To examine the link between human language and social network structure, a team of mathematical biologists analyzed 75 million tweets from 250,000 users.
The researchers, led by Princeton’s Sebastian Funk, located several tight-knit communities (and subcommunities) that tweet far more heavily to each other than to the rest of Twitter and revealed a slew of communities in which members shared a common dialect, ScienceNOW reports.
Besides identifying social groups, the team says applications of their method include: customizing online experiences, targeted marketing, and crowd-source characterization.
Pictured above are some communities sharing a special lingo and some top-ranked words. (Rather than apply a label to these communities, the researchers identified them by numbers, which you can see here.)
Some examples of shared words in communities with more than 250 users:
By using each tweeter's unique set of words, the researchers predicted their chosen communities correctly about 80% of the time -- suggesting that words help tweeters identify themselves as community members.
The work was published in EPJ Data Science last month.
Image: Bryden et al. EPJ Data Science
This post was originally published on Smartplanet.com