Our main goal for this project was to train the most efficient topic evolution model on large
datasets consisting of news articles. Efficiency was determined based on the following criteria:
model perplexity, topic coherence, topic diversity, and topic interpretability.
This project collected data that was associated with the current Covid-19 pandemic which brings
a new and relevant addition to existing research on topic evolution models. In contrast to most
existing research, we took a more globalized approach and tracked the development of narrow
topics like ‘anxiety’, ‘mental health’, and ‘lockdown’ in the context of countries like India and
Brazil.
Term
Spring 2021
Topic
Public Health