This app enables interactive exploration of Structural Topic Models. For most of the sidebar inputs a short info is available if you hover over them. In case you are not familiar with STM the authors describe the model in the corresponding package vignette.
The model included in this app was fitted on TED talk transcripts from 2006 until 2017.
Contact me: Email email@example.com / Twitter c_schwemmer
Inspect the most important terms for the selected topic.
Inspect the documents with the highest proportions for the selected topic.
Use the boxes to enter labels for your topics. It is suggested to keep the labels short, such that plot outputs will not be flooded with text.
Select the STM model and corresponding effect estimates here.
This plot shows the topic proportions over all documents.
Create a network based upon correlations between topics. Node sizes are scaled by topic proportions. Edge thickness is determined by pairwise topic correlations.
Parameters for the selected model and effect estimation are listed below.
Semantic Coherence and Exclusivity for all models in your environment are plotted if desired. This might take a while.
If you assigned labels to your topics you can download a corresponding table here.Download topic labels