Bio-inspired Methods for Dynamic Network Analysis in Science Mapping

TL;DR

This paper explores bio-inspired methods to analyze dynamic bibliometric networks, comparing different classification techniques over time to identify coherent research structures and their evolution.

Contribution

It introduces a novel bio-inspired framework for comparing bibliometric classifications using biological species concepts and similarity indexes.

Findings

Different bibliometric indicators can detect convergent research structures.

Classification comparisons reveal how research specialties evolve over time.

Bio-inspired methods effectively analyze dynamic science mapping.

Abstract

We apply bio-inspired methods for the analysis of different dynamic bibliometric networks (linking papers by citation, authors, and keywords, respectively). Biological species are clusters of individuals defined by widely different criteria and in the biological perspective it is natural to (1) use different categorizations on the same entities (2) to compare the different categorizations and to analyze the dissimilarities, especially as they change over time. We employ the same methodology to comparisons of bibliometric classifications. We constructed them as analogs of three species concepts: cladistic or lineage based, similarity based, and "biological species" (based on co-reproductive ability). We use the Rand and Jaccard indexes to compare classifications in different time intervals. The experiment is aimed to address the classic problem of science mapping, as to what extent the various techniques based on different bibliometric indicators, such as citations, keywords or authors are able to detect convergent structures in the litrerature, that is, to identify coherent specialities or research directions and their dynamics.