stminsights

Travis-CI Build Status AppVeyor Build Status

A Shiny Application for Structural Topic Models

This app enables interactive validation, interpretation and visualisation of Structural Topic Models (STM). In case you are not familiar with STM, the package vignette is an excellent starting point. In contrast to stmgui, which also allows to preprocess text data and fit models, stminsights is focused more on making your life easier after these steps.

How to Install

You can download and install the latest development version of the app by running devtools::install_github('methodds/stminsights').

For Windows users installing from github requires proper setup of Rtools, for which a tutorial is available here.

stminsights can also be installed from CRAN by running install.packages('stminsights').

How to Use

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: - one or several stm objects. - one or several estimateEffect objects. - an object 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)

out <- list(documents = poliblog5k.docs,
            vocab = poliblog5k.voc,
            meta = poliblog5k.meta)

poli <- stm(documents = out$documents, 
            vocab = out$vocab,
            data = out$meta, 
            prevalence = ~ rating * s(day),
            K = 20)
prep_poli <- estimateEffect(1:20 ~ rating * s(day), poli,
                            meta = out$meta)

poli_content <-  stm(documents = out$documents, 
                     vocab = out$vocab,
                     data = out$meta, 
                     prevalence = ~ rating + s(day),
                     content = ~ rating,
                     K = 15)  
prep_poli_content <- estimateEffect(1:15 ~ rating + s(day), poli_content,
                                    meta = out$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).

How to Deploy on Shiny Server

Deploying stminsights to your own shiny server is simple: 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.

To test stminsights on my server, you can download stm_poliblog5k.RData here and upload it at www.polsoz.uni-bamberg.de/stminsights.

Citation

If you use stminsights for your publications please consider citing it:

  Carsten Schwemmer (2018). stminsights: A Shiny Application for Inspecting
  Structural Topic Models. R package version 0.2.0.
  https://github.com/methodds/stminsights

A BibTeX entry for LaTeX users is:

  @Manual{,
    title = {stminsights: A Shiny Application for Inspecting Structural Topic Models},
    author = {Carsten Schwemmer},
    year = {2018},
    note = {R package version 0.2.0},
    url = {https://github.com/methodds/stminsights},
  }