Introduction to Emoji Data Science (Open Data Science Conference, 2017)
- 1. Hamdan Azhar hamdan@prismoji.com @hamdanazhar // November 3, 2017 From data with ⤠// An introduction to Emoji Data Science
- 2. đ đđ% Ă & đđ% Industry data science -> Indie data science
- 9. While emojis might seem fun and trivial, they are at their core vessels for conveying human emotion. Emojis are also data. And data science can help us understand them.
- 10. The birth of
- 12. My reaction to this article, in emoji
- 14. How do people use emojis to react to a major political event in real time?
- 15. Getting the data â Use Twitter API to sample 100,000 tweets for five hashtags related to Britainâs EU Referendum § Hashtags: #NotMyVote, #VoteRemain, #EURef, #Brexit, #VoteLeave § Data pulled for June 24, 2016 (the day after the referendum) § English language tweets only § After removing retweets, weâre left with 23,989 unique tweets, i.e. the âBrexit datasetâ § Of these, 1,505 tweets (6.3%) contain at least one emoji
- 16. Analyzing the data § Use regular expressions in R, along with Unicode emoji dictionaries, to extract emojis from tweets § Compute emoji counts in the Brexit dataset § Compare with counts for all >10B emoji tweets on Twitter since 2013 (from emojitracker.com) § Extract hashtags from tweets and compute hashtag profiles for various emojis
- 17. emoji emoji name brexit rank brexit count brexit density (per 1,000) đ face with tears of joy 1 285 11.9 flag of united kingdom 2 249 10.4 đ thumbs up sign 3 75 3.1 đ clapping hands sign 4 70 2.9 ⤠heavy black heart 5 60 2.5 đ loudly crying face 6 48 2.0 đ pensive face 7 40 1.7 đŠ weary face 8 38 1.6 đ˘ crying face 9 34 1.4 đ see-no-evil monkey 10 34 1.4 Brexit chatter includes a wide variety of emojis Of every 1,000 tweets in our dataset, 11.9 of them contain đ
- 19. Can we use hashtag data to understand what these emojis mean?
- 20. Finding the âhashtag signatureâ of a given emoji § We know the distribution of hashtags in our entire dataset § We can pick a given emoji and compute the distribution of hashtags for tweets that use that emoji § By comparing these two distributions, we can estimate which hashtags an emoji is most likely to be used with 15% 17% 20% 29% 19%
- 23. Hearts vs. Snakes: The emoji battle underyling the epic Taylor Swift â Kanye West feud Source: Analysis of 100,000 public tweets mentioning @taylorswift13 and @kanyewest from Aug. 1-4, 2016. (PRISMOJI)
- 24. Taylor Swift is winning hearts (and minds) Source: Analysis of 100,000 public tweets mentioning @taylorswift13 and @kanyewest from Aug. 1-4, 2016. (PRISMOJI) equal higher association with @taylorswift13 higher association with @kanyewest
- 25. Our common emoji language of #fanlove Source: Analysis of 250,000 public tweets mentioning @beyonce, @justinbieber, @djkhaled, @drake, and @rihanna from Aug. 1-4, 2016. (PRISMOJI)
- 26. Sometimes love hurts Examples of in tweets involving #fanlove Source: Analysis of 250,000 public tweets mentioning @beyonce, @justinbieber, @djkhaled, @drake, and @rihanna from Aug. 1-4, 2016. (PRISMOJI)
- 27. Now letâs turn to something more serious
- 28. Emoji reactions to Election Night results Source: Analysis of sampled Election Night tweets, from Nov. 9, 2016 12:00am â 3:30am EST. (PRISMOJI) More common in tweets mentioning TrumpMore common in tweets mentioning Clinton https://prismoji.com/2016/11/26/the-top-emojis-of-election-day-2016/
- 29. Election Day emojis in tweets mentioning Hillary Clinton More common AFTER midnightMore common BEFORE midnight Source: Analysis of sampled Election Day tweets from Nov. 8, 2016 6:00am to Nov. 9, 2016 3:30am. (PRISMOJI)
- 30. Source: Analysis of sampled Election Day tweets from Nov. 8, 2016. (PRISMOJI)
- 31. The divided emoji states of America Source: Analysis of sampled Election Night tweets, from Nov. 9, 2016 12:00am â 3:30am EST. Word clouds shown for tweets mentioning Hillary Clinton or Donald Trump as well as the given emoji above. Randomly selected tweet shown for each emoji.
- 34. Emojis of the resistance #NoBanNoWall #TheResistance #NotMyPresident #WomensMarch Source: Analysis of sampled English- language tweets from Jan. 28-29, 2017 (Emojipedia)
- 35. The media doesnât always appreciate emojis
- 36. 36
- 37. 37
- 38. 38
- 39. Latest work
- 40. The Top Emojis of #MeToo Source: Analysis of 28,629 English-language tweets sampled on Oct. 16, 2017 using the Twitter Search API. Top 10 emojis shown in descending order of frequency.
- 41. The Top Emojis of #MeToo Source: Analysis of 28,629 English-language tweets sampled on Oct. 16, 2017 using the Twitter Search API. Top 20 emojis shown, grouped by emoji type, in descending order of frequency. â¤đđđđđđđ đđ˘đĄđđđđĽ đâđŞđđ
- 42. The Top Emojis of #MeToo: Skin Tone Distributions Source: Analysis of 28,629 English-language tweets sampled on Oct. 16, 2017 using the Twitter Search API. Color 1 Color 2 Color 3 Color 4 Color 5 Mean Skin Tone â 9% 37% 29% 14% 11% 2.8 đ 29% 29% 18% 12% 12% 2.5 đ 20% 50% 15% 15% 0% 2.2 đŞ 38% 28% 25% 9% 0% 2.0 đ 50% 25% 25% 0% 0% 1.8 đ 50% 36% 14% 0% 0% 1.6 For emojis that are most frequently used with skin tone modifiers, we look at the distribution of skin tone usage for each emoji. The yellow band indicates the âmean skin toneâ for a given emoji.
- 43. While emojis might seem fun and trivial, they are at their core vessels for conveying human emotion. Emojis are also data. And data science can help us understand them.
- 44. You too can become an emoji data scientist! https://prismoji.com/2017/02/06/emoji-data-science-in-r-tutorial/
- 47. Next steps â Journalism vs. Academic & Industry â Study emoji usage across countries & cultures â Study private emoji usage vs. public emoji usage â Using this as a kick-off point to dive head-in into data journalism and storytelling about unexplored questions in society and culture
- 48. Thank you! Send me questions and tips! Or reach out if youâd like to collaborate J On Twitter: @hamdanazhar Or by email: hamdan@prismoji.com Check out more emoji data science at prismoji.com