Wed September 4, 2013
Do You Have A Twitter 'Accent'?
With 500 million users and 500 tweets a day, the social networking site Twitter has changed the way we communicate. It also changes the way we write.
This year alone there were more than 100 Twitter-based studies. One study found that tweets often use words and spellings that are consistent with — and unique to — the user’s region, reflecting local accents and terminology.
Jacob Eistenstein, who was one of the researchers, says the team is continuing its research to find out how new words, terms and acronyms move between cities.
- Jacob Eisenstein, assistant professor in the School of Interactive Computing at Georgia Tech. He was part of the research team that produced the study.
ROBIN YOUNG, HOST:
It's HERE AND NOW.
And we all know that Twitter has changed the way we communicate with 500 million users, over 500 million messages a day. And tweets also have changed the way we write, forcing us to condense thoughts to 140 characters. But is Twitter also changing the way we speak? There have been more than 150 Twitter-based studies. One of them focuses on how tweets are more like speech than writing.
Jacob Eisenstein is assistant professor in the School of Interactive Computing at Georgia Tech and part of the research team studying how tweets are like speech. And he joins us in the studios of Georgia Public Radio in Atlanta. And, Jacob, we know one thing you're looking at is how regionalisms are moving around through Twitter. Is there a Twitter dialect or accent?
JACOB EISENSTEIN: So there's not just one Twitter accent, but many. This is one sort of popular idea about social media that there's a dialect of how you would write in social media. Some call it netspeak. But, in fact, what we've seen in the work that we've done with my colleagues at Carnegie Mellon and here at Georgia Tech is that the way that people write in social media, not only is it different from the way that they write otherwise, but it'll differ markedly across the United States, even more so than more traditional forms of written language would differ.
YOUNG: Give us some examples.
EISENSTEIN: So, you can think about a few different types. In some cases, it would appear that regional pronunciations, or pronunciations that reflect other types of dialect are getting transcribed into social media. So for example, a word like something could be written S-U-T-T-I-N. What happens is the G gets dropped and TH gets stopped. So you go from something to suttin, and that's something you can see written out. Another example, or another type of word is a word that we might know from spoken language, like hella, which means very in Northern California. Can I say hella?
YOUNG: Yeah, you can. You just did.
EISENSTEIN: We know from spoken language to expect it to be centered in Northern California. In fact, that's what we see in social media, too. And then the third type are words that really seemed to be social media specific, or at least specific to computer-mediated communication. These are words like abbreviations from multiword expressions, like, you know, LOL. LOL can be used across the United States. Well, there are alternative forms of LOL, some of which I couldn't spell out for you on the radio. Some of those are very, very geographically specific.
YOUNG: Well, you mentioned suttin. Time Magazine inaccurately attributed that to our region, up here in Boston. But you say that that actually - you could see that coming from New York. Another example, IKR, for I know, right.
YOUNG: Popular in Detroit?
EISENSTEIN: The Michigan area, yeah.
YOUNG: So you can see - you can sort of track some of these were actually regionalisms. What are some of the other things you're looking at? We know you have a big study coming out, and you're not prepared to fully divulge. But we get a hint in Time Magazine that you're looking at race. You might see a word move from a city that has a large black population to another city, let's say Memphis in Tennessee, that has a black - a large black population, but not move as quickly to Nashville, which has more of a white population, which tells you that words can move racially. How do you see all these tweets?
EISENSTEIN: So what we're trying to do with that study is - and this is some work in progress - we want to see how words spread between places. So we know that it's not just static. We know it's not just the case that a word just appears in one city and never - either never leaves, or continues to exist there without spreading to other parts of the U.S. And what we believe is that by studying the way that words move between cities, we can identify some of the demographic factors that really, that shape people's perceptions of culture and similarity.
So if here in Atlanta, you know, it was found that we are more likely to adopt words from, for example, Memphis than Nashville, that would tell you something about the relationship between, you know, Memphis and Atlanta. And then we'd want to look at the demographic characteristics of those cities to explain that relationship.
YOUNG: Mm-hmm. And you might see that Memphis and Atlanta have higher African-American populations than Nashville, and that might tell you something about the tweets going back and forth.
EISENSTEIN: Yeah. So we're trying to really aggregate information across many different words whose geographical trajectories we've been able to identify and try to see what's the high-level picture that comes out in terms of the transmission of linguistic change, at least in terms of words.
