word tree

R code of the ggplot2 example

library(ggplot2)
library(dplyr)
library(tidyr)
library(tidytext)
library(igraph)
library(ggraph)
library(HistData)

## What follows "the power of" in the 85 essays: the eight most frequent next
## words, then, for each, the next word when it occurs at least twice.
phrases <- Federalist_text |>
  unnest_tokens(phrase, text, token = "ngrams", n = 5) |>
  filter(startsWith(phrase, "the power of ")) |>
  separate(phrase, c(NA, NA, NA, "w1", "w2"), sep = " ")
first <- phrases |>
  count(w1, sort = TRUE) |>
  slice_head(n = 8)
second <- phrases |>
  filter(w1 %in% first$w1) |>
  count(w1, w2) |>
  filter(n >= 2) |>
  group_by(w1) |>
  slice_max(n, n = 1, with_ties = FALSE) |>
  ungroup()
edges <- bind_rows(tibble(from = "the power of", to = first$w1),
                   tibble(from = second$w1, to = paste(second$w1, second$w2)))
nodes <- bind_rows(tibble(name = "the power of", label = "the power of", n = sum(first$n)),
                   tibble(name = first$w1, label = first$w1, n = first$n),
                   tibble(name = paste(second$w1, second$w2), label = second$w2, n = second$n))
tree <- graph_from_data_frame(edges, vertices = nodes)

p <- ggraph(tree, layout = "tree") +
  geom_edge_diagonal(colour = "#a6a6a6", width = 0.3) +
  geom_node_label(aes(label = label, size = n), family = "Quicksand", colour = "#9b2367",
                  label.size = 0, fill = "white") +
  scale_size(range = c(2.4, 5)) +
  coord_flip() +
  scale_y_reverse(expand = expansion(mult = c(0.45, 0.2))) +
  labs(title = "“The power of the” and “the power of making” lead",
       subtitle = "Words that follow “the power of” in The Federalist, size by count",
       caption = "Data: The Federalist Papers, via the HistData R package") +
  theme_lex_viz(grid = "", arrows = "") +
  theme(legend.position = "none", axis.text = element_blank(), axis.title = element_blank())