A project about capturing political polarization in news coverage by Topic Modeling using Latent Dirichlet Allocation (LDA) and Latent Semantic Indexing (LSI).
To run this project, open client.ipynb
. This will import all the necessary modules and load pretrained models from pickle.
The notebook provides a short plot demo for a topic search using LSI that suggests which network to watch given a particular query. For example, you can enter a query like stormy_daniels russia putin
and find out which network has aired more episodes that resemble the query.
The excellent pyLDAvis
package is used to provide interactive tools for topic exploration on the two networks.