Characterization of researchers using multidisciplinary measures: A mining-based approach to scientific publications
The area of scientometrics has been studying for decades robust ways of measurement of science in several aspects. In particular, the most visited subjects detail science from a qualitative and quantitative perspective using information from bibliometric sources. The Trans-, inter-, pluri- and multidisciplinary research are different forms of relationship or interaction between different scientific disciplines. Currently, these forms of interaction are desirable in academia and motivated by different science and technology agencies as well as research funding agencies. In this context, the characterization of researchers by multidisciplinary scientometric indicators (which corresponds to the simultaneous grouping of disciplines without necessarily representing a strong integration), is a first step towards the discovery of knowledge about the relationship of disciplines from bibliographic data sources and/or incomplete curricula.
In this work we present the creation of computational methods to analyze, over time, how multidisciplinary the scientific performance of research groups has been. We also formed a multidisciplinary network where we verify the integrations: researcher-discipline, researcher-researcher and discipline-discipline, using the structure of heterogeneous and homogeneous graphs. The computational procedure considers the traditional approach to knowledge discovery where mining techniques (e.g., time series and pattern recognition) are applied to bibliographic and curriculum data are used.