[Code]

---
title: "Star Trek"
author: "NearAndDistant"
date: "17/08/2021"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

#### Import
```{r}
library(tidyverse)

computer_raw <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2021/2021-08-17/computer.csv')

```

*Person*
Command: Utterances that directly tell the computer what to do. 
Question: Utterances that ask the computer for something.
Statement: Utterances tell the computer or ask it but meaning is inferred.
Password: Utterances that contain a password.
Wake Word: Key phrases used to activate the computer VUI.
Comment: Utterances that have no intended action for the computer
Conversation: Utterances that are more like human conversation, such as phatic expressions, formalities, and colloquial speech

*Domain*
InfoSeek: Querying any information from the computer, somewhat similar to a Google search, but also status updates, etc.
Locate: Querying the physical location of a person, usually on board the ship
Entertainment: Queries related to music, holodeck, other pastimes.
Music: Queries related to music
Holodeck: Queries related to operating the holodeck.
IoT: Queries that activates or uses another piece of hardware, like turning on lights or heating by voice in a smart house. 
Replicator: Queries relating to using the replicator
Turbolift: Queries relating to using the turbolift
Analysis: Queries that ask the computer to perform an
Communication: Queries that interact with digital mmunications
Help: Queries that seek help in using the system
Emergency: Queries around emergencies like setting the auto-destruct

*Computer*
Clarification: Utterances asking for more information.
Response: Utterances that respond to a person's query or action
System Alert: Utterances that either respond to a person requested action (often a warning) or are prompted by the system, rather than a person
Information: Utterances that inform users of what is happening without being prompted to do so (which would be a response)
Countdown: Utterances that are part of a countdown.
In Progress: Utterances that indicates that a process is ongoing or updates on that process
Conversation: Utterances that are more like human conversation, such as phatic expressions, formalities, and colloquial speech


#### Clean & Wrangle
```{r}

computer <-
  computer_raw %>%
    rename(character_type   = char_type ,
           character        = char      ,  
           stage_direction  = direction ,  
           interaction_type = type      ,  
           com_type         = pri_type  ,  
           com_domain       = domain    ,  
           com_sub_domain   = sub_domain,  
           no_response_comp = nv_resp   , 
           error_response   = error) %>%
  mutate(character = if_else(str_detect(character , "Computer"), "Computer", character),
         character = if_else(str_detect(character , "Com Panel"), "Computer", character),
         character = if_else(str_detect(character , "Picard"), "Picard", character),
         character = str_remove(character , " \\(O.S.\\)"),
         character = str_remove(character , " \\(V.O.\\)"),
         character = str_remove(character , " \\(Cont'D\\)"),
         character = str_remove(character , " \\(O.C.\\)"),
         character = str_remove(character , "Young ")) %>%
  mutate(interaction_type = str_to_title(interaction_type),
         com_domain       = if_else(str_detect(com_domain , "Iot"), "IoT", com_domain),
         com_sub_domain   =  str_remove(com_sub_domain , "\\?")) %>%
  mutate(species = case_when(character == "Worf"           ~ "Humanoid",
                             character == "Troi"           ~ "Humanoid",
                             character == "Beverly"        ~ "Humanoid",
                             character == "Riker"          ~ "Humanoid",
                             character == "Data"           ~ "Android",
                             character == "Picard"         ~ "Humanoid",
                             character == "Geordi"         ~ "Humanoid",
                             character == "Computer"       ~ "Computer")) %>%
  mutate(species = if_else(is.na(species), "Humanoid" , species)) %>%
  select(character_type , character, species , interaction_type , com_type , com_domain , com_sub_domain , line , no_response_comp , error_response) 

```

#### EDA
```{r}
theme_set(theme_minimal())

computer %>%
  filter(!is.na(species)) %>%
ggplot(aes(interaction_type , fill = interaction_type)) +
  geom_histogram(stat = "count")+
  coord_flip() +
  facet_wrap(~species , nrow = 1) +
  theme(legend.position = "none")

```

#### Text Analysis
```{r}

library(tidytext)
library(tidyr)

computer_unnested <- 
computer %>%
  unnest_tokens("word" , line) %>%
  anti_join(stop_words , by = "word") %>%
  add_count(word , name = "word_total" , sort = TRUE)

```

#### EDA on Tidy Text
```{r}

computer_unnested %>%
  filter(!is.na(species),
         word_total >= 35,
         word != "computer") %>%
    ggplot(aes(word , fill = word)) +
      geom_histogram(stat = "count")+
      coord_flip() +
      facet_wrap(~species) +
      theme(legend.position = "none")

```

```{r}

computer_unnested %>%
  filter(!is.na(species),
         word_total >= 50,
         word != "computer") %>%
    ggplot(aes(reorder(word , word_total) , fill = species)) +
      geom_histogram(stat = "count") +
      coord_flip() +
      facet_wrap(~interaction_type , scale = "free_y")

```

#### Palette
```{r}

background <- "#C3CBD4"

```

#### Frequency Analysis
```{r}

st_frequencies <-
computer_unnested %>%
  filter(word != "computer", word != "program") %>%
  mutate(humanoid = if_else(species == "Humanoid" , "Humanoid" , "Non-Humanoid")) %>%
  select(humanoid , species , word) %>%
  mutate(word_n = n()) %>% # total words  = 11,185 (minus "computer")
  add_count(word , name = "word_total" , sort = TRUE) %>% # frequency of words across dataset
  group_by(word) %>%
  mutate(word_pc = word_total / word_n) %>% # test: distinct(word , word_pc) & sum == 1
  ungroup() %>%
  group_by(species) %>%
  mutate(species_n = n()) %>% # test: distinct(species_n) %>% summarise(species_tot = sum(species_n))
  add_count(species,  word , name = "species_word_total" , sort = TRUE) %>% # frequency of words across species
  mutate(species_pc = species_word_total / species_n) %>%
  ungroup() %>%
  distinct(species , word , .keep_all = TRUE)

