Learn R Programming

⚠️There's a newer version (0.3.1.1) of this package.Take me there.

Rainette

The package website can be found at : https://juba.github.io/rainette/.

Rainette is an R package which implements a variant of the Reinert textual clustering method. This method is available in other software such as Iramuteq (free software) or Alceste (commercial, closed source).

Features

  • Simple or double clustering algorithms
  • Plot functions and shiny gadgets to visualise and explore clustering results
  • Utility functions to split a corpus into segments or import a corpus in Iramuteq format

Installation and usage

The package is installable from CRAN :

install_packages("rainette")

The development version is installable from Github :

remotes::install_github("juba/rainette")

Let's start with an example corpus provided by the excellent quanteda package :

library(quanteda)
data_corpus_inaugural

First, we'll use split_segments to split each text in the corpus into segments of about 40 words (punctuation is taken into account) :

corpus <- split_segments(data_corpus_inaugural, segment_size = 40)

Next, we'll compute a document-term matrix and apply some treatments with quanteda functions :

dtm <- dfm(corpus, remove = stopwords("en"), tolower = TRUE, remove_punct = TRUE)
dtm <- dfm_wordstem(dtm, language = "english")
dtm <- dfm_trim(dtm, min_termfreq = 3)

We can then apply a simple clustering on this dtm with the rainette function. We specify the number of clusters (k), the minimum size for a cluster to be splitted at next step (min_split_members) and the minimum number of forms in each segment (min_uc_size) :

res <- rainette(dtm, k = 6, min_uc_size = 15, min_split_members = 20)

We can use the rainette_explor shiny interface to visualise and explore the different clusterings at each k :

rainette_explor(res, dtm)

We can then use the generated R code to reproduce the displayed clustering visualisation plot :

rainette_plot(res, dtm, k = 5, type = "bar", n_terms = 20, free_scales = FALSE,
    measure = "chi2", show_negative = "TRUE", text_size = 10)

Or cut the tree at chosen k and add a group membership variable to our corpus metadata :

docvars(corpus)$group <- cutree_rainette(res, k = 5)

In addition to this, you can also perform a double clustering, ie two simple clusterings produced with different min_uc_size which are then "crossed" to generate more solid clusters. To do this, use rainette2 either on two rainette results :

res1 <- rainette(dtm, k = 10, min_uc_size = 10, min_split_members = 10)
res2 <- rainette(dtm, k = 10, min_uc_size = 15, min_split_members = 10)
res <- rainette2(res1, res2, max_k = 10, min_members = 20)

Or directly on a dtm with uc_size1 and uc_size2 arguments :

rainette2(dtm, max_k = 10, uc_size1 = 10, uc_size2 = 15, min_members = 20)

You can then use rainette2_explor, rainette2_plot and cutree_rainette2 to explore and visualise the results.

Tell me more

Three vignettes are available, an introduction in english :

And an introduction and an algorithm description, in french :

Credits

This classification method has been created by Max Reinert, and is described in several articles. Here are two references :

  • Reinert M, Une méthode de classification descendante hiérarchique : application à l'analyse lexicale par contexte, Cahiers de l'analyse des données, Volume 8, Numéro 2, 1983. http://www.numdam.org/item/?id=CAD_1983__8_2_187_0
  • Reinert M., Alceste une méthodologie d'analyse des données textuelles et une application: Aurelia De Gerard De Nerval, Bulletin de Méthodologie Sociologique, Volume 26, Numéro 1, 1990. https://doi.org/10.1177/075910639002600103

Thanks to Pierre Ratineau, the author of Iramuteq, for providing it as free software and open source. Even if the R code has been almost entirely rewritten, it has been a precious resource to understand the algorithms.

Many thanks to Sébastien Rochette for the creation of the hex logo.

Many thanks to Florian Privé for his work on rewriting and optimizing Rcpp code.

Copy Link

Version

Install

install.packages('rainette')

Monthly Downloads

288

Version

0.1.2

License

GPL (>= 3)

Maintainer

Julien Barnier

Last Published

January 20th, 2021

Functions in rainette (0.1.2)

cutree_rainette

Cut a rainette result tree into groups of documents
order_docs

return documents indices ordered by CA first axis coordinates
rainette2_complete_groups

Complete groups membership with knn classification
import_corpus_iramuteq

Import a corpus in Iramuteq format
cutree_rainette2

Cut a rainette2 result object into groups of documents
cluster_tab

Split a dtm into two clusters with reinert algorithm
split_segments

Split a character string or corpus into segments
rainette

Corpus clustering based on the Reinert method - Simple clustering
rainette2_plot

Generate a clustering description plot from a rainette2 result
rainette2_explor

Shiny gadget for rainette2 clustering exploration
rainette_explor

Shiny gadget for rainette clustering exploration
switch_docs

Switch documents between two groups to maximize chi-square value
rainette_stats

Generate cluster keyness statistics from a rainette result
select_features

Remove features from dtm of each group base don cc_test and tsj
rainette_plot

Generate a clustering description plot from a rainette result
compute_uc

Merges uces into uc according to minimum uc size
rainette2

Corpus clustering based on the Reinert method - Double clustering
cutree

Cut a tree into groups