Untangling the Academic Network: Analyzing Scientific Trajectories With Graphs and Natural Language Processing
Abstract
We analyze 3,143 Chilean doctoral fellows from the 2014–2016 ANID cohorts through a Knowledge Graph that federates eight heterogeneous sources into 16 node types, 15 relationship types, and over 3 million edges. BERTopic yields 237 topics over 26,071 Scopus abstracts. In the intra-cohort co-authorship network, 84.2% of fellows share no publication with another fellow, and its largest component splits into 21 communities that are thematically coherent but heterogeneous in gender and socioeconomic composition.
Type
Publication
In 17th Alberto Mendelzon International Workshop on Foundations of Data Management
