This app enables interactive validation, interpretation and visualization of Structural Topic Models (STM). Stminsights is focused on making your life easier after fitting your STM models. In case you are not familiar with the STM package, the corresponding vignette is an excellent starting point.
You can download and install the latest development version of
stminsights by running
devtools::install_github('cschwem2er/stminsights')
.
For Windows users installing from github requires proper setup of Rtools.
stminsights can also be installed from CRAN by running
install.packages('stminsights')
.
After loading stminsights you can launch the shiny app in your browser:
library(stminsights)
run_stminsights()
You can then upload a .RData
file which should
include:
stm
objects.estimateEffect
objects.out
which was used to fit your stm
models.As an example, the following code fits two models and estimates
effects for the Political Blog Corpus. Afterwards, all objects required
for stminsights are stored in stm_poliblog5k.RData
.
library(stm)
<- list(documents = poliblog5k.docs,
out vocab = poliblog5k.voc,
meta = poliblog5k.meta)
<- stm(documents = out$documents,
poli vocab = out$vocab,
data = out$meta,
prevalence = ~ rating * s(day),
K = 20)
<- estimateEffect(1:20 ~ rating * s(day), poli,
prep_poli meta = out$meta)
<- stm(documents = out$documents,
poli_content vocab = out$vocab,
data = out$meta,
prevalence = ~ rating + s(day),
content = ~ rating,
K = 15)
<- estimateEffect(1:15 ~ rating + s(day), poli_content,
prep_poli_content meta = out$meta)
save.image('stm_poliblog5k.RData')
After launching stminsights and uploading the file, all objects are automatically imported and you can select which models and effect estimates to analyze.
In addition to the shiny app, several helper functions are available,
e.g. get_effects()
for storing effect estimates in a tidy
dataframe.
To deploy stminsights to your own shiny server, place the file
app.R
, which is located at inst/app
of this
package, to a folder in your server directory and you should be good to
go.
Please cite stminsights if you use it for your publications:
Carsten Schwemmer (2021). stminsights: A Shiny Application for Inspecting
Structural Topic Models. R package version 0.4.1.
https://github.com/cschwem2er/stminsights
A BibTeX entry for LaTeX users is:
@Manual{,
title = {stminsights: A Shiny Application for Inspecting Structural Topic Models},
author = {Carsten Schwemmer},
year = {2021},
note = {R package version 0.4.1},
url = {https://github.com/cschwem2er/stminsights},
}