Which letter pairs that are next to each other in the alphabet actually appear together in the English language?

fake_letter_friends.r

library(janeaustenr)
library(tidytext)
library(tidyverse)

theme_set(theme_bw())

cleaned_words <- austen_books() |>
    unnest_tokens(word, text) |>
    # Remove roman numerals
    filter(!str_detect(word, "[IVXLCDM]+")) |>
    select(-book)

# Get letters
cleaned_letters <- cleaned_words |>
    mutate(word_id = row_number()) |>
    unnest_tokens(letter, word, token = "characters", drop = FALSE) |>
    mutate(letter = toupper(letter)) |>
    filter(letter %in% LETTERS) |>
    select(-word)


# Count letters and pairs
letter_counts <- cleaned_letters |>
    count(letter) |>
    mutate(f = n / sum(n))

independence_frequencies <- letter_counts |>
    mutate(dummy = 0) |>
    (\(df) full_join(df, df, by = "dummy", suffix = c("_1", "_2")))() |>
    mutate(
        pair = str_c(letter_1, letter_2),
        f = f_1 * f_2
    ) |>
    select(pair, f)

pair_counts <- cleaned_letters |>
    group_by(word_id) |>
    mutate(letter_2 = lead(letter)) |>
    ungroup() |>
    filter(!is.na(letter_2)) |>
    select(-word_id) |>
    rename(letter_1 = letter) |>
    complete(letter_1, letter_2, fill = list(n = 0)) |>
    mutate(pair = str_c(letter_1, letter_2)) |>
    count(letter_1, letter_2, pair) |>
    left_join(rename(independence_frequencies, exp_f = f), by = "pair") |>
    mutate(
        f = n / sum(n),
        rel_f = f / exp_f
    )

letter_friends <- str_c(LETTERS, lead(LETTERS))[-26]

# Plot absolute friend frequencies
pair_counts |>
    filter(pair %in% letter_friends) |>
    gather(f, rel_f, key = "type", value = "f") |>
    mutate(type = case_when(
        type == "f" ~ "Absolute Frequences",
        type == "rel_f" ~ "Relative Frequencies*"
    )) |>
    ggplot(aes(pair, f)) +
        geom_col(aes(fill = type), show.legend = FALSE) +
        facet_wrap(~type, ncol = 1, scales = "free_y") +
        labs(
            x = "Letter Pair ('friends' only)",
            y = "Frequency (out of all pairs)",
            title = "Fake Letter Friends",
            subtitle = str_c(
                "Which letters pairs that are next to each other in the",
                "alphabet actually appear together in the English language?",
                sep = " "
            ),
            caption = str_c(
                "Data Source: `janeaustenr`",
                "*Relative to expected frequencies if letters were independent",
                sep = "\n"
            )
        ) +
        scale_x_discrete(expand = c(0, 1)) +
        scale_y_continuous(expand = c(0, 0, 0.1, 0)) +
        theme(
            plot.title = element_text(face = "bold"),
            plot.subtitle = element_text(size = 9)
        )
ggsave("fake_letter_friends.jpg")