Twimoji

A fluffy experiment in analyzing emoji use in tweets

Lauren Ancona @laurenancona Christopher Tufts @devlintufts

K.I.S.S.

Hands-free tweet collector

Overview

• May 9 - July 11, 2015
• 28,799 location-enabled tweets
• 7,363 unique users

Tweet Cleaning

emoji.frequency <- matrix(NA, nrow = nrow(ds), ncol = nrow(emoticons))
for(i in 1:nrow(emoticons)){
  emoji.frequency[,i] <- regexpr(emoticons$bytes[i],ds$text, useBytes = T )	
  }
emoji.per.tweet <- rowSums(emoji.frequency > -1)
emoji.indexes <- which( emoji.per.tweet > 0)
emoji.ds <- NULL
for(i in emoji.indexes){
  valid.cols <- which(emoji.frequency[i,]>-1)
  for(j in valid.cols){
    emoji.ds <- rbind(cbind(ds[i,], emoticons[j,]), emoji.ds)
  }
}

Emoji

• ~ 25% of tweets contained emoji
• 590 unique emoji
• 7,131 tweets

Top Emoji

Possibilities

• Geo-analysis with available open data
  • True on-the-fly rendering as map markers (JS is hard)
• Lexical research
  • gender, age
  • psychological characteristics

Caveats

• Estimated only 1% of tweets have location
• Emoji meaning vs. use
• Inconsistent rendering across platforms

Resources

• Emoji Cheat Sheet
• Tim Whitlock: Emoji Utils
• CartoDB (Torque.js)

Friday, July 17, 2015

World Emoji Day

Code + data:

bit.ly/twimoji

@laurenancona | @devlintufts