Package installation

The package can be installed from CRAN by

install.packages("ggwordcloud")

or the development version from the github repository

devtools::install_github("lepennec/ggwordcloud")

Wordcloud

As an example, we will use the mtcars dataset

dat <- mtcars
dat$name <- row.names(mtcars)

Let’s load the package and set the random seed.

library(ggwordcloud)
#> Loading required package: ggplot2
set.seed(42)

A basic wordcloud can be obtained by:

ggplot(data = dat, aes(label = name)) + geom_text_wordcloud() + theme_minimal()

We have used a minimal theme to display only the words.

Because there is some randomness in the algorithm, the same command can yield a different result:

ggplot(data = dat, aes(label = name)) + geom_text_wordcloud() + theme_minimal()

Wordcloud and text size

We will use the mpg variable to define a size and modify two cars so that they are much larger than the other ones:

dat$size <- dat$mpg
dat$size[c(1,4)] <- dat$size[c(1,4)] + 100

We can now add this new variable to the aesthetic:

ggplot(data = dat, aes(label = name, size = size)) + geom_text_wordcloud() +
  theme_minimal()

In order to obtain a better picture, one can play with the size scale:

ggplot(data = dat, aes(label = name, size = size)) + geom_text_wordcloud() +
  scale_size(range = c(2,12)) +
  theme_minimal() 

Note that words that cannot be placed due to a lack of space are displayed at their original position:

ggplot(data = dat, aes(label = name, size = size)) + geom_text_wordcloud() +
  scale_size(range = c(4,20)) +
  theme_minimal() 

It is up to the user to avoid this behavior by either removing some words or changing the size scale.

Wordcloud and rotation

Let’s start by creating a rotation angle of 90 for 40 % of the words:

dat$rot <- 90*(runif(nrow(dat))>.6)

We can use this variable in the aesthetic:

ggplot(data = dat, aes(label = name, size = size, angle = rot)) +
  geom_text_wordcloud() +
  scale_size(range = c(2,12)) +
  theme_minimal() 

ggwordcloud allows arbitrary rotations:

dat$rot <- (-90+180*runif(nrow(dat)))*(runif(nrow(dat))>.2)

ggplot(data = dat, aes(label = name, size = size, angle = rot)) +
  geom_text_wordcloud()  +
  scale_size(range = c(2,12)) +
  theme_minimal() 

Wordcloud and eccentricity

The ggwordcloud algorithm moves the text around a spiral until it finds a free place for it. This spiral has by default a vertical eccentricuty of .65, so that the spiral is 1/.65 wider than taller.

ggplot(data = dat, aes(label = name, size = size, angle = rot)) +
  geom_text_wordcloud()  +
  scale_size(range = c(2,12)) +
  theme_minimal() 

This can be changed using the eccentricity parameter:

ggplot(data = dat, aes(label = name, size = size, angle = rot)) +
  geom_text_wordcloud(eccentricity = 1)  +
  scale_size(range = c(2,12)) +
  theme_minimal() 

ggplot(data = dat, aes(label = name, size = size, angle = rot)) +
  geom_text_wordcloud(eccentricity = .3)  +
  scale_size(range = c(2,12)) +
  theme_minimal() 

Advanced wordcloud

geom_text_wordcloud is compatible with the facet system of ggplot2:

ggplot(data = dat, aes(label = name, size = size, angle = rot)) +
  geom_text_wordcloud()  +
  scale_size(range = c(2,12)) +
  facet_wrap(~am) +
  theme_minimal() 

One can also specify an original position for each label that what will be used as the starting point of the spiral algorithm for this label:

ggplot(data = dat, aes(x = factor(am), label = name, size = size, angle = rot,
                       color = factor(am))) +
  geom_text_wordcloud()  +
  scale_size(range = c(2,12)) +
  theme_minimal()