joni_vader_sentiment.R

library(tidyverse)
library(tidytext)
library(vader)
library(glue)

theme_alex <- function() {
  font <- "Arial"
  theme_minimal()
  theme(
    strip.text = element_text(
      hjust = 0,
      color = "#460069",
      size = 12
    ),
    axis.ticks = element_blank(),
    plot.title = element_text(
      family = font,
      size = 20,
      face = "bold",
      color = "#460069"
    ),
    plot.subtitle = element_text(
      family = font,
      size = 14,
      color = "#6a1c91",
      hjust = 0.5
    ),
    plot.caption = element_text(
      family = font,
      size = 9,
      hjust = 1,
      color = "#460069"
    ),
    axis.title = element_text(
      family = font,
      size = 10,
      color = "#460069"
    ),
    axis.text = element_text(
      family = font,
      size = 9,
      color = "#460069"
    ),
    axis.text.x = element_text(
      margin = margin(5, b = 10)
    ),
    legend.text.align = 0,
    legend.background = element_blank(),
    legend.title = element_blank(),
    legend.key = element_blank(),
    legend.text = element_text(
      family = font,
      size = 18,
      color = "#4B636E"
    )
  )
}

joni_spotify <- readRDS(url("https://github.com/farach/data/blob/master/joni_spotify.rds?raw=true", "rb"))
joni_genius_df <- readRDS(url("https://github.com/farach/data/blob/master/joni_genius_df.rds?raw=true", "rb"))

joni_genius_clean <- joni_genius_df %>%
  filter(
    !album_name %in% c(
      "Shadows and Light", "Travelogue", "Both Sides Now"
    ),
    !is.na(lyric)
  ) %>%
  group_by(album_name, track_n, artist, track_title) %>%
  summarise(lyric = glue_collapse(lyric, sep = " ")) %>%
  ungroup() %>%
  mutate(
    track_title = tolower(
      str_trim(str_remove(track_title, "\\s*\\([^\\)]+\\)\\s*$"), side = "both")
    )
  )

joni_spotify_clean <- joni_spotify %>%
  mutate(
    track_title = tolower(
      str_trim(str_remove(track_name, "\\s*\\([^\\)]+\\)\\s*$"), side = "both")
    ),
    track_title = str_remove(track_title, " - live")
  ) %>%
  filter(
    !album_name %in% c(
      "Shadows and Light", "Travelogue", "Both Sides Now",
      "Shine [Standard Jewel - Parts Order Only]"
    ),
    !track_title %in% c(
      "Coin In The Pocket (Rap)", "Funeral (Rap)",
      "Happy Birthday 1975 (Rap)", "I's A Muggin' (Rap)",
      "Lucky (Rap)"
    )
  )

joni_full_clean <- joni_spotify_clean %>%
  right_join(joni_genius_clean %>% select(-album_name)) %>%
  select(album_name, track_title, lyric, track_number, valence, 
         album_release_year) %>%
  distinct() %>%
  filter(!is.na(album_name))

joni_vader <- joni_full_clean %>%
  mutate(
    vader = map(lyric, get_vader),
    track_title = str_trim(str_remove(track_title, "\\s*\\([^\\)]+\\)\\s*$"), 
                           side = "both")
  ) %>%
  unnest_wider(vader)

joni_p <- joni_vader %>%
  mutate(
    track_number = as.numeric(track_number),
    compound = as.numeric(compound),
    track_title = str_to_title(glue("{track_number} - {track_title}")),
    track_title =
      reorder_within(track_title, -track_number, within = album_name),
    album_name = paste(album_release_year, "-", album_name)
  ) %>%
  distinct(track_title, compound, valence, album_name)

joni_p_final <- joni_p %>%
  mutate(
    valence_scaled = scales::rescale(valence, to = c(-1, 1)),
    valence_1 = "Spotify valence score",
    compound_1 = "VADER compound score"
  ) %>%
  group_by(track_title) %>%
  add_count() %>%
  filter(valence == max(valence)) %>%
  ungroup() %>%
  ggplot(aes(x = track_title, y = compound)) +
  geom_segment(aes(
    x = track_title,
    xend = track_title,
    y = valence_scaled,
    yend = compound
  ),
  linetype = "dashed",
  alpha = 0.75
  ) +
  geom_point(aes(color = compound_1), size = 2) +
  geom_point(aes(y = valence_scaled, color = valence_1), size = 2) +
  facet_wrap(~album_name, scales = "free_y") +
  theme_alex() +
  theme(
    legend.position = "bottom",
    panel.grid.minor = element_blank(),
    panel.grid.major.x = element_line(
      color = "#cbcbcb"
    ),
    panel.grid.major.y = element_blank(),
    panel.background = element_blank(),
    plot.title = element_text(hjust = 0.5)
  ) +
  scale_x_reordered() +
  scale_color_manual(values = c("steelblue", "red")) +
  coord_flip() +
  labs(
    x = NULL,
    y = "Normalized valence and vader score",
    title = "VADER sentiment vs. Spotify valence score of Joni Mitchell songs by album",
    subtitle = "Valence describes the musical positiveness conveyed by a track. Tracks with high valence sound more\npositive (happy, cheerful, euphoric), while tracks with low valence sound more negative (sad, depressed, angry).\nVADER (Valence Aware Dictionary and sEntiment Reasoner) is a lexicon and rule-based sentiment analysis\ntool that is specifically attuned to sentiments expressed in social media, and works well on texts from other domains.",
    caption = "Source:\nGenius.com\nSpotify.com"
  ) +
  guides(color = guide_legend(override.aes = list(size = 5)))

ggsave("./joni_p_final.png", plot = joni_p_final, width = 6000, height = 4000, 
       units = "px")