The circlize package allows to create chord diagrams with R. It accepts two types of input formats: either a matrix showing the strength of connections between pairs of nodes, or a data frame listing the relationships.
The chordDiagram() function from circlize creates the
visualization with entities arranged in a circle, connected by ribbons
whose width represents the strength of their relationship. The function
offers extensive customization options for colors, transparency, and
other visual properties. Find a full documentation here.
Note: The circlize package is a project by Zuguang Gu. Here, we will use a matrix where each cell represents how many times one character mentions another character in The Office. This example comes from Ansgar Wolsing.
Data Preparation:
Processing Character Mentions from The Office
# Load required packages
library(tidyverse) # Data manipulation and visualization
library(ggtext) # Text formatting in ggplot2
library(tidytext) # Text mining tools
library(here) # File path management
library(circlize) # Circular visualization
library(schrute) # The Office dataset
# Set base path for output files
base_path <- here("faceted-chord-diagram")
# Clean and standardize character names in the dataset
theoffice <- theoffice %>%
mutate(
character = str_remove_all(character, "[\":]"), # Remove quotes and colons
character = str_remove_all(character, "\\\""), # Remove escaped quotes
character = case_match(
character,
"(Pam's mom) Heleen" ~ "Pam's mom",
"AJ" ~ "A.J.",
"abe" ~ "Gabe",
.default = character
))
# Get top 10 characters by number of lines
top_character_names <- theoffice %>%
count(character, sort = TRUE) %>%
slice_max(order_by = n, n = 10) %>%
arrange(-n) %>%
pull(character)
# Create dataset of character mentions
theoffice_mentions <-
theoffice %>%
filter(character %in% top_character_names) %>% # Keep only top 10 characters
unnest_tokens(word, text, token = "words", to_lower = FALSE) %>% # Split text into words
filter(word %in% top_character_names) %>% # Keep only mentions of top characters
# Convert characters to ordered factors
mutate(
character = factor(character, levels = top_character_names),
character_mentioned = factor(word, levels = top_character_names)) %>%
select(index, season, episode, character, character_mentioned)
# Aggregate mentions and create complete matrix
theoffice_mentions_agg <- theoffice_mentions %>%
group_by(character, character_mentioned) %>%
summarize(
total_mentions = n(), # Count total mentions
mentions_lines = n_distinct(index), # Count unique lines with mentions
.groups = "drop"
) %>%
# Ensure all character combinations exist (including zeros)
complete(character, character_mentioned, fill = list(total_mentions = 0, mentions_lines = 0)) %>%
arrange(character, character_mentioned)Creating the Basic Chord Diagram
This code creates our first chord diagram showing character
interactions in The Office. We first convert our aggregated data into a
matrix format required by the chordDiagram() function, then
set up a custom color palette. The diagram is rendered using
ragg for high-quality output, with carefully chosen visual
parameters like transparency and border width to ensure readability.
# Create a square matrix from mentions data, with character names as row and column names
mat <- matrix(theoffice_mentions_agg$total_mentions, ncol = length(top_character_names))
rownames(mat) <- top_character_names
colnames(mat) <- top_character_names
# Define color palette for characters and background
pal_office <-
c("#FBA93A", "#cc2d36", "#93BFE5", "#D8ACD8", "#f0813c", "#F0F4EC", "#2AA3A6",
"#4BD9EF", "#4bad6d", "#5D77AA")
bg_color <- colorspace::darken("#435774", 0.2)
# Create and save the chord diagram
ragg::agg_png(here(base_path, "17-network-the-office-mentions.png"),
res = 500, width = 8, height = 8, units = "in", bg = bg_color)
par(
family = "Outfit", cex = 2, col = "white", # font family, size, color
bg = bg_color,
mai = rep(0.5, 4) # plot margin in inches
)
chordDiagram(mat, transparency = 0.3,
grid.col = pal_office[1:10], # assign colors to each character
link.border = "white", link.lwd = 0.2, # add thin white borders to connections
annotationTrack = c("name", "grid"), # show only name and grid tracks (excluding axis)
annotationTrackHeight = mm_h(c(3, 5)),
link.largest.ontop = TRUE
)
title(
main = "Who speaks about whom?",
sub = "top 10 characters in The Office\nSource: {schrute} R package. Visualisation: Ansgar Wolsing",
col.main = "white", cex.main = 1.3)
invisible(dev.off())
# Display the saved plot in R Markdown
knitr::include_graphics("https://github.com/holtzy/R-graph-gallery/blob/master/character-interaction-analysis/17-network-the-office-mentions.png?raw=true")
Creating Individual Character-Focused Diagrams
Here we create a custom function to generate individual chord diagrams, each highlighting a specific character’s interactions. The function applies unique coloring to emphasize the focal character while de-emphasizing others. We then use this function in a loop to create a 3x3 facet of diagrams, one for each of the top 9 characters, offering a detailed view of each character’s mention patterns.
