Vampire Weekend Music & Lyrics
Juliet Goodman
5/7/2020
Background
Vampire Weekend is an indie band that formed in 2006. Since then, they have released four albums. “Vampire Weekend” was released in 2008, then “Contra” in 2010, “Modern Vampires of the City” in 2013, and “Father of the Bride” in 2019. There is a sense to fans that the overall image and tone of the band started very happy and light with the first two albums, got very dark at the third, then became more positive again but not to the point where they began.
This report investigates the hypothesis that the music and lyrics both start positive with the first two albums, become very negative at the third album, and even out with the fourth album.
Lyrics
Using the Genius library to find and break down Vampire Weekend’s discography into lyrics by album
Separating lyrics from lines into words for each album
First Album “Vampire Weekend”
Second Album “Contra”
Third Album “Modern Vampires of the City”
Most Common Words
Vampire Weekend
vw_word_album1 %>%
head(10) ->album1top10
album1top10 %>%
ggplot(aes(reorder(word, -n), n, fill=word)) + geom_col() + scale_fill_tableau() + ggtitle("Top 10 Words of 'Vampire Weekend'") + labs(y= "Number of Appearances", x = "Words") 
Contra
vw_word_album2 %>%
head(10) ->album2top10
album2top10 %>%
ggplot(aes(reorder(word, -n), n, fill=word)) + geom_col() + scale_fill_tableau() + ggtitle("Top 10 Words of 'Contra'") + labs(y= "Number of Mentions", x = "Words") 
Sentiments
The sentiment analysis was accomplished with the Afinn lexicon. Afinn ranks words on a scale of -5 to 5 based on postive or negative connotation. -5 is the most negative rating and 5 is the most positive.
The tables below are the 10 most positive and 10 most negative words from each respective album. If multiple words have the same ranking in the lexicon, they are then ordered in the table based on frequency.
Vampire Weekend
Positive
vw_song_lyrics %>%
filter(title %in% "Vampire Weekend") %>%
unnest_tokens(word, lyric) %>%
anti_join(stop_words) %>%
count(word, sort = TRUE) %>%
inner_join(get_sentiments("afinn")) %>%
arrange(desc(value)) -> vw_word_album_afinn_pos
vw_word_album_afinn_pos %>%
head(10) %>% kable| word | n | value |
|---|---|---|
| funny | 1 | 4 |
| charm | 2 | 3 |
| praise | 2 | 3 |
| perfect | 1 | 3 |
| chance | 3 | 2 |
| fine | 2 | 2 |
| smile | 2 | 2 |
| tops | 2 | 2 |
| true | 2 | 2 |
| cares | 1 | 2 |
There are 18 occurrences of the 10 unique most positive words. The median value is 2 and mean is 2.5.
Negative
vw_song_lyrics %>%
filter(title %in% "Vampire Weekend") %>%
unnest_tokens(word, lyric) %>%
anti_join(stop_words) %>%
count(word, sort = TRUE) %>%
inner_join(get_sentiments("afinn")) %>%
arrange(desc(-value)) -> vw_word_album_afinn_neg
vw_word_album_afinn_neg %>%
head(10) %>% kable| word | n | value |
|---|---|---|
| fuck | 5 | -4 |
| shit | 1 | -4 |
| cruel | 3 | -3 |
| dumb | 2 | -3 |
| evil | 1 | -3 |
| lost | 1 | -3 |
| murdering | 1 | -3 |
| racist | 1 | -3 |
| insane | 6 | -2 |
| collapse | 1 | -2 |
There are 22 occurrences of the 10 unique most negative words. The mean and median values are both -3.
Contra
Positive
vw_song_lyrics %>%
filter(title %in% "Contra") %>%
unnest_tokens(word, lyric) %>%
anti_join(stop_words) %>%
count(word, sort = TRUE) %>%
inner_join(get_sentiments("afinn")) %>%
arrange(desc(value)) -> c_word_album_afinn_pos
c_word_album_afinn_pos %>%
head(10) %>% kable| word | n | value |
|---|---|---|
| funny | 2 | 4 |
| love | 1 | 3 |
| lovely | 1 | 3 |
| chance | 3 | 2 |
| care | 2 | 2 |
| honest | 2 | 2 |
| brave | 1 | 2 |
| enjoy | 1 | 2 |
| fair | 1 | 2 |
| fine | 1 | 2 |
There are 14 occurrences of the 10 unique most positive words. The median and mean values are both 2.
