In tulliapadellini/rainbaud: Color Palettes from Texts

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "README-",
  fig.align = "center"
)

Rainbaud

An R package for converting text into colors.

Color theory has been a largely explored field of research, in which the relationship between color and other forms of perception stands as an under-defined and still unsolved problem. Many have been addressing the issue of associating colors to other domains of perception. For example Kandinsky (1911) relates colors to moods, while Ramachandran and Hubard investigate the connection between sounds and color in brain attivation patterns (Chromesthesia).

We built on the framework introduced by the work of Rimbaud (1871), who proposes to link colors to text, by associating each vowel to a chromatic shade:

Vowels - Arthur Rimbaud (1871)

A black, E white, I red, U green, O blue: vowels,\ I shall tell, one day, of your mysterious origins:\ A, black velvety jacket of brilliant flies\ which buzz around cruel smells,

Gulfs of shadow; E, whiteness of vapours and of tents,\ lances of proud glaciers, white kings, shivers of cow-parsley;\ I, purples, spat blood, smile of beautiful lips\ in anger or in the raptures of penitence;\

U, waves, divine shudderings of viridian seas,\ the peace of pastures dotted with animals, the peace of the furrows\ which alchemy prints on broad studious foreheads;\

O, sublime Trumpet full of strange piercing sounds,\ silences crossed by [Worlds and by Angels]:\ –O the Omega! the violet ray of [His] Eyes!

The core of our work is to extend such idea by considering words as a whole, rather than just the letters they are composed of. To this aim, we expoit Picular, which allows to retrieve the most prominent colors in Google Images searches for any text query.

The Practice

The package consists of only 3 functions:

get.color, which builds the connection between R and Picular, and for each given word, returns the first 20 results of the corresponding Picular search.

library(rainbaud)
pal1 = get.color("persistence")
plot(1:20, rep(1, 20), col = pal1, xlab = "", ylab = "", pch = 15, cex = 7, axes = FALSE)

By default, colors are sorted according to their hue value. Setting the argument sorted = FALSE however, returns the same colors ordered by their prevalence in the Google Images search result.

pal2 = get.color("persistence", sorted = FALSE)
plot(1:20, rep(1, 20), col = pal2, xlab = "", ylab = "", pch = 15, cex = 7, axes = FALSE)

create.palette, which is the core of the package as it is the function that converts any collection of words into a color-palette of a given size. We show a toy example of how to use to extract a color-palette from another of Rimbaud's poems, The Drunken Boat. The full-text of this poetry is provided in the package, as the data-object drunkenboat. In the create.palette function, the whole poem is summarized into n.colors words, representing the key topic of the composition, by means of Latent Dirichlet Allocation. Each of this word is then converted to colors through Picular. As we have seen before, however, the Picular call returns 20 colors for each word, hence it is necessary to summarize them into one. Here we propose 3 aggregation strategies, which can be selected by specifying the argument aggreg.method of the function. Possible choices are:

• "most-prominent" - for each word, consider the first color returned by Picular (the default option);

data("drunkenboat")
db.palette1 = create.palette(drunkenboat, n.colors = 5, aggreg.method = "most-prominent")
plot(1:5, rep(1, 5), col = db.palette1, xlab = "", ylab = "", pch = 15, cex = 7, axes = FALSE)

• "mean" - for each word, take the mean of the 20 colors retured by Picular in the hsv space.

db.palette2 = create.palette(drunkenboat, n.colors = 5, aggreg.method = "mean")
plot(1:5, rep(1, 5), col = db.palette2, xlab = "", ylab = "", pch = 15, cex = 7, axes = FALSE)

• "median" - for each word, take the median of the 20 colors retured by Picular, sorting them according to their hue value;

db.palette3 = create.palette(drunkenboat, n.colors = 5, aggreg.method = "median")
plot(1:5, rep(1, 5), col = db.palette3, xlab = "", ylab = "", pch = 15, cex = 7, axes = FALSE)

aggreg.colors, which is mostly a utility function, which can be used to sort colors according to their hue value or compute summaries in the hsv-space of colors.

Installation

rainbaud is still not on CRAN, but you can install it from github using the following:

# install.packages("devtools")
devtools::install_github("tulliapadellini/rainbaud")

tulliapadellini/rainbaud documentation built on July 12, 2019, 9:55 p.m.