Slashdot Mirror
Measuring Real Time Public Opinion With Twitter
Hugh Pickens writes "The NY Times reports that statisticians from the University of Vermont are hoping to harness the stream of messages flowing through Twitter to read public opinion and sentiment in real time. '"Twitter is a reflection of what people are interested in right now," says Peter Dodds, adding that the goal is to establish an index, akin to the Dow Jones industrial average, that can "give an overall sense of how a collective body of people are feeling at any given point in time.' Dodds says he and his colleagues are analyzing about 1,000 tweets each minute, or about a million a day, looking for trends in descriptive words and phrases that indicate moods and emotions. In addition, the two can monitor the public reaction to news or policy announcement and track it over time. The tool is still in its early stages, but eventually Dodds hopes that it could work similarly to Google Flu Trends, a Web tool that doubles as an early-warning system for flu outbreaks by detecting spikes in certain search terms. Since relationships and conversations are so intrinsic to how people communicate on Twitter, the researchers hope that observing how one user's mood is affected by another might shed some light on crowd behavior and emotional contagion. 'All of this data serves as a remote sensor of well-being,' Dodds says."
Wait until the astroturfers and lobyists discover this.
I just wrote on twitter about my opinion about harnessing the stream of messages flowing through twitter to read public opinion and sentiment in real time.
Everything I write is lies, read between the lines.
Only 98% of Twitter updates are "pointless babble," says a new report that studied 2,000 tweets over a period of two weeks.
The top category was "pointless babble" tweets, with nearly 98% of tweets being inanity no sane person could want to read, retweets of inanity, links to inanity, retweets of links to inanity and retweets of retweets of links to links to the reretweet itself. And camera phone pictures of bowel movements on Twitpic.
Almost 2% was Stephen Fry, Neil Gaiman or retweets thereof and the rest was Warren Ellis posting scatological abuse of his fans.
Botnet command messages were becoming more popular, many disguised as combinations of the syllables "lol" "wtf" "d00d" "RT" and "#fb" or scatological abuse of Warren Ellis's fans.
Twitter's demographics as of June 2009 were 55% female, 43% ages 18 to 34, 78% white, and 99.5% of such short attention spans that Facebook might as well be War and Peace. Botnet readership was considered likely to rise as soon nothing with organic intelligence would be able to cope.
Twitter recently redesigned its homepage, changing the tag "What are you doing now?" to "Post tomorrow's CNN headlines, particularly about #goatse."
http://rocknerd.co.uk
Twitter goes down a lot now (or at least partially). For example, I cannot post at this moment. It's going to be hard to base serious programs on a service that is down so frequently.
Comment removed based on user account deletion
"Twitter is a reflection of what people are interested in right now"
=>
"Twitter is a reflection of what the twits are twatting in right now"
Can you see the problem?
Or the collected opinions of twits, er twitterers, twats, or whatever. It's a self-selected group, whose collective opinions are no more representative of the general public (or voters or any other subset of the general public) than, say, the opinions of slashdotters. And although there is much drivel on slashdot, I suspect it is nothing compared to the twaddle on twitter.
Those who can make you believe absurdities can make you commit atrocities. - Voltaire
Assuming that "normal" people behaves and think about subjects like the average high activity twitter user in all cases is a somewhat a risky choice, throwing dices could be more exact.
People who use twitter and read tweets should be gathered in a room and be shot.
Great, more and more people are beginning to believe that Twitter is brain flu! "The tool is still in its early stages, but eventually Mr. Dodds hopes that it could work similarly to Google Flu Trends, a Web tool that doubles as an early-warning system for flu outbreaks by detecting spikes in certain search terms."
Twitter is a reflection of what people are interested in right now
Correction: Twitter is a reflection of what morons are interested in right now. Still, useful marketing information.
Negative moral value of force outweighs the positive value of good intentions.
I hear you like twiting, so we built a twiter with twats built in so that you can twit while something or ohter.
Fuck systemd. Fuck Redhat. Fuck Soylent, too. Wait, scratch the last one.
Another case of life imitating Onion.
I, for one, am completely fed-up with polling agencies and the media trying to front-run public debate. We dont allow ballot stuffing, and we should not allow twitter stuffing either, though I do like the sound of that.
The fact that we have endless Astroturfing here tells you that this is a very bad idea. We will soon have the cable news media reporting this crap and the sheeple will lap it up, AGAIN.
The DOJ would be better looking at this than writing amicus curia brief supporting the MPIA/RIAA.
Twitter sucks and twitter users need to get a life.
This seems open to gaming the system since there are no controls on the users and Twit-bots are easily possible. Given that politicians are so d@mn poll-driven these days, this idea seems dangerous.
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
Replying to remove an improper mod of this comment.
VLC FOR MAC IS DYING! IF YOU DEVELOP, PLEASE SAVE IT!!
http://www.mediaite.com/tv/the-daily-show-explains-how-fox-self-pollinates-outrage/>the daily show is already outlining this happening on fox news.
VLC FOR MAC IS DYING! IF YOU DEVELOP, PLEASE SAVE IT!!
slashdot seems to have improperly published the link?.. bug anyone?
http://www.mediaite.com/tv/the-daily-show-explains-how-fox-self-pollinates-outrage/
VLC FOR MAC IS DYING! IF YOU DEVELOP, PLEASE SAVE IT!!
Webtrends has a product that provides social measurement via blog posts and tweets and bring it all together to tell you the overall web "buzz". There's sweet twitter feeds of everyone that mentions you.
about 1,000 tweets each minute, or about a million a day
Either someone really sucks in maths, or there is only 1000 mins in a day....
The ABC (Australian Broadcast Commission) did this for a live state -vs- state football match we have here. It was quite good. Measuring sentiment of who would win based on the positiveness or negativeness of 'tweets' (or whatever the fuck they're called).
Here
.
I'm reminded of watching a documentary on ex-pres Reagan's speeches being realtime tracked with "satisfaction knobs" by focus group members, allowing specific words and phrases to be chosen for future speeches for certain demographics. Reagan said "A thousand stars", Bush Sr said " A thousand points of light" and so on.
/quick patent this in my name please. available for business model consultation.
So that Selection Bias can equate to a target demographic.
I expect a pay per tweet model to be offered as an incentive, for instance as a rebate against cell phone charges within certain demographics, at certain times ie elections or by "lottery", soliciting participation with a text message.
bug anyone?
No. As you can see in your OP, there is a closing angle bracket after your link:
If you look at the source code, you see this:
Notice the href="a". This means you somehow mangled the HTML link. The "a" is probably from your link text; did it start with a lowercase "a"?
It happens a lot. Forget a quote here, confuse the order of punctuation, it could be a lot of reasons. That's why there is a preview function. Why didn't you use it?
Holy batman, therefore from now on we can expect politicians to become Twitter sock puppets, no longer poll slaves?
Catalin Braescu
Ofaly.com
'"Twitter is a reflection of what people are interested in right now," So true. The world needs to know about #MrsSlocombesPussy, um, like, RIGHT NOW.
Is it really a true sample of PUBLIC opinion when only a limited portion of the public actually uses Twitter?
For the most part twitter is a measure of what a vanishingly tiny and very niche chunk of the population incorrectly believes to be interesting.
...breach of privacy?
TFA is misleading, this time due to the researchers trying to make their science sexy when it's really just trying to ride a social networking wave. There is no "now" involved. They are going to scan across time looking for posts on a particular subject to see what peoples "reactions" are. People don't post these millions per day simultaneously, and they may think for an hour before they post or the may post immediately after a particular thing of interest occurs.
The 'self-selection' bias is true but isn't nearly the issue people assume. What is there about being a twit that makes people think the same about a particular thing? The evidence is in the variance of opinion measures as compared to ... what? People who are not twits (group) who'll agree to participate (subgroup) in a study comparing their opinions (subsubgroup), with those of twats (subsubsubgroup, plus forcing a result because they may never really have had such an opinion before being asked to here). All such studies have to contend with some biases such as self-selection. They have to report these as part of their operationalizations. As long as the statistical testing produces results that *could* be an accurate subset of the general population (it may just happen to be a result that looks like this, you can't tell) it can be considered a useful result. This depends on the statistical testing. And this project is being run by statisticians. I've no doubt they know what they're doing in this respect. Now when it comes to examining peoples' response to a tragedy, and finding they expect something worse to happen, well good luck on them finding the well understood basis for this and, failing to incorporate this, end up with a result that's way off compared to their others.
Why aren't they scanning the far broader user base of another system that allows the posting of more emotionally laden material by not restricting how much they post? Why aren't they using a collection of posts that cover decades instead of months? Because they're relying on making their science sexy to make it relevant and noticed. Any grouping creates bias, so the larger grouping makes better science, but Usenet isn't sexy, and Twatterizing is. At least this week. There's another problem. Twits move from one social networking site to another as each becomes the next best thing. Twitter will become obsolete, and so will their results.
"I may be synthetic, but I'm not stupid." -- Bishop 341-B
That will be fucking hilarious to watch!
Obama, 10:27am: Shut down the investigation into Black Panther voter intimidation. ...
Obama, 10:28am: Reopen the investigation into Black Panther voter intimidation.
Obama: 10:29am: Shut down the investigation into Black Panther voter intimidation.
Won't work: people use opposite meanings, such as bad=good. Impossible to automate public opinion this way.