Wordcloud test

cloud.R

library(tidyverse)
library(ggwordcloud)
library(tidytext)

# https://github.com/lepennec/ggwordcloud
# https://cran.r-project.org/web/packages/tidytext/vignettes/tidytext.html

# get some text, remove newline/return chars
text <- read_file("https://www.gutenberg.org/cache/epub/84/pg84.txt") |> 
  str_replace_all(pattern = "\r\n|\n|\r", replacement = " ") |> 
  str_squish()

# put in dataframe, one row per word
df <- tibble(text = text) |> 
  unnest_tokens(output = word, input = text)

# just combine all the en language stopword dictionaries
stopwords <- get_stopwords("en", "snowball") |> 
  bind_rows(get_stopwords("en", "smart")) |>
  bind_rows(get_stopwords("en", "stopwords-iso")) |> 
  bind_rows(get_stopwords("en", "marimo")) |> 
  bind_rows(get_stopwords("en", "nltk")) |> 
  bind_rows(tibble(word = c("my", "custom", "stopwords"), lexicon = "custom")) |> 
  group_by(word) |> summarise(lexicon = first(lexicon), .groups = "drop")

# remove stopwords and count remaining word freqs
df_cleaned_word_freqs <- df |> 
  anti_join(stopwords) |> 
  count(word) |> 
  arrange(desc(n)) 

# plot n most common words in text
n <- 30
max_size <- 15 # size limit of plot, may need tweaking

df_cleaned_word_freqs |> 
  slice_head(n = n) |> 
  ggplot(aes(label = word, size = n, colour = n)) +
  geom_text_wordcloud() +
  scale_size_area(max_size = max_size) +
  scale_colour_viridis_c(option = "magma", direction = -1, end = 0.9) +
  theme_minimal()