Taylor Swift Profanity Rate

tay_distribution.R

## Imports ----
library(tidyverse)
library(genius)
library(here)
library(fs)
data("stop_words")
source("https://raw.githubusercontent.com/mkiang/airline_testing_strategies/master/code/mk_nytimes.R")

dir_create(here("data"))

## Set up a skeleton df ----
taytaframe <- tribble(
    ~ artist,      ~ album,          ~ release_date,
    "Taylor Swift", "Taylor Swift",  "2006-10-24",
    "Taylor Swift", "Fearless",      "2008-11-11",
    "Taylor Swift", "Speak Now",     "2010-10-25",
    "Taylor Swift", "Red",           "2012-10-22",
    "Taylor Swift", "1989",          "2014-10-27",
    "Taylor Swift", "Reputation",    "2017-11-10",
    "Taylor Swift", "Lover",         "2019-08-23",
    "Taylor Swift", "Folklore",      "2020-07-24",
    "Taylor Swift", "Evermore",      "2020-12-11" 
)
write_csv(taytaframe, here("data", "disco_info.csv"))

## Add lyrics ----
if (!file_exists(here("data", "working_data.RDS"))) {
    swiftframe <- taytaframe %>%
        select(-release_date) %>% 
        add_genius(artist, album, "album")
    
    ## Save them ----
    write_csv(swiftframe, here("data", "lyrics_by_line.csv"))
    saveRDS(list(taytaframe = taytaframe,
                 swiftframe = swiftframe),
            here("data", "working_data.RDS"))
}

swiftframe <- readRDS(here("data", "working_data.RDS"))$swiftframe
taytaframe <- readRDS(here("data", "working_data.RDS"))$taytaframe %>%
    mutate(release_date = lubridate::ymd(release_date),
           album_cat = factor(album, ordered = TRUE),
           label = sprintf("%s (%s)", album, format(release_date, "%b %Y"))) %>% 
    mutate(label = case_when(album == "Evermore" ~ paste0("\n\n", label), 
                             album == "Folklore" ~ paste0("\n", label), 
                             TRUE ~ label))

words <- swiftframe %>% 
    unnest_tokens(word, lyric) %>% 
    mutate(stopword = (word %in% stop_words) + 0,
           shit = grepl("\\<shit", word) + 0,
           fuck = grepl("\\<fuck", word) + 0,
           damn = grepl("damn", word) + 0,
           crazy = grepl("\\<crazy", word) + 0)

curses <- words %>%
    filter(stopword == 0) %>%
    group_by(artist, album) %>%
    summarize(
        n_words = n(),
        shit = sum(shit),
        fuck = sum(fuck),
        damn = sum(damn),
        crazy = sum(crazy)
    ) %>%
    gather(curse, n_curse, shit:crazy) %>%
    ungroup() %>%
    left_join(taytaframe) %>%
    mutate(
        rate_per_5k = n_curse / n_words * 5000,
        curse_cat = factor(
            curse,
            levels = c("crazy", "damn", "fuck", "shit"),
            labels = c("crazy", "[god]damn", "fuck[ed]", "shit"),
            ordered = TRUE
        )
    ) %>% 
    arrange(album_cat, curse_cat)

p1 <- ggplot(curses,
       aes(x = release_date,
           y = rate_per_5k,
           color = curse_cat,
           group = curse_cat)) + 
    geom_line(size = 1) + 
    geom_point(color = "white",
               size = 3.5) + 
    geom_point(size = 2) + 
    mk_nytimes() + 
    scale_color_brewer(NULL, palette = "Set1") + 
    scale_y_continuous("Profanity rate (curse word per 5,000 words") + 
    scale_x_date(NULL,
                 breaks = taytaframe$release_date,
                 labels = taytaframe$label) + 
    labs(title = "Taylor Swift's Profanity Rate, 2006—2020")
ggsave("./shits_over_time.jpg", p1, width = 6, height = 3.5, scale = 2)