Sentiment analysis of Avatar

Understanding Characters through Avatar Episode Scripts.
tidytuesday
Author

Ted Laderas

Published

August 11, 2020

Look at the available datasets

library(tidytuesdayR)

Loading the Data

#incoming data comes in as a list
datasets <- tidytuesdayR::tt_load("2020-08-11")
--- Compiling #TidyTuesday Information for 2020-08-11 ----
--- There are 2 files available ---
--- Starting Download ---

    Downloading file 1 of 2: `avatar.csv`
    Downloading file 2 of 2: `scene_description.csv`
--- Download complete ---
#show the names of the individual datasets
names(datasets)
[1] "avatar"            "scene_description"
avatar <- datasets$avatar
avatar[1:5,]
# A tibble: 5 × 11
     id book  book_num chapter   chapter_num character full_text character_words
  <dbl> <chr>    <dbl> <chr>           <dbl> <chr>     <chr>     <chr>          
1     1 Water        1 The Boy …           1 Katara    "Water. … Water. Earth. …
2     2 Water        1 The Boy …           1 Scene De… "As the … <NA>           
3     3 Water        1 The Boy …           1 Sokka     "It's no… It's not getti…
4     4 Water        1 The Boy …           1 Scene De… "The sho… <NA>           
5     5 Water        1 The Boy …           1 Katara    "[Happil… Sokka, look!   
# ℹ 3 more variables: writer <chr>, director <chr>, imdb_rating <dbl>
scenes <- datasets$scene_description
scenes[1:5,]
# A tibble: 5 × 2
     id scene_description                                                       
  <dbl> <chr>                                                                   
1     3 [Close-up of the boy as he grins confidently over his shoulder in the d…
2     5 [Happily surprised.]                                                    
3     6 [Close-up of Sokka; whispering.]                                        
4     6 [A look of bliss adorns his face. He licks his lips and wiggles his fin…
5     8 [Struggling with the water that passes right in front of her.]          

My Research Question

Does the sentiment of each character change over the multiple seasons? That is, does a character become more positive or more negative as their character develops?

I will attempt to summarize the sentiment of each character across each episode.

Using tidytext to unnest_tokens() - that is, split each line into 1 word per row.

library(tidytext)
library(tidyverse)
── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
✔ dplyr     1.1.4     ✔ readr     2.1.5
✔ forcats   1.0.0     ✔ stringr   1.5.1
✔ ggplot2   3.5.1     ✔ tibble    3.2.1
✔ lubridate 1.9.3     ✔ tidyr     1.3.1
✔ purrr     1.0.2     
── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag()    masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
avatar_words <- avatar %>%
  select(id, book, book_num, chapter, chapter_num, character, character_words) %>%
  filter(character != "Scene Description") %>%
  unnest_tokens(word, character_words)

avatar_words[1:10,]
# A tibble: 10 × 7
      id book  book_num chapter                chapter_num character word       
   <dbl> <chr>    <dbl> <chr>                        <dbl> <chr>     <chr>      
 1     1 Water        1 The Boy in the Iceberg           1 Katara    water      
 2     1 Water        1 The Boy in the Iceberg           1 Katara    earth      
 3     1 Water        1 The Boy in the Iceberg           1 Katara    fire       
 4     1 Water        1 The Boy in the Iceberg           1 Katara    air        
 5     1 Water        1 The Boy in the Iceberg           1 Katara    my         
 6     1 Water        1 The Boy in the Iceberg           1 Katara    grandmother
 7     1 Water        1 The Boy in the Iceberg           1 Katara    used       
 8     1 Water        1 The Boy in the Iceberg           1 Katara    to         
 9     1 Water        1 The Boy in the Iceberg           1 Katara    tell       
10     1 Water        1 The Boy in the Iceberg           1 Katara    me         

Characters by Episode/Chapter

episode_count <- avatar %>%
  count(character, chapter) %>%
  select(character, chapter) %>%
  filter(character != "Scene Description") %>%
  distinct() %>%
  count(character) %>%
  arrange(desc(n))

episode_count %>% 
  DT::datatable()

Bit Players

episode_count %>%
  filter(n == 1) %>%
  arrange(character)
# A tibble: 272 × 2
   character         n
   <chr>         <int>
 1 Aang and Zuko     1
 2 Aang:             1
 3 Actor Bumi        1
 4 Actor Iroh        1
 5 Actor Jet         1
 6 Actor Ozai        1
 7 Actor Sokka       1
 8 Actor Toph        1
 9 Actor Zuko        1
10 Actress Azula     1
# ℹ 262 more rows

Cabbage merchant

The cabbage merchant appears in 4 episodes, and you can see his path to resignation as Aang and company keep busting up his cabbage kiosk.

avatar %>%
  filter(character == "Cabbage merchant") %>%
  select(chapter, character_words) %>%
  gt::gt()
chapter character_words
The King of Omashu No! My cabbages!
The King of Omashu My cabbages! You're gonna pay for this!
The King of Omashu Off with their heads! One for each head of cabbage!
The King of Omashu My cabbages!
The Waterbending Scroll My cabbages! This place is worse than Omashu!
The Serpent's Pass Ahhh! My cabbages!
The Tales of Ba Sing Se My cabba- Oh, forget it.

Who Spoke the Most?