YOUNG: Yeah. Well, other researchers have looked at how men and women tweet. There's a difference. Women use the word I more. Men are more likely to send links or use numbers. They looked at the difference between older and younger Twitter users. Younger users are more likely to use all caps. Dutch researchers found that younger tweeters are more likely to be very expressive, stretch out words like nice to have 10 I's in it. But what else are you looking at when you look at how different regions might share words? What else does that tell you?
EISENSTEIN: To me, it really comes back to the sense of identity. So I think language - we know this with spoken language, it's really deeply connected to people's sense of personal identity. And there's a great line of research from social linguistics focusing on speech that's shown us again and again over the last few decades that generally was thought not to be true in written language.
And I think the reason is that, until recently, written language was very constrained. So you learned to write in school. You use writing for official documents, you know, writing to your attorney or something or writing, you know, in the context of your work. So written language wasn't really free to express personal identity in the way that spoken language has been throughout history.
Now, with the advent of social media and computer-media communication generally, that's changed, and people are able to use written language as a key component of their daily social interactions. And that's, I think, the reason we're seeing this kind of variation according to geography, of all the dimensions that you mentioned - geography, age, gender and so on.
YOUNG: Well, and in other words, people from Cleveland might never have had the reason to write a formal letter to someone in Boston or to speak to someone in Boston, but now they do, through Twitter. So whatever the language uses in Cleveland are, they're going to have a better chance of making their way to Boston than they ever did.
EISENSTEIN: Perhaps. So I think the premise of social media is that it could connect anybody anywhere. But the reality is that most people have social media networks that are actually incredibly constrained geographically. And I think if that wasn't true, it wouldn't be possible to even detect this kind of regional differences, because the change would spread so quickly.
What we're finding is that if you look over a period of two or three years - which is already perhaps half the life span of Twitter - you can see this kind of geographical differences. Some of them will change over that time period. Some of them won't. But people's networks seem to be compact enough that they can preserve some regional differences.
My personal favorite one is an abbreviation. This is - you know LOL. That stands for laugh out loud. So this one, I can't tell you what it stands for. It's CTFU. It's cracking up, and there's - you can...
YOUNG: Oh, boy. Yeah. Don't - please don't tell us what that stands for.
EISENSTEIN: So it's cracking up. It indicates that something's funny. The reason I like this one is that when we first came across this, it was 2010, and it was popular in Cleveland - where we now know it seems to have originated - Pittsburgh and Philadelphia. And I presented this work to a group of linguists, and this was the one example that really bothered them.
The reason is if you've spoken to people from these cities, you realize that the spoken dialects are completely different across these three cities. So someone from Cleveland really has a Great Lakes accent. Someone from Philadelphia emphatically does not, and Pittsburgh is doing its own thing.
So the idea that these three cities with three different spoken dialects would share something that the rest of the country did not share in terms of social media, I think, you know, reveals the fact that, in some ways, social media is going to parallel spoken language. But in other ways, it really reflects new kinds of identity and new kinds of groups.
YOUNG: Interesting. Jacob Eisenstein, assistant professor in the School of Interactive Computing at Georgia Tech. Thanks so much.
EISENSTEIN: Thank you.
YOUNG: So, Jeremy, it sounds as if Twitter is spreading regionalisms across state and city lines, but they're still staying within social and demographic groups.
JEREMY HOBSON, HOST:
Yeah. You also hear about this phenomenon, Robin, with television and the idea that maybe some of the regional accents that we have around this country are disappearing because so many of the people that we watch on TV are from California.
HOBSON: And so maybe we're all going to start to sound like we're from Southern California.
YOUNG: And will social media mean more of that? Well, we just know that if you start saying wicked or the more forceful wicked pissah, it's not a bad word, parents, but it did come from here. Oh, and my tweets tomorrow will come from Mississippi. Going to visit listeners at Mississippi Public Broadcasting in Jackson. I cannot wait. I'm excited about that.
HOBSON: You'll probably have a very thick Southern accent on Twitter.
YOUNG: When I come back.
HOBSON: From NPR and WBUR Boston, I'm Jeremy Hobson.
YOUNG: I'm Robin Young. This is HERE AND NOW. Transcript provided by NPR, Copyright NPR.