Want to be Retweeted?Large Scale Analytics on Factors Impacting Retweet Rate in Twitter Network<br />Bongwon Suh, Lichan H...
Research Motivation<br />How and why certain information spreads more widely than others in Twitter network?<br />
Followee and Follower<br />
Retweet and Information Diffusion<br />Kwak et al. WWW10<br />
How to Retweet<br />Old Style<br />
How to Retweet<br />New Style<br />Feature Retweet<br />
Research Question<br />Which tweet features are associated with retweet?<br />Retweet Model<br /># Retweet ~ function(f1, ...
Features of Tweets<br />Content features<br />URL<br />Hashtag<br />Mention<br />Contextual features<br /># of followers<b...
Example Tweet<br />Content Features<br />Mention<br />Hashtag<br />URL<br /># Followees: 395<br /># Followers: 1,400<br />...
Data Collection<br />74M tweets from Twitter Stream API<br />Characterization<br />2~3 % sample<br />Incomplete retweet in...
Two Analyses<br />Exploratory Analysis <br />Principle component analysis (PCA)<br />Retweet + 8 features<br />Examine how...
PCA Factor Map<br />Content Factor<br />
PCA Factor Map<br />Content Factor<br />Contextual Factor<br />
PCA Factor Map<br />Content Factor<br />Contextual Factor<br />
Generalized Linear Model<br />Content <br />Features<br />Contextual<br />Features<br />
OK. Having URL matters… Then?<br />All of them are equal?<br />What kind of content do people use?<br />Does retweeting va...
Calculating Retweet Rate for Domain<br />Group URLs by domain<br />475,509 tweets have URLs from www.youtube.com (e.g. htt...
Normalized Retweet Rate<br />Relative ratio of retweet rate compared to the global average<br />Global Average Retweet Rat...
URL vs. Retweet Rate<br />twitlonger.com<br />Normalized Retweet Rate<br />mashable.com<br />huffingtonpost.com<br />guard...
Varying Retweet Rate per Domain Type<br />Personal media ( < 0.15)<br />Justin.tv<br />twitcam.com<br />Trivia & Discovery...
Hashtag vs. Retweet Rate<br />#vouconfessarque<br />Normalized Retweet Rate<br />#idothat2<br />#jafizisso<br />#ohjustlik...
Patterns in Hashtag<br />Personal media<br />#nowplaying: 0.75<br />Social meme<br />#idothat2: 7.6<br />#ohjustlikeme: 7....
Retweet Rate vs. Contextual Features<br /># of followers<br /># of followees<br />Account seniority (days)<br /># of past ...
Follower vs. Retweet Rate<br />Normalized Retweet Rate<br /># of Follower <br />
Followee vs. Retweet Rate<br />Normalized Retweet Rate<br /># of Followee<br />
Age of Twitter Account vs. Retweet Rate<br />Normalized Retweet Rate<br />Days<br />
Past Tweets vs. Retweet Rate<br />Normalized Retweet Rate<br /># of Past Tweet<br />
Want to be Retweeted - Summary<br />Content<br />URL, Hashtag<br />Context<br />More # of followers, make many friends<br ...
Thank You!<br />Want to be Retweeted?Large Scale Analytics on Factors Impacting Retweet Rate in Twitter Network<br />Bongw...
Upcoming SlideShare
Loading in...5
×

Want to be Retweeted? Large Scale Analytics on Factors Impacting Retweet Rate in Twitter Network

5,074

Published on

Presented at SocialCom 2010 conference Aug 21st 2010

Published in: Technology
1 Comment
4 Likes
Statistics
Notes
No Downloads
Views
Total Views
5,074
On Slideshare
0
From Embeds
0
Number of Embeds
4
Actions
Shares
0
Downloads
0
Comments
1
Likes
4
Embeds 0
No embeds