# Define function to create individual chord diagrams for each character
plot_chordDiagram <- function(name, rank, color = NULL) {
# Set up colors for current character's diagram
row_colors <- rep("#EEEEEE33", 10) # Default transparent gray for all rows
stopifnot(rank <= length(pal_office)) # Ensure rank is valid
row_colors[rank] <- pal_office[rank] # Highlight current character's row
# Set up grid colors
grid_colors <- alpha(pal_office, 0.2) # Set base transparency for all segments
stopifnot(rank <= length(pal_office)) # Ensure rank is valid
grid_colors[rank] <- pal_office[rank] # Highlight current character's segment
# Create chord diagram
chordDiagram(mat,
grid.col = grid_colors, # Assign segment colors
# Optional: Add thin white borders around connections
# link.border = "white",
# link.lwd = 0.1,
annotationTrack = c("name", "grid"), # show only name and grid tracks (excluding axis)
annotationTrackHeight = mm_h(c(2, 2)),
link.largest.ontop = TRUE,
row.col = row_colors # Assign row colors
)
title(main = name, col.main = "white", cex.main = 1.3) # Add character name as title
}
# Set up output file for faceted plot
filepath_facets_chart <- here(base_path, "17-network-the-office-mentions-facets.png")
ragg::agg_png(filepath_facets_chart,
res = 500, width = 10, height = 10, units = "in", bg = bg_color)
# Set up 3x3 plot grid with formatting parameters
par(
mfrow = c(3, 3), # Arrange plots in 3x3 grid
family = "Outfit", cex = 0.8, col = "white", # Font family, size, color
bg = bg_color,
mai = rep(0.5, 4) # Set uniform margins
)
# Create individual chord diagrams for first 9 characters
for (i in 1:9) {
plot_chordDiagram(top_character_names[i], i)
}
# Save the plot
invisible(dev.off())
# Display the saved plot in R Markdown
knitr::include_graphics("https://github.com/holtzy/R-graph-gallery/blob/master/character-interaction-analysis/17-network-the-office-mentions-facets.png?raw=true")
Adding Title and Annotations to the Final Visualization
This final code block uses the magick package to add professional finishing touches to our visualization. We extend the canvas to accommodate text elements, then add a main title, an explanatory subtitle, and source information. The annotations are carefully positioned and styled using appropriate fonts and sizes to create a polished graphic.
# Load magick package for image manipulation
library(magick)
# Read the generated chart
img_chart <- image_read(filepath_facets_chart)
img_chart_info <- image_info(img_chart)
# Calculate new dimensions to add space for titles (11% taller)
img_new_geometry <- sprintf("%sx%s", img_chart_info$height, 1.11 * img_chart_info$height)
# Add text elements to the chart
img_chart %>%
# Extend canvas vertically while maintaining aspect ratio
image_extent(img_new_geometry, gravity = "south", color = bg_color) %>%
# Add main title at the top
image_annotate("Who speaks about whom in The Office",
gravity = "north", size = 200, color = "white",
font = "American Typewriter") %>%
# Add explanatory subtitle below main title
image_annotate("The width of the connecting lines indicates how often a character
mentions another character's name",
location = "+0+250", # Position 250 pixels from top
gravity = "north", size = 100, color = "white",
font = "Outfit") %>%
# Add source attribution at bottom
image_annotate("Source: {schrute} R package. Visualisation: Ansgar Wolsing",
location = "+0+50", # Position 50 pixels from bottom
gravity = "south", size = 80, color = "white", font = "Outfit") %>%
# Save final image
image_write(here(base_path, "17-network-who-speaks-office.png"))
# Display the final image in R Markdown
knitr::include_graphics("https://github.com/holtzy/R-graph-gallery/blob/master/character-interaction-analysis/17-network-who-speaks-office.png?raw=true")
