A Scalable Tool for Longitudinal Twitter Analysis

Abstract: The open forum of social media provides an opportunity to collect novel data that can by used to study the patterns of social discourse around the COVID-19 pandemic.   We have developed a simple and scalable framework for longitudinal analysis of Twitter data that combines latent topic models (LDA) with computational geometry embeddings to propagate and associate topics over time.   In this talk, we illustrate how the framework can be applied to capture natural progressions of latent topics that are either explicitly or implicitly COVID-related.