stm: An R Package for Structural Topic Models

Abstract

This paper demonstrates how to use the R package stm for structural topic modeling. The structural topic model allows researchers to flexibly estimate a topic model that includes document-level metadata. Estimation is accomplished through a fast variational approximation. The stm package provides many useful features, including rich ways to explore topics, estimate uncertainty, and visualize quantities of interest.

Adding rooms onto a house we love: Central banking after the global financial crisis

“…Core includes those that capture the key elements of the original monetary policy paradigm 4 We use software by Benoit et al (2018) to remove stop words, numbers, and punctions; and to remove inflections from words in order to reduce them to their roots. We use software by Roberts et al (2018) to estimate a Correlated Topic Model (Blei and Lafferty 2007). This model is preferable to alternatives like the Latent Dirichlet Allocation, because it allows for certain topics to appear together more frequently than others.…”

Analyzing big data in social media: Text and network analyses of an eating disorder forum

“…The dataset for topic modelling contains texts by N 5 3,009 users. The latent Dirichlet allocation implementation in the stm package (Roberts, Stewart, & Tingley, 2017) was applied as it allows for correlated topics and the optional addition of external covariates.…”

Central banks’ communication as reputation management: How the Fed talks under uncertainty

“…'Fiscal policy' is the fourth most prevalent topic, in a reflection of the importance of fiscal decisions for the conduct of monetary policy. 10 We estimate topic models using the stm package in R (Roberts et al 2018). 11 We also run different model specifications including a larger set of covariates.…”