Computing interdisciplinarity of scholarly objects using an author-citation-text model with a new measure

Abstract

In the 21st century, as the problems facing humanity grow, the importance of interdisciplinary research emerged as a solution. As a result, interdisciplinary research projects and support policies have increased, but there is a lack of research on indicators to measure the interdisciplinarity of proposed projects. Previous works measured the interdisciplinarity through two steps

they used a scholarly object model to calculate the distribution and then they measured its interdisciplinarity. However, while earlier scholarly objects use information such as authors or citations separately and previous interdisciplinarity measures shows low value when some salient disciplines are mixed with other disciplines. To tackle the problem, we propose a scholarly object model and an interdisciplinarity measure. This new model organizes types of information jointly through network/sentence embedding. The new interdisciplinarity measure separates certain salient disciplines and calculates the interdisciplinarity based on these disciplines. It can reduce the effect of non-salient disciplines during the calculation step. The experimental results show that using the joint information model can improve the performance of scholarly object classification and the results of the new measure are more similar to the results of human raters than are other measures.