Graphical Techniques for Geo-Semantic Analysis of the Shoah Foundation Visual History Archive and Other Collections

Peter Broadwell
University of California, Los Angeles (UCLA)

Four undergraduate participants in the UCLA Institute for Pure and Applied Mathematics' summer Research in Industrial Projects for Students program, in collaboration with the USC Shoah Foundation Institute for Visual History and Education, conducted knowledge discovery analyses to improve the topical and geospatial search capabilities of the Shoah Foundation's database of Holocaust-related video testimonies. The graph-based Louvain community detection algorithm and the Affinity Propagation clustering algorithm, when applied to geographical co-occurrence vectors and topical term frequency scores, proved useful for identifying and visualizing geo-semantic relationships. These techniques also yielded promising results when used to analyze a geo-referenced collection of folklore that had previously been studied via divisive graph clustering techniques.

Back to Networks and Network Analysis for the Humanities: Reunion Conference