<p>Interdisciplinarity has long been characterized and quantified through the notion of ‘knowledge integration’, yet this singular focus has created an imbalance by neglecting the equally vital aspect of ‘problem orientation’. To address this gap, we propose a taxonomic framework for characterizing interdisciplinarity from a ‘problem-solving’ perspective that holistically integrates ‘problem orientation’ with ‘knowledge integration’. This framework delineates distinct types based on the specific interplay between these two core aspects. We operationalize it by combining manual coding with deep learning techniques to analyze relevant textual data. Within this framework, we identify five distinct types of interdisciplinarity: Synthetic, Exploratory, Spillover, Peripheral, and Miscellaneous. An empirical inquiry focusing on COVID-19-related research is conducted to test the feasibility of the proposed methodology for characterizing interdisciplinarity and compare its results with those from reference-based diversity indicators. The&#xa0;results demonstrate that the values of diversity indicators cannot be directly compared without breaking interdisciplinary researches down into distinct types and contextualizing them. Finally, we discuss several limitations of the proposed taxonomy and highlight areas for further exploration and refinement in future work.</p>