```

#### Frequency Graphic
```{r}
library(scales)
library(ggrepel)

showtext::showtext_auto()
sysfonts::font_add_google("Orbitron" , "orbi")
sysfonts::font_add_google("Rationale" , "rationale")

set.seed(111)

plot_freq <-
st_frequencies %>%
  ggplot(aes(x = species_pc, y = word_pc)) + # color = abs(word_pc - species_pc
    #geom_jitter(size = 0.5, width = 0.75, height = 0.5, alpha = 0.10) +
    geom_text_repel(aes(label = word) , color = "grey20" , segment.color = NA , size = 3 , family = "rationale", max.overlaps = 8) +
    geom_abline(color = "#2070b6" , alpha = 0.5) +
    scale_x_log10(labels = percent_format()) +
    scale_y_log10(labels = percent_format()) +
    scale_color_viridis_c(option = "cividis" , begin = 0 , end = 0.75) +
    facet_wrap(~species , ncol = 3) +
    labs(x = "Species Specific Word Weighting", y = "Overall Word Weighting") + 
    theme(legend.position  ="none",
          panel.grid       = element_blank(),
          panel.background = element_rect(fill = background , color = background),
          plot.background  = element_rect(fill = background, color = background),
          text             = element_text(family = "rationale"),
          axis.title.x     = element_text(size = 14, vjust = -3),
          axis.title.y     = element_text(size = 12),
          strip.background = element_blank(), # fill = background , color = background
          strip.text       = element_text(family = "orbi" , face = "bold" , size = 18),
          plot.margin      = margin(1.5,1.2,0.5,0.1, unit = "cm")) # 1.5,1.5,0.5,1, # 1.5,1.5,0.5,1,

```

#### Sentiment Analysis
```{r}

st_sentiment_analysis <-
  computer_unnested %>%
  inner_join(sentiments) %>%
  count(species , interaction_type, sentiment , sort = TRUE) %>%
  pivot_wider(names_from = sentiment, values_from = n) %>% 
  mutate(sentiment = positive - negative)

```

#### Sentiment Graphic
```{r}

plot_sentiment <-
st_sentiment_analysis %>%
  ggplot(aes(interaction_type, sentiment, fill = species)) +
    geom_col(show.legend = FALSE) +
    geom_text(aes(label = interaction_type), family = "rationale" , size = 5, color = "grey30") +
    scale_y_continuous(breaks = seq(-40, 40, by = 10), limits = c(-40, 40)) +
    scale_fill_viridis_d(option = "cividis" , alpha = 0.35, begin = 0.75 , end = 0) +
    coord_flip() +
    facet_wrap(~species, ncol = 3) + 
  labs(x = "Interaction Type",
       y = "Sentiment") +
  theme(panel.background = element_rect(fill = background , color = background),
        plot.background  = element_rect(fill = background , color = background),
        panel.grid       = element_blank(),
        text             = element_text(family = "rationale"),
        axis.title.x     = element_text(size = 14 , vjust = -3),
        axis.ticks       = element_blank(),
        axis.title.y     = element_text(size = 12 , vjust = 7),
        axis.text.y      = element_blank(),
        strip.text       = element_blank(),
        plot.margin      = margin(1,1.5,1,0.8, unit = "cm"))

```

#### Panel
```{r}
library(gridExtra)

plot_arranged <-
grid.arrange(plot_freq , plot_sentiment,
             ncol = 1 , heights = c(1 , 0.50))

```

#### Cowplot
```{r}
library(cowplot)

ggdraw(plot_arranged) +
  draw_image("https://brandslogos.com/wp-content/uploads/images/large/star-trek-logo.png",
             height = 0.5 , width = 0.5,
             x = 0.26 , y = 0.24) +
  draw_image("https://i.pinimg.com/originals/d6/7d/5c/d67d5c43befed8feae8f60857c51b4a6.png",
             height = 0.25 , width = 0.25,
             x = -0.065 , y = 0.79) +
  draw_text("Using frequency analysis on Star Trek\ninteractions we can analyse the different\nwords Humanoids, Androids and Computers\nuse in Star Trek",
            x = 0.80 , y = 0.60 , size = 9 , family = "rationale" , hjust = 0 , color = "#2070b6") +
  draw_text("Using sentiment analysis\nwe can see the android Data\nuses far less emotive words\nacross all types of interactions\nwith Computer than Humanoids\nor even Computer does",
            x = 0.05 , y = 0.2 , size = 9 , family = "rationale" , hjust = 0 , color = "#2070b6") +
  draw_text("Using both Computer plots we can\n see that Computer uses a lot of negative\nsentiment and time words begging the\nquestion if Computer is a glorified clock",
            x = 0.52 , y = 0.24 , size = 9 , family = "rationale" , hjust = 0 , color = "#2070b6")

```

#### Saving
```{r}

# save plot
ggsave(here::here("Do Androids Dream of Star Trek?.png"), dpi = 360, height = 10, width = 16)

```