Catch up on stories from the past week (and beyond) at the Slashdot story archive


Forgot your password?
Twitter Social Networks

Data Mining Reveals How Wording Influences Tweet Propagation 21

KentuckyFC (1144503) writes "One of the most widely shared tweets in history is Obama's "Four more years", posted after his second presidential election victory and currently retweeted 775,000 times. But how would different wording have influenced this tweet's popularity and the way it spread? It's easy to imagine that there's no way of telling what might have been in such an alternative universe. But a surprising phenomenon on Twitter has allowed data scientists to study this kind of alternative reality and work out the factors that make one tweet more popular than another. It turns out that the twitter stream contains a surprisingly large number of tweets from the same authors, pointing to the same content but with different messages. That's a natural experiment in which factors such as the author, the URL, the number of followers and so on are all held constant while the message varies. By studying these pairs of tweets, researchers can measure how well each performs and then determine which factors contribute to their popularity. These turn out to be things like the amount of information the tweet contains, the language it uses and even whether it includes a request for a retweet. The team has developed an algorithm that predicts which of a pair of tweets is more likely to be successful with greater accuracy than humans. And they've even set up a website where anybody can test their tweet-rating ability and thereby improve their chances of writing the perfect tweet."
This discussion has been archived. No new comments can be posted.

Data Mining Reveals How Wording Influences Tweet Propagation

Comments Filter:
  • Someone comes along and claims to have the perfect formula for writing a book, or a song, or music.

    They distill it down to its dreary essence, optimize based on what the focus groups say, and then produce utter dreck.

    Some of the best movies, books, and songs would NOT have passed through these design by committee things. And many more which do pass through these things should have never seen the light of day.

    Every time I see one of these things I think "OK, we'll see everything done like this for a while,

    • There's another value of this research: if you take the test, you can see how attuned you are to social media (group) thinking. I took the test and selected answers that I genuinely thought were more interesting for retweeting, and I got success rate of less than 30% -- worse than chance. That tells me if I had to do a social marketing campaign, I better not do it myself.

    • by Anonymous Coward

      Their are formulas for successful books, music, and if you hear the shit they play on radio, or go out and read romance novels, or any genre it is all predictable. I really have no idea what your talking about? Those standards or formulas have been in place for years, how many books have been rejected from publishers, how many bands have been rejected by labels, how many big production studios turned down a script, all because they didn't fit their formula. That includes idiotic censorship, 'this books cont

    • by TheLink ( 130905 )

      I think some of the stuff works. But many Hollywood films fail because the people making them don't care or have other agendas. Many movie makers live in a "different world" and are not in touch - for example: [] []

      There are formulas and critics say too many movies nowadays are following the same formulas too strictly. []

      It's not necessary to follow the formula that strictly for success: []

    • If this shit worked, there wouldn't be huge Hollywood films which fall on their face because nobody is interested. All it really does it make lowest common denominator stuff which nobody actually likes.

      Yeah, and even if "these things" did work sometimes, there's little evidence here that this particular model is any good.

      From TFA:

      They found that humans successfully pick the more popular phrasing with an average accuracy of 61.3 per cent. âoeNot that high, but better than chance, indicating that it is somewhat possible for humans to predict greater message spread from different deliveries of the same information,â say Tan and co.


      [T]his algorithm searched for various phrases, positive or negative sentiment, requests to share and so on. And this algorithm achieved the success rate of about 66 per cent, somewhat better than humans.

      Seriously? They were comparing pairs of tweets. This sounds like a barely significant improvement over human intuition: the computer is only wrong 34% of the time on a true/false test, while humans are wrong 38-39%. Simply guessing randomly would be a 50% error rate. Doesn't sound very predictive to me.

      I'm not saying the model isn't a statistically significant improvement

    • by Bob9113 ( 14996 )

      If this shit worked, there wouldn't be huge Hollywood films which fall on their face because nobody is interested. All it really does it make lowest common denominator stuff which nobody actually likes.

      You may be right about the objective quality of the product, but if this stuff didn't work, the lowest common denominator drek that nobody actually likes wouldn't sell so fast. You can't use this stuff to write Milton, but you can use it to sell the shit out of whatever syrupy self-help book Oprah is hawking

  • by nimbius ( 983462 ) on Thursday May 15, 2014 @10:24AM (#47008707) Homepage
    "Four years, more please"
    "remember when i killed an american in yemen because i can?"
    "Guantanamo...why does that sound familiar"
    "If uncle Joe says OK i guess gays are maybe kinda ok."
    "I swear to god if i hear benghazi one more time..."
    "I can haz budget?"
  • by Curunir_wolf ( 588405 ) on Thursday May 15, 2014 @10:26AM (#47008723) Homepage Journal

    As a Social Media Strategy consultant, this information is very useful to me. I will study it closely.

  • by Anonymous Coward

    Apparently, I am in a minority in that I do not reflexively favor excessive use of capital letters and exclamation marks. I also tend to pick the ones that get to the point (I had not realized it was possible to ramble when limited to 140 characters), which also makes me unpopular.

    This tells me that my very low opinion of Twits was still based on an excess of optimism, I am correcting that error in my judgment.

  • by kruach aum ( 1934852 ) on Thursday May 15, 2014 @10:29AM (#47008745)

    The twitter stream
    a frog tumblr
    the sound of retweets

  • to tweet to perfect tweet.

  • The "compare two tweets" functionality is broken.
    Submitting the form leads to a 403 Forbidden error.

    • by Piata ( 927858 )
      As is the quiz. Either that or I'm really bad at picking good tweets as I'm 0/20.
  • by CambodiaSam ( 1153015 ) on Thursday May 15, 2014 @10:42AM (#47008861)
    Click on the ad to see the trick they don't want you to know about...
  • by mysterons ( 1472839 ) on Thursday May 15, 2014 @01:36PM (#47010441)
    We did a study on predicting when a tweet would be retweeted (this paper cites us). The dominant factor is not what you write, but how many followers you have.

    Basically, a famous person can write anything and it will be retweeted. An unknown person can write the same tweet and it will be ignored.

    Link to paper:

    Sasa Petrovic, Miles Osborne and Victor Lavrenko. RT to win! Predicting Message Propagation in Twitter. ICWSM, Barcelona, Spain. July 2011. []

  • Fuck Tweets.

In less than a century, computers will be making substantial progress on ... the overriding problem of war and peace. -- James Slagle