Bojack the Data Scientist

New Season of Bojack Horseman – NO SPOILERS!

A new season of my favorite show, Bojack Horseman, just dropped on Netflix, and I have absolutely zero time to sit down and watch it. However, in some magical way there was time for me to write this blog post and do this analysis. What analysis you ask? I’ll tell you in a minute, I ain’t Horsin’ Around.

Bojack Horseman, in case you got stuck in 2013, is an animated comedy-drama about a washed up ’90s sitcom star, who is also a horse (animals are anthropomorphized in the show’s universe). If you follow Bojack, you know that it isn’t your everyday animated series. It’s deep, cynical and lights up some of the darkest corners of damaged human behavior. Moreover, it keeps getting better, or should I say – heavier.

The psychologist in me got triggered, if indeed Bojack is getting deeper and deeper, perhaps this pattern could be observed in the show’s script? I mean, would the show runners use more abstract language compared to concrete language as the show progresses? The data scientist in me said “I’m on it”.

Scrape Bojack Script and Language Analysis

First, I had to scrape Bojack’s script. I’ll do it the rvest, xml2 and stringer packages. It’s actually my first doing web scarping so apologies for the non-elegant code

library(xml2)
library(rvest)
library(stringr)

#data frame pre-allocation
m <- matrix(nrow = 5*10, ncol = 3)
colnames(m) <- c("season", "episode", "text")
m <- as.data.frame(m)

count = 1
for (i in 1:5){
  for (j in 1:10){
    if (j<10){
      scrappedurl <-paste0("https://www.springfieldspringfield.co.uk/view_episode_scripts.php?tv-show=bojack-horseman-2014&episode=s0",
                           i, "e0", j)
      html.raw <- read_html(scrappedurl)
      n <- html_nodes(html.raw, "div#content_container")
      txt <- html_text(n)
      txt <- str_replace_all(txt, "[rnt]" , "")
      t <- str_split(txt, "Episode Script", simplify = T)[3]
      m$season[count] <- i
      m$episode[count] <- j
      m$text[count] <- t
    }
    else{
      scrappedurl <-paste0("https://www.springfieldspringfield.co.uk/view_episode_scripts.php?tv-show=bojack-horseman-2014&episode=s0",
                           i, "e", j)
      html.raw <- read_html(scrappedurl)
      n <- html_nodes(html.raw, "div#content_container")
      txt <- html_text(n)
      txt <- str_replace_all(txt, "[rnt]" , "")
      t <- str_split(txt, "Episode Script", simplify = T)[3]
      m$season[count] <- i
      m$episode[count] <- j
      m$text[count] <- t
    }
    count <-  count+1
  }
}

library(kableExtra)
#let's see the dataframe
kable(m) %>%
  kable_styling() %>%
  scroll_box(width = "100%", height = "500px")