Negative
vw_song_lyrics %>%
filter(title %in% "Contra") %>%
unnest_tokens(word, lyric) %>%
anti_join(stop_words) %>%
count(word, sort = TRUE) %>%
inner_join(get_sentiments("afinn")) %>%
arrange(desc(-value)) -> c_word_album_afinn_neg
c_word_album_afinn_neg %>%
head(10) %>% kable| word | n | value |
|---|---|---|
| cruel | 2 | -3 |
| lost | 2 | -3 |
| worse | 2 | -3 |
| bad | 1 | -3 |
| desperate | 1 | -3 |
| die | 1 | -3 |
| fake | 1 | -3 |
| horrified | 1 | -3 |
| victim | 1 | -3 |
| bitter | 4 | -2 |
There are 16 occurrences of the 10 unique most negative words. The mean median is -3 and the mean is -2.9.
Modern Vampires of the City
Positive
vw_song_lyrics %>%
filter(title %in% "Modern Vampires of the City") %>%
unnest_tokens(word, lyric) %>%
anti_join(stop_words) %>%
count(word, sort = TRUE) %>%
inner_join(get_sentiments("afinn")) %>%
arrange(desc(value)) -> mvotc_word_album_afinn_pos
mvotc_word_album_afinn_pos %>%
head(10) %>% kable| word | n | value |
|---|---|---|
| rejoicing | 1 | 4 |
| love | 20 | 3 |
| praise | 4 | 3 |
| excited | 3 | 3 |
| blessing | 2 | 3 |
| charming | 1 | 3 |
| luck | 1 | 3 |
| pleasant | 1 | 3 |
| won | 1 | 3 |
| stronger | 4 | 2 |
There are 39 occurrences of the 10 unique most positive words. The median and mean values are both 3.
Negative
vw_song_lyrics %>%
filter(title %in% "Modern Vampires of the City") %>%
unnest_tokens(word, lyric) %>%
anti_join(stop_words) %>%
count(word, sort = TRUE) %>%
inner_join(get_sentiments("afinn")) %>%
arrange(desc(-value)) -> mvotc_word_album_afinn_neg
mvotc_word_album_afinn_neg %>%
head(10) %>% kable| word | n | value |
|---|---|---|
| damn | 2 | -4 |
| hell | 1 | -4 |
| die | 8 | -3 |
| idiot | 3 | -3 |
| died | 2 | -3 |
| bad | 1 | -3 |
| hate | 1 | -3 |
| lost | 1 | -3 |
| fire | 14 | -2 |
| fool | 5 | -2 |
There are 38 occurrences of the 10 unique most negative words. The mean and median values are both -3.
Father of the Bride
Positive
vw_song_lyrics %>%
filter(title %in% "Father of the Bride") %>%
unnest_tokens(word, lyric) %>%
anti_join(stop_words) %>%
count(word, sort = TRUE) %>%
inner_join(get_sentiments("afinn")) %>%
arrange(desc(value)) -> fotb_word_album_afinn_pos
fotb_word_album_afinn_pos %>%
head(10) %>% kable| word | n | value |
|---|---|---|
| win | 5 | 4 |
| funny | 1 | 4 |
| triumph | 1 | 4 |
| love | 7 | 3 |
| affection | 4 | 3 |
| grand | 1 | 3 |
| loved | 1 | 3 |
| perfect | 1 | 3 |
| sympathy | 6 | 2 |
| proud | 4 | 2 |
There are 31 occurrences of the 10 unique most positive words. The median value is 3 and mean is 3.1.
Negative
vw_song_lyrics %>%
filter(title %in% "Father of the Bride") %>%
unnest_tokens(word, lyric) %>%
anti_join(stop_words) %>%
count(word, sort = TRUE) %>%
inner_join(get_sentiments("afinn")) %>%
arrange(desc(-value)) -> fotb_word_album_afinn_neg
fotb_word_album_afinn_neg %>%
head(10) %>% kable| word | n | value |
|---|---|---|
| die | 8 | -3 |
| worried | 4 | -3 |
| violence | 3 | -3 |
| anger | 2 | -3 |
| cruel | 2 | -3 |
| evil | 2 | -3 |
| fake | 2 | -3 |
| hate | 2 | -3 |
| kill | 2 | -3 |
| lost | 2 | -3 |
There are 29 occurrences of the 10 unique most negative words. The mean and median values are both -3.
Lyrics Conclusion
Frequency
Most of the top 10 most common words among the four albums are filler like “ooh” and “la” or simply have no connatation either way.
Sentiment Analysis
Of the 10 most positive words that appear in each individual album, the number of occurrences does not follow a clear negatively or positively sloped line. A line of best fit would be positively sloped as they go from 18 to 15 to 39 to 31. Meaning, the most postive words are used more often as time goes on generally.
