lyrical-masterminds.R

library(rio)
library(tidyverse)
library(spotifyr)
library(fuzzyjoin)
library(ggrepel)

#lets get the rapper's vocabulary count data
rappers <- import("https://docs.google.com/spreadsheets/d/1HIIfgDpNMM-j0hoQHN-yP5P1lNOfJuvym0u0sdWwD9g/edit#gid=737896402") %>%
  filter(!is.na(recalc)) %>% #rm those that we don't have word count data for
  as_tibble() %>%
  mutate(rapper_clean = recode(rapper_clean, "BoB" = "B.o.B", # or we'll get Bob Marley from spotify
                               "Big KRIT" = "Big K.R.I.T.",
                               "Brockhampton"  = "BROCKHAMPTON",
                               "Del the Funky Homosapian" = "Del the Funky Homosapien",
                               "Jay-Z" = "JAY Z",
                               "KAAN" = "K.A.A.N.",
                               "Missy Elliot" = "Missy Elliott",
                               "Puff Daddy" = "Diddy",
                               "Young Jeezy" = "Jeezy"
                               ))

#cool lets see what data we can get from Spotify, first we need to set up a developer
#id to get access to the API see https://www.rcharlie.com/spotifyr/ for guide

Sys.setenv(SPOTIFY_CLIENT_ID = 'insert_your_client_id_here')
Sys.setenv(SPOTIFY_CLIENT_SECRET = 'insert_your_client_secret_here')

access_token <- spotifyr::get_spotify_access_token(client_id = Sys.getenv("SPOTIFY_CLIENT_ID"),
                                                   client_secret = Sys.getenv("SPOTIFY_CLIENT_SECRET"))

# lets get a vector with the rapper names
artists <- rappers$rapper_clean

# ok lets write a loop that fetches the data for all rappers in artists
df <- NULL

for (i in 1:length(artists)) {
  tryCatch({
    m <-  spotifyr::get_artists(artists[i], return_closest_artist = TRUE)
    df <- rbind(df, data.frame(m))
  }, error=function(e){})
}

# "Murs" need special attention or we get "Olly Murs", and we don't want that!
df <- spotifyr::get_artists("Murs") %>%
  filter(artist_name=="Murs") %>%
  rbind(df)

# same goes for Rakim or we'll get RKM & Ken-Y
df <- spotifyr::get_artists("Rakim") %>%
  filter(artist_name=="Rakim") %>%
  rbind(df)

# lets join with fuzzy join's string dist, because there are so many ways to spell these names...
df <- rappers %>%
  fuzzyjoin::stringdist_left_join(df, by = c("rapper_clean"="artist_name")) %>%
  select(rapper, rapper_clean, artist_name, era, words, recalc, artist_popularity, artist_num_followers)

#cool but we have ten too many, i.e. there are duplicates introduced
df %>% filter(duplicated(rapper_clean)) %>% select(rapper_clean, artist_name)

# ironically these all would have worked perfectly with a normal left_join()
# lets remove these, filter out those that are incorrect and merge back again
df_duplicates <- df %>%
  filter(rapper_clean == "GZA" | rapper_clean == "Murs" | rapper_clean == "Nas" |
           rapper_clean == "NF" | rapper_clean == "Russ" | rapper_clean == "RZA" |
           rapper_clean == "YG" | rapper_clean == "UGK") %>%
  filter(rapper_clean == artist_name)

df <- df %>%
  filter(!(rapper_clean == "GZA" | rapper_clean == "Murs" | rapper_clean == "Nas" |
           rapper_clean == "NF" | rapper_clean == "Russ" | rapper_clean == "RZA" |
           rapper_clean == "YG" | rapper_clean == "UGK")) %>%
  rbind(df_duplicates)

# perhaps not so smooth but at least its right
cols <- viridis::viridis(1, begin = .9)

label_these <- c("Wu-Tang Clan","Kendrick Lamar", "Jedi Mind Tricks", "Immortal Technique", "Eminem",
                 "2Pac", "Nas", "Beastie Boys", "Busdriver", "Aesop Rock", "Kanye West", "Run-D.M.C.",
                 "NF", "Diddy", "Childish Gambino", "Snoop Dogg", "Mos Def",
                 "GZA", "A Tribe Called Quest", "Wiz Khalifa", "A$AP Rocky", "Common", "Cypress Hill", "DMX", "Ice T",
                 "Killah Priest", "Kool G Rap", "KRS-One", "Lil Uzi Vert",  "Method Man", "MF DOOM",
                 "Outkast", "Travis Scott", "Too Short", "K-Rino", "Ice Cube", "Canibus", "Lil Wayne",
                 "Rakim", "Missy Elliott", "The Roots", "21 Savage", "Busta Rhymes", "J Cole", "Logic")

df <- df %>%
  mutate(high_light = ifelse(rapper_clean == "Drake", "Drake is the most popular artist on Spotify",
                             ifelse(rapper_clean %in% label_these, rapper_clean, "")))

ggplot(df, aes(recalc, artist_popularity)) +
  geom_smooth(size = 1.3, color = cols, se = FALSE, linetype = 2, method = "lm") +
  geom_point(aes(color = era), size = 4) +
  theme_classic(base_size = 14) +
  scale_color_viridis_d(end = .8) +
  ylab("Popularity on Spotify") +
  xlab("Lyrical largesse") +
  labs(title = "How are the lyrical masterminds doing on Spotify?",
       subtitle = "Code at: https://gist.github.com/tannenberg",
       caption = "Source: Vocabulary size from @matthew_daniels; popularity from Spotify's API") +
  guides(color=guide_legend(title="Era")) +
  geom_text_repel(aes(label = high_light), size = 4,  nudge_y = 1.5, nudge_x = 17) +
  NULL