. Applying Latent Semantic Analysis to Track Trends in Geoscience Research

Abstract
We describe the application of Latent Dirichlet Allocation (LDA), a generative probabilistic form of Latent Semantic Analysis, to discover topics and trends in scientific abstracts submitted to the European Geophysical Union (EGU) General Assemblies between 2009 and 2018, and to abstracts submitted to the Arctic Science Summit Week in 2018 (ASSW'18). The goal is to reveal the evolution of research topics being discussed in the abstracts and to assess appropriateness of the themes around which the various programme groups are organized. Topics derived from ASSW abstracts were analyzed to identify interdisciplinary themes that could be proposed for sessions in future ASSW meetings. The study developed Jupyter notebooks utilizing open source software routines for text preparation, analysis and visualization. These are being made available to the community for adaptation to any other document corpus.