Day 77: Modularizing functionality & Better emoji displays

2017-05-09_emoji.png

2017-05-10_emoji.png

emoji-data-science-twitter.R

library(twitteR)
library(dplyr)
library(readr)
library(stringr)
library(ggplot2)


# Prereq ------------------------------------------------------------------

# For visualization of emojis, you will need to download the
# EmojiOne*.ttf file from https://github.com/eosrei/emojione-color-font


# Functions ---------------------------------------------------------------

pull_tweets <- function(search_string, ...){
  set.seed(20170202)
  # Pull raw tweets
  tweets_raw <-
    searchTwitter(search_string, ...)

  # Remove retweets & convert twitteR lists to data.frames
  tweets_tidy1 <-
    strip_retweets(tweets_raw, strip_manual = TRUE, strip_mt = TRUE) %>%
    twListToDF()

  tweets_tidy2 <-
    tweets_tidy1 %>%
    mutate(
      # Add new columns containing the hashtag & tweet url
      hashtag = search_string,
      url = paste0('https://twitter.com/', screenName, '/status/', id),

      # Convert character vector between encodings
      text = iconv(text, from='latin1', to='ASCII', sub='byte'),

      # Update type
      created = lubridate::ymd_hms(created, tz = "UTC")
    ) %>%
    rename(retweets = retweetCount)

  list(
    raw = tweets_tidy1,
    tidy = tweets_tidy2
  )
}


load_dict_emoji <- function(emoji_file = "emDict.csv"){
  # Download emoji_file if it does not exist
  if (!file.exists(emoji_file)) {
    emoji_file <- "emDict.csv"
    emoji_file_url <-
      paste0(
        "https://raw.githubusercontent.com/today-is-a-good-day/emojis/master/",
        emoji_file)

    download.file(emoji_file_url, destfile = emoji_file)
  }

  readr::read_delim(
    emoji_file,
    delim = ";",
    col_names = c("description", "native", "bytes", "r_encoding"),
    skip = 1,
    progress = FALSE
  ) %>%
    mutate(description = tolower(description))
}


count_emoji <- function(emoticons, the_tweets){
  # Helper to: Count number of times a pattern occur in string
  count_pattern <- function(string, pattern){
    counts <- str_count(string, pattern)
    data.frame(
      counts,
      tweet_id = 1:length(counts))
  }

  # Count number of emojis for each tweet
  emoji_counts <-
    emoticons %>%
    mutate(
      counts = purrr::map(r_encoding, ~count_pattern(the_tweets, .x))
    ) %>%
    tidyr::unnest(counts) %>%
    filter(counts != 0)

  # Summarize the counts per emoji
  (emoji_counts_summary <-
    emoji_counts %>%
    group_by(description, native) %>%
    summarise(count = sum(counts)) %>%
    arrange(desc(count))
  )
}

visualize_counts <- function(count_summary, filename = NULL, titlename = "",
                             use_cowplot = TRUE, font_style = "EmojiOne", ...){
  if (use_cowplot) {
    library(cowplot)
  }

  plt <-
    count_summary %>%
    ggplot(aes(x=reorder(description, -count),
               y=count,
               label=native)) +
    geom_bar(stat="identity", fill="grey88") +
    xlab("") +
    ggtitle(titlename) +
    theme(axis.text.x = element_text(angle=90, hjust=1, vjust=.5))

  if (require("emojifont")) {
    # EmojiOne.ttf file downloaded from
    # https://github.com/eosrei/emojione-color-font
    paste0(font_style, ".ttf") %>%
      load.emojifont()
    plt + geom_text(family=font_style, size=5)
  } else {
    plt
  }
  if (!is.null(filename)) ggsave(filename, ...)
}


# Do something ------------------------------------------------------------

# Create API keys from https://apps.twitter.com
#   api_key <- 'XXX'
#   api_secret <- 'XXX'
#   access_token <- 'XXX'
#   access_token_secret <- 'XXX'
# To load true values:
source("twitter_api_key.R")
setup_twitter_oauth(api_key, api_secret, access_token, access_token_secret)


emoticons <- load_dict_emoji()
search_string <- "#BCVotes2017"

## During election day
tweets <- pull_tweets(search_string,
                      n = 10000,
                      lang = 'en',
                      since = '2017-05-09',
                      until = '2017-05-10')
emoji_counts_summary <-
  count_emoji(emoticons, the_tweets = tweets$tidy$text)
visualize_counts(
  emoji_counts_summary %>% filter(count > 1),
  filename = "~/Desktop/2017-05-09_emoji.pdf",
  titlename =
    paste(search_string, "during Election Day (May 9, 2017)"),
  height = 7)

## After election day
tweets <- pull_tweets(search_string,
                      n = 10000,
                      lang = 'en',
                      since = '2017-05-10',
                      until = '2017-05-11')
emoji_counts_summary <-
  count_emoji(emoticons, the_tweets = tweets$tidy$text)
visualize_counts(
  emoji_counts_summary %>% filter(count > 1),
  filename = "~/Desktop/2017-05-10_emoji.pdf",
  titlename =
    paste(search_string, "a day after Election Day (May 10, 2017)"),
  scale = 1.5, width = 7)