No notes for slide
  • Tweet about“airfrance flight” June 1st 2009
  • rank DOMAIN Retweet Rate1 http://twitpic.com 1.465929872 http://myloc.me 2.0522640113 http://www.facebook.com 1.0285236584 http://www.youtube.com 1.4980595675 http://formspring.me 0.0505511466 http://www.twitlonger.com 6.0643892537 http://tweetphoto.com 1.7269902328 http://youtu.be 0.342674039 http://twitcam.com 0.12286631210 http://twitter.com 2.43754964511 http://www.plurk.com 0.086860112 http://fun140.com 0.03234741213 http://www.formspring.me 0.15897076314 http://foursquare.com 0.17509565215 http://www.ustream.tv 1.0249179716 http://tinychat.com 0.08524135517 http://blip.fm 0.39707552918 http://www.funwebsites.org 0.01857939419 http://www.flickr.com 1.4448898720 http://king-soukutu.com 0.31556806321 http://mashable.com 3.64777737322 http://twittascope.com 0.04802722723 http://pollpigeon.com 0.08245854324 http://www.nicovideo.jp 0.54207940725 http://news.bbc.co.uk 1.5088626826 http://friendfeed.com 0.43967589727 http://www.askbiography.com 0.05334752928 http://www.etsy.com 1.92588300629 http://www.nytimes.com 2.58635696730 http://digg.com 1.50888092531 http://oohja.com 0.13218834332 http://www.amazon.co.jp 0.51427291333 http://news.yahoo.com 0.58829323634 http://links.assetize.com 0.2507329635 http://meadd.com 0.44673997736 http://www.cnn.com 2.93557682737 http://activities.myspace.com 0.1135431938 http://itunes.apple.com 1.9515056939 http://www.myspace.com 1.1631308640 http://www.epicpetwars.com 0.00206165741 http://tweetmyjobs.com 0.06686426442 http://www.freelancer.com 0.07898495843 http://www.huffingtonpost.com 2.66464827444 http://www.guardian.co.uk 2.7082061545 http://roflquiz.com 0.05226299846 http://ameblo.jp 0.94685734147 http://techcrunch.com 2.48535160548 http://twitgoo.com 0.89076354249 http://www.orkut.com.br 0.98507791650 http://dezireweb.info 0.05667828251 http://cgi.ebay.com 0.37291388252 http://ardenal.info 0.06602368253 http://trtools.com.br 0.02573072154 http://blog.livedoor.jp 1.10301649855 http://www.google.com 1.21834285556 http://www.engadget.com 1.325103232
  • 1 #nowplaying 355147 29846 0.7539168112 #ff 224760 62331 2.4878862183 #jobs 124728 2173 0.1562936064 #fb 87959 10994 1.12129765 #tinychat 67225 273 0.0364315196 #vouconfessarque 51578 43628 7.588330717 #fail 49248 9759 1.7777151078 #tcot 47394 18527 3.5069306729 #1 47373 9124 1.7278253110 #followfriday 39986 11170 2.50605532611 #news 38573 3630 0.84424529912 #shoutout 30633 928 0.27177148613 #tweetmyjobs 30594 165 0.04838303714 #bbb 28590 5877 1.84411066115 #haiti 28563 13829 4.34342573216 #letsbehonest 27926 7214 2.31746320517 #iranelection 27611 12334 4.00744205318 #quote 27541 11475 3.73782041919 #followmejp 25940 4494 1.55420585120 #follow 24166 8084 3.00100678221 #musicmonday 22044 2881 1.17246071322 #oscars 21843 4235 1.7393483623 #epicpetwars 21740 6 0.00247592324 #2 21157 2686 1.13893087325 #petpeeve 20959 6015 2.5746044626 #imattractedto 20606 5415 2.35749189727 #olympics 20514 3832 1.6757938528 #random 20412 4873 2.14168845229 #ohjustlikeme 20332 16531 7.29397831830 #mm 20141 3174 1.41374582731 #ipad 19866 4032 1.82077187132 #retweetthisif 19832 12602 5.70057159833 #idothat2 19632 16583 7.57781500134 #maisfollowers 19545 149 0.06839054135 #p2 19403 8719 4.03128253636 #bbb10 18039 3710 1.84504470837 #job 17625 935 0.47591339538 #iphone 17211 3241 1.68934507339 #terremotochile 16980 8892 4.69793724640 #jafizisso 16859 15564 8.28199291541 #lost 16387 2492 1.36425020442 #nicovideo 15217 953 0.56183566343 #music 14172 1732 1.09638273344 #apple 13867 2611 1.68915615345 #3 13796 1195 0.77707003646 #freevenezuela 12544 4708 3.36702120947 #90stweet 12485 2696 1.93721036648 #van2010 12484 2860 2.05521715549 #saints 12281 3500 2.5566999150 #imthetypeto 12169 2725 2.008894171
  • 1 #nowplaying 355147 29846 0.7539168112 #ff 224760 62331 2.4878862183 #jobs 124728 2173 0.1562936064 #fb 87959 10994 1.12129765 #tinychat 67225 273 0.0364315196 #vouconfessarque 51578 43628 7.588330717 #fail 49248 9759 1.7777151078 #tcot 47394 18527 3.5069306729 #1 47373 9124 1.7278253110 #followfriday 39986 11170 2.50605532611 #news 38573 3630 0.84424529912 #shoutout 30633 928 0.27177148613 #tweetmyjobs 30594 165 0.04838303714 #bbb 28590 5877 1.84411066115 #haiti 28563 13829 4.34342573216 #letsbehonest 27926 7214 2.31746320517 #iranelection 27611 12334 4.00744205318 #quote 27541 11475 3.73782041919 #followmejp 25940 4494 1.55420585120 #follow 24166 8084 3.00100678221 #musicmonday 22044 2881 1.17246071322 #oscars 21843 4235 1.7393483623 #epicpetwars 21740 6 0.00247592324 #2 21157 2686 1.13893087325 #petpeeve 20959 6015 2.5746044626 #imattractedto 20606 5415 2.35749189727 #olympics 20514 3832 1.6757938528 #random 20412 4873 2.14168845229 #ohjustlikeme 20332 16531 7.29397831830 #mm 20141 3174 1.41374582731 #ipad 19866 4032 1.82077187132 #retweetthisif 19832 12602 5.70057159833 #idothat2 19632 16583 7.57781500134 #maisfollowers 19545 149 0.06839054135 #p2 19403 8719 4.03128253636 #bbb10 18039 3710 1.84504470837 #job 17625 935 0.47591339538 #iphone 17211 3241 1.68934507339 #terremotochile 16980 8892 4.69793724640 #jafizisso 16859 15564 8.28199291541 #lost 16387 2492 1.36425020442 #nicovideo 15217 953 0.56183566343 #music 14172 1732 1.09638273344 #apple 13867 2611 1.68915615345 #3 13796 1195 0.77707003646 #freevenezuela 12544 4708 3.36702120947 #90stweet 12485 2696 1.93721036648 #van2010 12484 2860 2.05521715549 #saints 12281 3500 2.5566999150 #imthetypeto 12169 2725 2.008894171
  • Want to be Retweeted? Large Scale Analytics on Factors Impacting Retweet Rate in Twitter Network

