Topic modeling in scientometrics: Community, connectivity, and change
- Author(s): Bergmann, Till Christian
- Advisor(s): Matlock, Teenie
- Dale, Rick
- et al.
This dissertation examines the complex structure of scientific organization and publication behavior. Since the last century, the number of scientific publications has exponentially risen, and researchers are now more connected than ever. This has led to an increasing interest in quantifying the structure of academia at var- ious levels, for exampling university rankings and journal impact factors. In the current work, abstracts of scientific publication will be analyzed with respect to three different features of academia. First, the internal structure of a scientific community will be examined. What are the research topics prevalent in a comunity? Which are neglected? Does collaboration between researchers facilitate or hinder topic popularity? We find that more central authors in the community publish on a distinct set on research areas than non-central authors. Second, the connectivity between different scientific communities will be analyzed. Using quantitative methods, the overlap in scientific content between related scientific fields will be measured. Despite claims from within the community, the quantitative analysis shows very little overlap between supposedly related areas. Third, temporal change of scientific fields will be investigated. Taking two unrelated fields, philosophy and biology, the change of topics is evaluated over time. It is shown that biology as a field undergoes more change than philosophy, and the patterns of topic change differ across the two fields. In biology, topics either in- crease or decrease in popularity, while in philosophy their popularity fluctuates up and down over time.
These different aspects of scientific organization will be examined using topic models, a tool from natural language processing, and extended by various methods for each chapter. The theoretical discussion will argue that the results obtained in the case studies are heavily influenced by group cognition, that is, pressures and influences inherent to social groups.
This dissertation, Topic modeling in scientometrics: Community, connectivity, and change, is submitted by Till Christian Bergmann in 2016 in partial fulfillment of the degree Doctor of Philosophy in Cognitive and Information Sciences at the University of California, Merced, under the guidance of dissertation committee co-chairs Teenie Matlock and Rick Dale.