2023-10-10, 09:30 authored by Shunshun Shi, Lin Zhang

Measuring interdisciplinarity is a significant but challenging task in science quantitative studies. Various indicators have been proposed for measurement, but recent studies showed the majority of these indicators are unsatisfactory and that some even produce contradicting results. This problem is largely due to the fact that interdisciplinarity is a complex and multifaceted concept, and it is difficult for indicators to capture this complexity. Therefore, in this study, we argue for classifying interdisciplinary research (IDR) rather than measuring it directly. A combined conceptual-empirical framework is proposed to classify IDR. Specifically, at the conceptual level, four ideal types of IDR—Synthetic, Discovery, Diffusion, and Background—are provided in terms of their knowledge integration patterns; at the empirical level, bibliometrics based on full-text are used to extract citation features (e.g., citation mentioned, shared, length, and function features) of categories from IMR&D structure to characterize different knowledge integration patterns. Finally, these elected features are fed into deep learning classifiers (e.g., CNN and RNN) to achieve the final result of IDR classification. Our result shows that the number of articles of Discovery and Diffusion types accounted for the largest proportion of the total. The proportion of articles of Synthetic type that well-satisfy the core definition of IDR is slightly lower. The Background type accounted for the lowest. We therefore argue that classifying IDR using a well-designed framework is a feasible and reasonable solution to the current “measurement trap” and may offer the opportunity and foundation for subsequent measurement research of different IDR types.

History

Usage metrics

    Categories

    Licence

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC