Tea, Earl Grey, Hot

My submission for #TidyTuesday 2021 week 34. A look at StarTrek TNG voice interactions with the Enterprise’s computer. In this submission I focus on the ‘locate’ command to find someone on the ship.

The Task

This week’s #TidyTuesday dataset is all about Star Trek The Next Generation.

In particular, the data collected by www.speechinteraction.org/TNG/ is about voice interactions of the characters with the ship’s computer. While the dataset comprises all kinds of voice interactions (questions, commands and other utterances) I focus on the ‘locate-command’ alone.

With it, characters can locate other people on the ship, if they are looking for them.¹

The way

Setup of the environment

First let’s load the required packages:

Show code

library("tidyverse")
library("ggrepel")
library("ggdark")
library("showtext")
library("reactable")

Then we need to setup the custom fonts for the plot. In this post I do not load the jolly_theme.R².

The Star Trek related fonts come with the {trekfont}-package.

Show code

font <- c("StarNext", "TNGcast")
path <- system.file(paste0("fonts/", font, ".ttf"), package = "trekfont")
for (i in 1:2) font_add(font[i], path[i])
font_add_google("Open Sans")
font_families()

#> [1] "sans"         "serif"        "mono"         "wqy-microhei"
#> [5] "StarNext"     "TNGcast"      "Open Sans"

Show code

showtext_auto()

Data preparation

It all begins with the download of the #TidyTuesday dataset from github:

Show code

#load the data and store locally for future runs of the code
tuesdata <- tidytuesdayR::tt_load(2021, week = 34)

#> 
#>  Downloading file 1 of 1: `computer.csv`

Show code

computer <- tuesdata$computer

Count the voice commands

Counting the number of location-commands is quite easy, as the dataset contains a column specifying who issues the command:

Show code

# count how often a character located someone else
searches_by_people <- computer %>%
  # ignore interactions by the computer and ignore the Wake Word "Computer" itself
  filter(sub_domain == "Locate", !str_detect(char, pattern = "[Cc]omputer"), type != "Wake Word")


searches_by_people_count <- searches_by_people %>%
  count(char, sort = TRUE) %>%
  mutate(
    char = str_to_lower(char),

    # use last name for later joining
    char = ifelse(char == "beverly", "crusher", char),
    char = ifelse(char == "geordi", "la forge", char)
  )

Checking, how often someone is being looked for is not as straight forward. Due to time limitations I took a shortcut and compromised possible mis-counts. I basically filter the voice commands for occurrences of the main characters’ names.

As the number of rows with location commands is <100 I skimmed the commands for the names used to locate people and put these in a vector.

Show code

# Define People of interest (this is not a complete cast list, but the result of skimming ~90 entries)
people <- str_to_lower(c(
  "data", "picard", "captain", "riker", "pulaski", "Goss", "Tam Elbrun", "Barclay",
  "Dalen Quaice", "Hill and Selar", "Worf", "La Forge", "Vash", "Diana", "Troi", "Crusher", "Ensign Ro",
  "Alexander Rozhenko", "Uhnari", "Morag"
))

# Create a Regex pattern by collapsing the vector with the "or" operator
people_pattern <- paste0(people, collapse = "|")


people_searched <- searches_by_people %>%
  mutate(
    # make the interactions strings to lower case
    interaction_lower = str_to_lower(interaction),

    # reduce the interactions strings to the searched person
    # e.g. from "computer, locate commander riker" --> "riker" is extracted.
    # Caution: This is not the best / generalizable way, but a rather hacky approach
    # due to limited time. It works for this use case / dataset.
    person_of_interest = str_extract(interaction_lower, pattern = people_pattern)
  ) %>%
  select(interaction, person_of_interest) %>%
  filter(!is.na(person_of_interest)) %>%
  count(person_of_interest, sort = TRUE) %>%
  mutate(person_of_interest = ifelse(person_of_interest == "captain", "picard", person_of_interest))

Enriching the dataset

I created a csv containing the glyphs used for the characters of interest in the TNGcast-font. In Addition I took the appropriate Federation Uniform Colors from the {trekcolors} package.

Show code

relevant <- read_csv2("res/relevant.csv", col_names = TRUE)

# let's take a look:
relevant %>%
  filter(!is.na(char)) %>%
  reactable(
    # global reactable options
    defaultSorted = "char",
    # defaultSortOrder = "desc",
    searchable = TRUE,
    highlight = TRUE,
    rowStyle = list(cursor = "pointer"),
    theme = reactableTheme(
      highlightColor = "#1BC7DC"
    ),

    # formatting individual columns
    columns =
      list(
        char = colDef(
          name = "Character Name",
          sortable = TRUE,
          minWidth = 150
        ),
        char_label = colDef(
          name = "Label glyph",
          minWidth = 50,
          sortable = TRUE
        ),
        char_col = colDef(
          name = "Uniform color HEX",
          minWidth = 100,
          sortable = TRUE,
          style = function(value) {
            list(background = value)
          }
        )
      )
  )

As last step before plotting the data is combined:

Show code

whereabouts <- searches_by_people_count %>%
  full_join(people_searched, by = c("char" = "person_of_interest")) %>%
  rename(searching = n.x, searched = n.y) %>%
  mutate(char = str_to_title(char)) %>%
  replace_na(list(searching = 0L, searched = 0L)) %>%
  inner_join(relevant, by = "char")

The result

Now, that the data has been prepared the plot can be drawn.

Show code

whereabouts %>%
  ggplot(aes(searching, searched)) +
  geom_point(size = 3) +
  geom_label_repel(
    aes(label = char_label, color = char_col),
    box.padding = 0.5,
    label.padding = 0.5,
    max.time = 1,
    max.iter = 100000,
    family = "TNGcast",
    size = 30
  ) +
  labs(
    title = "Where is Captain Picard?",
    subtitle = "How often did Characters in 'StarTrek TNG' ask the computer to locate someone on the Starship Enterprise\nvs. how often are they being located via the computer.\n",
    x = "Times searching someone",
    y = "Times being searched",
    caption = "\n@c_gebhard | #TidyTuesday Week 34 (2021)\nData source: http://www.speechinteraction.org/TNG/"
  ) +
  coord_trans(x = "sqrt", y = "sqrt") +
  scale_x_continuous(breaks = c(0:6, 10, 15, 18)) +
  scale_y_continuous(breaks = c(1:7)) +
  scale_color_identity() +
  dark_theme_minimal() +
  theme(
    plot.title = element_text(
      family = "StarNext",
      face = "bold",
      size = rel(3),
      hjust = 0,
      vjust = 5
    ),
    plot.subtitle = element_text(
      family = "Open Sans",
      size = rel(1.3),
      hjust = 0
    ),
    plot.caption = element_text(
      size = rel(1.1),
      face = "italic",
      hjust = 1
    ),
    plot.caption.position = "plot",
    plot.margin = margin(1.5, 0.4, 0.4, 0.4, unit = "cm"),
    axis.title = element_text(
      face = "bold",
      size = rel(1.3)
    ),
    axis.title.x = element_text(margin = margin(t = 15, r = 0, b = 0, l = 0)),
    axis.title.y = element_text(margin = margin(t = 0, r = 15, b = 0, l = 0), angle = 90),
    axis.text = element_text(
      size = rel(1.3)
    )
  )

Figure 1: Characters in StarTrek The Next Generatio (TNG) frequently interact with the shep’s computer via voice commands. One of the computer’s functions is to locate a person on the ship. Within the speechinteractions.org/TNG/ dataset, these ‘locate-commands’ were filtered and analysed. The characters are plotted in regard to how often the used the ‘locate-command’ to find someone vs. how often they are being located via the computer.

Show code

ggsave("tt21-34_picard.png", dpi = 96, height = 8, width = 10)

Note that the “officially” submitted plot³ differs from the one above. To meet the deadline I submitted a simpler version with a simple scatterplot.

Comments

Being a Star Trek fan I really enjoyed working on the dataset. In this post I shared what I learned in regard to custom fonts and using the {reactable} package. I hope it was informative to read. If there’s something missing, let me know:

---

1. Handy, if you really need to find someone at any time, yet kind of creepy if you think about it.↩︎

2. check out this post↩︎

3. aka the tweeted version↩︎