As for negative words, the total occurrences go from 22 to 16 to 38 to 29. This follows the exact same pattern. It appears that language with strong connotations (be it positive or negative) tend to appear with each album.
The absolute values of the means and medians for all eight tables are very similar to each other as well. For each album, the absolute values of the medians and means of the negative rankings are greater than or equal to their positive counterparts in every instance but one. The exception is the last album. The absolute values for both medians are 3 and the absolute value for the negative mean is 3 and the absolute value for the positive mean is 3.1. Close, but not equal.
Because the data showed the same patterns for the rise of positive and negative words and overall similar outcomes for measures of central tendencies, the hypothesis has been disproved. While the lyrics have become more negative, they have also become more positive with time.
Music
Accessing the Spotify API and Vampire Weekend’s discography and music
Valence
The SpotifyR package comes with a mix of standard metrics and API-specific metrics. The Spotify-specific measurement, valence, ascribes a number from 0.0-1.0 to a track based on “musical positivness” (per the API’s documentation). The closer to 1.0, the higher the valence and the more positive it is.
Five songs with the highest valence
Querying top ten because many have duplicates.
| track_name | valence |
|---|---|
| Oxford Comma | 0.974 |
| Oxford Comma | 0.973 |
| M79 | 0.948 |
| White Sky | 0.944 |
| M79 | 0.940 |
| Sunflower (feat. Steve Lacy) | 0.933 |
| Sunflower (feat. Steve Lacy) | 0.932 |
| White Sky | 0.908 |
| The Kids Don’t Stand A Chance | 0.906 |
| Holiday | 0.893 |
1. “Oxford Comma” (Vampire Weekend)
2. “M79” (Vampire Weekend)
3. “White Sky” (Contra)
4. “Sunflower” (Father of the Bride)
5. “The Kids Don’t Stand A Chance” (Vampire Weekend)
The third album is the only one not represented in that list.
Mean and median valences across albums
The results are displayed from highest to lowest number, by release chronology.
vampireweekend %>%
group_by(album_name) %>%
summarise(mean(valence)) %>%
arrange(desc(`mean(valence)`)) %>%
kable | album_name | mean(valence) |
|---|---|
| Contra | 0.7397500 |
| Vampire Weekend | 0.7362727 |
| Father of the Bride | 0.5201026 |
| Modern Vampires of the City | 0.4884615 |
vampireweekend %>%
group_by(album_name) %>%
summarise(median(valence)) %>%
arrange(desc(`median(valence)`)) %>%
kable | album_name | median(valence) |
|---|---|
| Contra | 0.8130 |
| Vampire Weekend | 0.7795 |
| Father of the Bride | 0.5250 |
| Modern Vampires of the City | 0.4910 |
Density plot
Spotify assigned each track a valence value. These density plots shows which valence values the tracks cluster at for each album.
Individual density plots
vampireweekend %>%
group_by(album_name) %>%
ggplot(aes(x = valence, y = album_name, fill = ..x..)) +
geom_density_ridges_gradient() +
xlim(0,1) +
theme(legend.position = "none") +
ggtitle("Density Plot of Valences For Each Album") +
labs(y= "Album Name", x = "Valence") 
All the albums stacked on top
vampireweekend %>% group_by(album_name) %>%
ggplot(aes(x = valence, fill = album_name)) +
geom_density(alpha=.4, color=NA) +
xlim(0,1) +
ggtitle("Density Plot of Valences Across Albums") +
labs(y= "Density", x = "Valence", fill = "Album Name") 
“Contra” has a small local maximum between 0 and 0.25, and a much higher maximum between 0.75 and 1.0. There is a valley that starts just past 0.25 and picks up close to 0.5. This means that a few songs from that album have low valence values but most have high values and almost none are in the middle. “Vampire Weekend” has a similar shape to Contra, but neither peak is as high, leading them to both be wider.
On the other end of the spectrum is “Modern Vampires of the City.” The plateau indicates that every measurement (except the highest) is met.
“Father of the Bride” has the opposite shape. Most of its tracks are around 0.50, few are low nor high.
Music Conclusion
Mean and Median
The mean and medians show that the hypothesis is correct. The first two albums for both are very high, the third album’s mean and median are much lower, and the fourth album’s values are in the middle.
Density Plot
The first two albums both cluster at high valences. The third album does not truly peak at all. The last album’s follow’s the same pattern here, of slightly peaking about halfway through.
Connecting Music and Lyrics
Based on the conclusion from the lyrics, about the use of polarized language in both directions increasing over time, the music does not follow the same pattern. If it did, the valence density plots would have started in the center then later albums would have two peaks: one on each side of the chart of similar sizes. In turned out to be the opposite.