UC Riverside

This dissertation presents an investigation into the structural dynamics of scientific semantic networks, examining these networks across 16 scientific fields at three distinct levels of analysis: individual terms, communities of terms, and broader fields. Utilizing Exponential Random Graph Models (ERGM), the research explores critical local structures within these networks, comparing the semantic structures of the hard and soft sciences. Contrary to expectations, the findings reveal minimal differences between these disciplines, although some notable structural features are identified. The analysis extends to track the evolution of term communities over time, employing Singular Value Decomposition (SVD) to identify core and peripheral semantic communities. Again, contrary to expectations, the findings reveal reveal minimal differences between different types of fields. Finally, the study delves into the genealogy of these semantic communities, exploring patterns of split and merge events that contribute to the formation of semantic lineages. The findings reveal patterns of semantic inheritance, with some lineages enduring over long periods, others fading or disappearing, some undergoing division, and others merging or being subsumed under new terminological frameworks. Additionally, certain terms remain isolated, demonstrating the varied trajectories of semantic evolution within these networks. The insights garnered from this research not only shed light on the static and dynamic aspects of scientific semantic networks but also contribute to a deeper understanding of the interdisciplinary similarities and evolutionary trends in scientific terminology.