    1. 1. Want to be Retweeted?Large Scale Analytics on Factors Impacting Retweet Rate in Twitter Network<br />Bongwon Suh, Lichan Hong, Peter Pirolli, and Ed Chi<br />suh@parc.com<br />@billsuh<br />Augmented Social Cognition Research Group<br />(Xerox) PARC<br />
    2. 2. Research Motivation<br />How and why certain information spreads more widely than others in Twitter network?<br />
    3. 3. Followee and Follower<br />
    4. 4. Retweet and Information Diffusion<br />Kwak et al. WWW10<br />
    5. 5. How to Retweet<br />Old Style<br />
    6. 6. How to Retweet<br />New Style<br />Feature Retweet<br />
    7. 7. Research Question<br />Which tweet features are associated with retweet?<br />Retweet Model<br /># Retweet ~ function(f1, f2, …., fn), where fiare simple features extracted from a tweet<br />Data Driven<br />
    8. 8. Features of Tweets<br />Content features<br />URL<br />Hashtag<br />Mention<br />Contextual features<br /># of followers<br /># of followees<br />Account seniority (days)<br /># of past tweets (or statuses)<br /># of favorite items<br />Retweet<br />E.g. Feature retweet, RT: @username, Via: @username, etc<br />Image: http://astore.amazon.com/buy.victorinox.swiss.army.knives-20/<br />
    9. 9. Example Tweet<br />Content Features<br />Mention<br />Hashtag<br />URL<br /># Followees: 395<br /># Followers: 1,400<br /># Favorite: 1,657<br /># Day: (since June 17, 2008)<br /># Past tweets: 21,000<br />Contextual Features<br />User “shefaly”<br />(through Twitter API)<br />
    10. 10. Data Collection<br />74M tweets from Twitter Stream API<br />Characterization<br />2~3 % sample<br />Incomplete retweet information<br />10K tweets with accurate # of retweets for each of them<br />Hadoop / Hbase / MapReduce <br />Image: http://covers.oreilly.com/images/9780596154622/lrg.jpg<br />
    11. 11. Two Analyses<br />Exploratory Analysis <br />Principle component analysis (PCA)<br />Retweet + 8 features<br />Examine how they are related<br />Retweet Model<br />Generalized Linear Model (GLM)<br /># retweets as a linear sum of function(feature)<br />
    12. 12. PCA Factor Map<br />Content Factor<br />
    13. 13. PCA Factor Map<br />Content Factor<br />Contextual Factor<br />
    14. 14. PCA Factor Map<br />Content Factor<br />Contextual Factor<br />
    15. 15. Generalized Linear Model<br />Content <br />Features<br />Contextual<br />Features<br />
    16. 16. OK. Having URL matters… Then?<br />All of them are equal?<br />What kind of content do people use?<br />Does retweeting vary depending on types of information?<br />Further analysis<br />Content features<br />Contextual features<br />Image: http://techfilipino.com/how-much-is-my-website-domain-name<br />
    17. 17. Calculating Retweet Rate for Domain<br />Group URLs by domain<br />475,509 tweets have URLs from www.youtube.com (e.g. http://www.youtube.com/watch?v=xyz1234)<br />Unshortening 15.6M+ URLs!<br />Identify retweets<br />79,404 of them are in retweets.<br />Reweet Rate for Youtube.com<br />79,404 (in retweets) / 475,509 (in all tweets) = 16.6%<br />
    18. 18. Normalized Retweet Rate<br />Relative ratio of retweet rate compared to the global average<br />Global Average Retweet Rate<br />8.24M total retweet<br />73.9M total tweet <br />11.2% = 8.24M / 73.9M<br />Retweetablity<br />Normalized retweet rate for a domain<br />Retweet rate for youtube.com: 16.6%<br />Youtube.com: 1.50 = 16.6% / 11.2% <br />
    19. 19. URL vs. Retweet Rate<br />twitlonger.com<br />Normalized Retweet Rate<br />mashable.com<br />huffingtonpost.com<br />guardian.co.uk<br />techcrunch.com<br />cnn.com<br />nytimes.com<br />twitter.com<br />twitpic.com<br />Popularity Rank<br />formspring.me<br />twitcam.com<br />fun140.com<br />foursquare.com<br />tweetmyjobs.com<br />
    20. 20. Varying Retweet Rate per Domain Type<br />Personal media ( < 0.15)<br />Justin.tv<br />twitcam.com<br />Trivia & Discovery ( > 4.0)<br />Omg-facts.com<br />Holykaw.alltop.com<br />News media with High Retweet Rate ( > 2.5)<br />Mashable.com<br />TheOnion.com<br />NYTimes.com<br />News Aggregator ( < 0.6)<br />news.yahoo.com<br />news.google.com<br />
    21. 21. Hashtag vs. Retweet Rate<br />#vouconfessarque<br />Normalized Retweet Rate<br />#idothat2<br />#jafizisso<br />#ohjustlikeme<br />#retweetthisif<br />#terremotechile<br />#iranelection<br />#haiti<br />#quote<br />#tcot<br />#followfriday<br />#ff<br />#fail<br />#nowplaying<br />Popularity Rank<br />#jobs<br />#tinychat<br />#fb<br />
    22. 22. Patterns in Hashtag<br />Personal media<br />#nowplaying: 0.75<br />Social meme<br />#idothat2: 7.6<br />#ohjustlikeme: 7.3<br />#fail: 1.78<br />#haiti: 4.3<br />#iranelection: 4.0<br />#tcot: 3.5<br />Being Social <br />#ff: 2.5, #followfriday:2.5, #follow: 3.0<br />Spam / Robots<br />#epicpetwars : 0.002<br />#tinychat: 0.03<br />#tweetmyjobs: 0.05<br />
    23. 23. Retweet Rate vs. Contextual Features<br /># of followers<br /># of followees<br />Account seniority (days)<br /># of past tweets (or statuses)<br /># of favorite items<br />
    24. 24. Follower vs. Retweet Rate<br />Normalized Retweet Rate<br /># of Follower <br />
    25. 25. Followee vs. Retweet Rate<br />Normalized Retweet Rate<br /># of Followee<br />
    26. 26. Age of Twitter Account vs. Retweet Rate<br />Normalized Retweet Rate<br />Days<br />
    27. 27. Past Tweets vs. Retweet Rate<br />Normalized Retweet Rate<br /># of Past Tweet<br />
    28. 28. Want to be Retweeted - Summary<br />Content<br />URL, Hashtag<br />Context<br />More # of followers, make many friends<br />Implication<br />Be an interesting person!<br />Up-to-date, opinionated, insightful, serendipitous, and/or funny information<br />
    29. 29. Thank You!<br />Want to be Retweeted?Large Scale Analytics on Factors Impacting Retweet Rate in Twitter Network<br />Bongwon Suh, Lichan Hong, Peter Pirolli, and Ed Chi<br />suh@parc.com<br />@billsuh<br />Augmented Social Cognition Research Group<br />
    30. 30.
    31. 31. Extra Slides<br />
    32. 32. Retweet Rate for URL Domain<br />
    33. 33. Retweet Rate for Hashtag<br />

    ×