The algorithm of research submission and publication is strict and highly formal. This produces lots of metadata accompanying research papers. The analysis of such datasets is not a mainstream research field nowadays. Yet such an analysis could help us better understand certain large-scale patterns of contemporary academic publishing, and maybe even the dynamics of science.
The paper presents experiments in mix-method analysis of academic research papers and respective metadata. We investigate a corpus of abstracts and metadata for 31, 000+ papers from the arXiv.org e-prints archive. Our experiments reveal the properties of interinstitutional research and the collaboration of the industry and academy. We also do visualize and interpret the temporal properties of research trends.