He's a cold hearted snake!
I have been organizing a series of R workshops at work to introduce my colleagues to R, as well as the tidyverse. The most recent workshop focused on working with text data where we looked at the lyrics to Paula Abdul's "Cold Hearted", a song supposedly written about her ex-boyfriend, John Stamos.
Packages
library(tidyverse)
library(geniusr)
library(magick)
Obtain the lyrics
The lyrics can be obtained using the geniusr package and the code below. An API token is not
required for the searching of lyrics. I usually use genius for the scraping of lyrics but lately
I've been using geniusr since genius has been encountering very strange issues with the scraping
of lyrics.
cold <- get_lyrics_search(artist_name="Paula Abdul", song_title="Cold Hearted")
head(cold)
## # A tibble: 6 x 5
## line section_name section_artist song_name artist_name
## <chr> <chr> <chr> <chr> <chr>
## 1 He's a cold-hearted snake Chorus Paula Abdul Cold Hea~ Paula Abdul
## 2 Look into his eyes, uh-oh Chorus Paula Abdul Cold Hea~ Paula Abdul
## 3 He's been tellin' lies Chorus Paula Abdul Cold Hea~ Paula Abdul
## 4 He's a lover boy at play Chorus Paula Abdul Cold Hea~ Paula Abdul
## 5 He don't play by rules, uh-~ Chorus Paula Abdul Cold Hea~ Paula Abdul
## 6 Girl, don't play the fool, ~ Chorus Paula Abdul Cold Hea~ Paula Abdul
Since this data only contains one artist and one song, we can actually just drop all columns except the one containing the lyrics. I'm also going to convert everything to lowercase to make the later regexing a bit easier.
cold <- cold %>%
select(line) %>%
mutate(line = str_to_lower(line))
I'm going to do some additional light cleaning to make the text a bit more formal.
cold <- cold %>%
mutate(
line = str_replace_all(line, pattern="in'", replacement="ing"),
line = str_remove_all(line, pattern="(c-)+"),
line = str_replace_all(line, pattern="snnnnnake", replacement="snake")
)
What is she saying about "him"?
I'm mostly interested in the snippets of the lines that mention he/he's ..., e.g. "he's a cold-hearted snake", "he don't play by rules", etc.
To do:
First we need to get all the lyrics that mention he/he's. Then, we're going to extract the part that comes after he/he's, e.g. "he's a cold-hearted snake" → "a cold-hearted snake".
Before we can visualise it, we're going to need to obtain the counts. In addition:
filter(cold, str_detect(line, pattern="don't play by"))
## # A tibble: 6 x 1
## line
## <chr>
## 1 he don't play by rules, uh-oh
## 2 he don't play by rules, uh-oh
## 3 he don't play by rules, uh-oh
## 4 he don't play by the rules, uh-oh
## 5 ooh, ahh, he don't play by rules
## 6 he don't play by the rules, uh-oh
some of the lyrics will need to be collapsed since we have "don't play by rules, uh-oh", "don't play by the rules, uh-oh", and "don't play by rules". In the initial data, there was also "he's cold as ice" and "he's c-cold as ice", but that was taken care of in the initial cleaning.
Finally, we can visualise the counts. I also want to include a picture of John Stamos' face in the plot, since this song is supposedly about him.
Get the relevant lyrics and extract the descriptions
Technically, we should only remove the first instance of he/he's in a line, e.g. "do you really think he thinks about you when he's out?".
he <- cold %>%
filter(str_detect(line, pattern="\\bhe('s)?\\b")) %>%
mutate(
line = str_extract(line, pattern="\\bhe('s)?\\b.+"),
line = str_remove(line, pattern="\\bhe('s)?\\b"),
line = str_squish(line)
)
he
## # A tibble: 34 x 1
## line
## <chr>
## 1 a cold-hearted snake
## 2 been telling lies
## 3 a lover boy at play
## 4 don't play by rules, uh-oh
## 5 needs it
## 6 off and running with the crowd
## 7 thinks about you when he's out?
## 8 a cold-hearted snake
## 9 been telling lies
## 10 a lover boy at play
## # ... with 24 more rows
Collapse similar lyrics
Since the lyric lines are eventually going to be plotted, I'm going to convert them to sentence case.
he <- he %>%
mutate(
line = if_else(str_detect(line, pattern="don't play by"), "don't play by rules", line),
line = str_to_sentence(line)
)
Get the counts
To avoid overplotting, I'm only going to keep the lines that appeared more than once.
he_counts <- he %>%
count(line) %>%
arrange(desc(n)) %>%
filter(n > 1)
he_counts
## # A tibble: 5 x 2
## line n
## <chr> <int>
## 1 Been telling lies 7
## 2 A cold-hearted snake 6
## 3 Don't play by rules 6
## 4 A lover boy at play 5
## 5 Cold as ice 3
Visualisation
Prep the image of John Stamos
Grabbed his headshot from his Wikipedia page.
I don't want him to be the centre of attention in the plot. So I'm going to fade his picture out a bit to match the plotting background (white).
stamos_face <- image_read("./img/stamos-face.png") %>%
image_colorize(opacity=85, color="#ffffff")
plot(stamos_face)
Sort the lyric lines by their counts
A bar chart of unsorted counts is no fun to read. To ensure that our plot will have bars sorted by the counts, we will need to convert the text to factor and define an order within them.
The lines should be sorted by ascending count, since we'll have text on the y-axis and counts on the x-axis.
he_counts_reordered <- he_counts %>%
mutate(line = fct_reorder(line, n))
Make the plot
Since order matters, I'm first going to initialise the canvas, plot his face, then draw the bars on top.
ggplot(he_counts_reordered, aes(x=n, y=line)) +
annotation_raster(stamos_face, xmin=4.75, xmax=7.5, ymin=-0.5, ymax=2.75) +
geom_col() +
labs(
title="Sketch of an ex-boyfriend",
x="Frequency", y="",
caption="Source: Paula Abdul, \"Cold Hearted\""
) +
theme_minimal()
Conclusions
If your boyfriend has been telling lies, doesn't play by rules, is a cold-hearted snake, a lover boy at play, and/or is cold as ice, you've found yourself the perfect ex-boyfriend.
"Girl, don't play the fool" — Paula Abdul
