Abstract
When quantifying the level of interdisciplinarity for scientific research, most established indicators employ a three‐element diversity framework, namely variety, balance, and disparity, each of which captures a distinct but insufficient element. Among three, disparity, i.e. how different (dissimilar) the categories within a system are, is the most challenging one due to its calculation cost and conceptual ambiguity. The discriminative power for disparity is found to be weakened in more fine‐grained science classification schemes. To address this issue, this paper proposes a new method for quantifying disparity by applying Node2vec on the discipline citation network and retrieving distance between disciplines using embeddings vectors. Compared to cosine‐based dissimilarity for disparity, our proposed method exhibited broader distribution and less skewness for disparity values, which could potentially lead to higher discriminative power of interdisciplinarity. A case study for Linguistics is also conducted to show the capability of detecting variations in disparity of the proposed method.
Funding Information
  • Vlaamse regering

This publication has 4 references indexed in Scilit: