Slashdot Mirror
Using Twitter Data To Approximate a Telephone Survey
cremeglace writes "A team led by a computer scientist at Carnegie Mellon University has used text-analysis software to detect tweets pertaining to various issues — such as whether President Barack Obama is doing a good job — and measure the frequency of positive or negative words ranging from 'awesome' to 'sucks.' The results were surprisingly similar to traditional surveys. For example, the ratio of Twitter posts expressing either positive or negative sentiments about President Obama produced a 'job approval rating' that closely tracked the big Gallup daily poll across 2009. The analysis also produced classic economic indicators like consumer confidence." By averaging several days' worth of tweets on presidential job approval, the researchers got results that correlated 79% with daily Gallup polling. Lead researcher Noah Smith said, "The results are noisy, as are the results of polls. Opinion pollsters have learned to compensate for these distortions, while we're still trying to identify and understand the noise in our data. Given that, I'm excited that we get any signal at all from social media that correlates with the polls." Here is CMU's press release.
It seems that the age demographics of twitter users wouldn't be very representative of the population as a whole.
I'm guessing it will take no more than a month for a combination of "conservative" and "progressive" blogs to rev up their teams of dittoheads to start flooding Twitter with politically themed messages, thus totally skewing the results. Same principle as Google-bombing, I guess. As someone who already views Twitter as almost entirely content-free, I can't say I'm particularly dismayed by this possibility. . . but anything that encourages the self-absorbed political zealots of this country can't possibly be good.
Just like traditional pollsters, social media researchers will have to address how representative Twitter users are of the general population. And unlike telephone surveys, small groups of people can wildly skew the results of Internet data,
Yes I did STFA (Skim the fucking article).
It mentioned the two main problems I see with this, cheating the system and whether twitter really is a large enough sample and a random enough sample to be considered a viable alternative.
Twitter has a whole range of people who don't actually use the damned thing. As with any poll though, people are going to say that the minority polled is what everyone says.
"The American people want to do x! Our poll says 80% of the American people want it!" No. No it doesn't. It just means 80% of the people you polled want it.
I despise how easy it is to use statistics and polls to manipulate people.
I've collaborated on research using Twitter traffic as a predictor so I applaud their efforts, but a 79% correlation with telephone responses is not as high as it sounds. For example, the minimum acceptable correlation for interrater reliability is typically 80%.
Put simply, the Twitter data can only account for about two thirds of the variation in phone responses. That's useful but there's still a lot of unexplained variance -- we have a long way to go.
The noise coming from one group of twits is the same as the noise coming from another group of twits.
Film at 11.
Science advances one funeral at a time- Max Planck
As soon as something like this comes to light it's only a short matter of time until turfers screw it up. Turfers are like spammers, as soon as there's a new medium they abuse it into uselessness.
The preceding post was not a Slashvertisement.
As someone who spends a lot of their time working to uncover endogeneities in statistical analysis, I feel that analyzing Tweets will never be a viable measure of general American opinions.
Remember when The Literary Digest predicted Alf Landon would crush FDR in the 1936 presidential election based on a poll of its subscribers? Okay, you don't *remember* that, but you've probably heard of it. Same problem here.
The readers of Literary Digest were not representative of the average American in 1936.
The users of Twitter are not representative of the average American in 2010.
Twitter polling is no better than straw polling, which is usually worse than nothing.
Umm... from TFA:
Likewise, both the Twitter-derived sentiments and the traditional polls reflected declining approval of President Obama's job performance during 2009, with a 72 percent correlation between them.
Okay... not a great correlation, but let's continue....
But the researchers found that their sentiment analysis did not correlate as well with election polling during 2008. For instance, increased mentions of "Obama" tended to correlate with rises in Barack Obama's polling numbers, but increased mentions of "McCain" also correlated with rises in Obama's popularity.
WTF? Is all of this built on how many times "Obama" or "McCain" is uttered on Twitter? And, given the obvious skewed demographics on Twitter (i.e., younger people, which tended to poll way toward Obama), increased conversations about McCain probably were bad in general.
Well, how do they explain this? Ah, the next sentence....
Improved computational methods for understanding natural language, particularly the unusual lexicon of microblogs, will be necessary before Twitter feeds can be reliably mined to predict elections, the researchers concluded.
Ah yes, the "unusual lexicon of microblogs," which probably consisted of sentiments like "I luv Obama!" and "McCain too old - WTF?"
Perhaps if they bothered to measure more than "mentions" of a candidates name, the data might have some (albeit still vague) meaning...
If this is the best stuff from the study which they actually mention in a press release, how much crap results are they not reporting?
There weren't enough mentions of Apple either in the summary or the comments posted.
To have a right to do a thing is not at all the same as to be right in doing it