Random and fickle as social networks may seem, the success of an individual Tweet or Facebook post can be hugely dependent on the way it's worded. Some sentences will work well, while others will quietly implode in a puff of indifference. However, researchers from Google and Cornell University now claim that they can predict this outcome using straightforward linguistic rules. The rules may not be especially practical when it comes to actually crafting a Tweet (the main tips are: make it sound like a news headline, follow "community norms," and add more detail where possible), but there does seem to be decent statistical evidence to back them up.
When the rules were applied to a real-world sample of Twitter "twins" -- that is, two published Tweets by the same user, about the same subject, where the only big difference is in the wording -- they were 66 percent accurate in predicting which twin got more re-tweets. By contrast, average human readers were only 61 percent accurate in spotting the better Tweet, which backs up the idea that social networking success is more about science than about emotion. Then again, the remaining 34 percent of inaccurate predictions delivered by the researchers' model means there's still a whole lot of literary flair which remains unaccounted for, and which probably can't be condensed to a simple set of rules.
If you reckon you know what a good Tweet looks like, to the point where you might be able to beat the human average and the computer model, then the researchers have built an addictive little quiz to give you the chance to prove your skills. (Just to get things going, this author scored 70 percent, after ten trials. Beat that, ppl.)
Via: Washington Post
Source: Chenhao Tan et al. [PDF]
0 comments:
Post a Comment