Detecting Pandemic Related R&D Trends using Dynamic Topic Modeling

Report
Authors:Linehan, Kathryn, PV-BII SDADUniversity of Virginia ORCID icon orcid.org/0000-0001-9012-6261SIWE, Guy Leonel, PV-BII SDADUniversity of Virginia ORCID icon orcid.org/0000-0002-9275-6416Thurston, Joel, PV-BII SDADUniversity of Virginia ORCID icon orcid.org/0000-0002-3923-9065Shipp, Stephanie, PV-BII SDADUniversity of Virginia ORCID icon orcid.org/0000-0002-2142-2136Kindlon, AudreyNational Center for Science and Engineering Statistics Jankowski, John National Center for Science and Engineering Statistics
Abstract:

This report explores the use of dynamic nonnegative matrix factorization (D-NMF) to identify topics and trends in federally funded R&D related to pandemics. We utilized scientific grant abstract data and compared D-NMF and static NMF topic model results on this data. We report topics discovered, topic prevalence, and provide an initial exploration into measures of D-NMF topic content change. While
we found that D-NMF and static NMF identify many similar topics, we show that the D-NMF model can provide insightful supplementary results to those produced by the static NMF, and that the synthesis of results from D-NMF and static NMF can offer a more complete topic and trend analysis.
 

Rights:
Attribution 4.0 International (CC BY)
Contributor:Lyman, Kimberly, PV-BII SDADUniversity of Virginia
Language:
English
Source Citation:

Linehan K, Siwe GL, Thurston J, Shipp S, Kindlon A, Jankowski J. Detecting Pandemic Related R&D Trends using Dynamic Topic Modeling. [Other]. Biocomplexity Institute, University of Virginia; 2023 February. Available from: https://doi.org/10.18130/zz1b-ww76

Publisher:
University of Virginia
Published Date:
January 31, 2023
Sponsoring Agency:
National Center for Science and Engineering Statistics (NCSES)
Persistent Link:
https://doi.org/10.18130/zz1b-ww76