Visalization of learned topics for the paper "A Geometry-Drive Longitudial Topic Model" submitted to HDSR by Y. Wang, C. Hougen, B. Oselio, W. Dempsey, A.O.Hero.
By default the main notebook scripts/topicMatching.ipynb loads learned topics and associated data from a folder learned_topics. This folder can be downloaded here. Note that in the data folder we have included learned topics from an extended period, i.e., May 15 to Aug 15, which was not included in the main paper.
The notebook scripts/TopicFlow.ipynb loads the same set of learned topics, but assembles topic trends using the method called TopicFlow. The notebook scripts/simulation_phate.ipynb presents simulation studies for PHATE-Helliner dimensionality reduction.
An associated web application built by James Chu and Yu Wang via shinyapp is hosted here. The web app presents spatio-temporal visualization of the topic modelling results.