Surprisingly, Sokka has the most lines.

line_count <- avatar_words %>% 
  count(character) %>%
  arrange(desc(n)) 

line_count[1:20,] %>%
  gt::gt()
character n
Sokka 18293
Aang 17821
Katara 14961
Zuko 8972
Toph 5434
Iroh 5252
Azula 3299
Zhao 1607
Jet 1604
Suki 1221
Hakoda 1065
Pathik 1030
Roku 1015
Ozai 1002
Hama 955
Mai 844
Bumi 818
Long Feng 757
Warden 722
Ty Lee 705

Understanding Each Character’s Journey

Using tidytext, I do a sentiment analysis of each episode (here called a chapter) to determine the overal sentiment for a character.

bing <- get_sentiments("bing")

characters <- c("Aang", "Katara", "Zuko", "Toph", "Iroh", "Sokka", "Azula", "Mai", "Ty Lee")

sentiment_summary <- avatar_words %>%
  inner_join(bing) %>%
  count(book_num, chapter_num, chapter, character, sentiment) %>%
  filter(character %in% characters) %>%
  arrange(book_num, chapter_num) %>%
  pivot_wider(names_from = sentiment, values_from = n) %>%
  mutate(positive = tidyr::replace_na(positive, 0),
         negative = tidyr::replace_na(negative, 0)) %>%
  mutate(sentiment = positive - negative)
Joining with `by = join_by(word)`
index_chapters <- avatar_words %>%
  select(book_num, chapter_num) %>%
  distinct() %>%
  mutate(index = row_number())
sentiment_index <- sentiment_summary %>% 
  inner_join(y= index_chapters, by=c("book_num", "chapter_num"))
out_plot <- ggplot(sentiment_index) +
  aes(x=index, y=sentiment, fill=character, episode=chapter, book = book_num, episode_number=chapter_num) +
  geom_col(show_legend = FALSE) +
  facet_wrap(~character, ncol=2) +
  labs(title= "Each Character's Sentiment Journey", x="Episode Number",
       subtitle = "mouse over each graph for more information") + 
  geom_vline(xintercept = 21, lty=2) +
  geom_vline(xintercept = 41, lty=2)
Warning in geom_col(show_legend = FALSE): Ignoring unknown parameters:
`show_legend`
plotly::ggplotly(out_plot)

Sentiment Heatmap

s_index <- sentiment_index %>%
  tidyr::complete(chapter_num, character)

ggplot(s_index) +
  aes(x=index, y=character, fill=sentiment) +
geom_tile() +
  scale_fill_viridis_b(na.value="black") 
Warning: Removed 30 rows containing missing values or values outside the scale range
(`geom_tile()`).

Zuko has the most interesting journey

Zuko has many ups and downs, which may reflect his overall lack of confidence and his tendency for self-loathing.

zuko <- sentiment_index %>%
  filter(character=="Zuko")

out_plot <- ggplot(zuko) +
  aes(x=index, y=sentiment, fill=character, episode=chapter, book = book_num, group=character, episode_number=chapter_num) +
  geom_col(show_legend = FALSE) +
  facet_wrap(~character, ncol=2) +
  annotate(geom="text", x=27, y= -8 , label = "Zuko Alone\nA Turning Point") +
  annotate(geom="text", x=53, y = 11, label = "Where Zuko\ntrains Aang") +
  labs(title= "Zuko has lots of ups and downs", x="Episode Number",
       subtitle = "mouse over for more episode information") +
  ylim(c(-13, 13)) +
  geom_vline(xintercept = 21, lty=2) +
  geom_vline(xintercept = 41, lty=2)
Warning in geom_col(show_legend = FALSE): Ignoring unknown parameters:
`show_legend`
plotly::ggplotly(out_plot)

Aang and Zuko’s Journeys

Plotting the sentiment journey of Zuko and Aang together shows that they often mirror each other, except in the last parts of Book 3.

zuko_aang <- sentiment_index %>%
  filter(character %in% c("Zuko", "Aang"))

out_plot <- ggplot(zuko_aang) +
  aes(x=index, y=sentiment, fill=character, episode=chapter, book = book_num, episode_number=chapter_num) +
  geom_col(show_legend = FALSE, alpha=0.7) +
  labs(title= "Aang and Zuko's Journeys Often Mirror Each Other", 
       x="Episode Number",
       subtitle = "mouse over for more episode information") +
  ylim(c(-13, 13)) +
    geom_vline(xintercept = 21, lty=2) +
  geom_vline(xintercept = 41, lty=2)
Warning in geom_col(show_legend = FALSE, alpha = 0.7): Ignoring unknown
parameters: `show_legend`
plotly::ggplotly(out_plot)

Iroh is so chill and positive

iroh <- sentiment_index %>%
  filter(character=="Iroh")

out_plot <- ggplot(iroh) +
  aes(x=index, y=sentiment, fill=character, episode=chapter, book = book_num, episode_number=chapter_num) +
  geom_col(show_legend = FALSE) +
  labs(title= "Iroh is just so chill and positive", x="Episode Number",
       subtitle = "mouse over for more episode information") +
  ylim(c(-13, 13)) +
    geom_vline(xintercept = 21, lty=2) +
  geom_vline(xintercept = 41, lty=2)
Warning in geom_col(show_legend = FALSE): Ignoring unknown parameters:
`show_legend`
plotly::ggplotly(out_plot)

Citation

BibTeX citation:
@online{laderas2020,
  author = {Laderas, Ted and Laderas, Ted},
  title = {Sentiment Analysis of {Avatar}},
  date = {2020-08-11},
  url = {https://laderast.github.io/articles/2020-08-11-avatar_last_airbender/},
  langid = {en}
}
For attribution, please cite this work as:
Laderas, Ted, and Ted Laderas. 2020. “Sentiment Analysis of Avatar.” August 11, 2020. https://laderast.github.io/articles/2020-08-11-avatar_last_airbender/.