What's the buzz? Text analysis technology tracks who's saying what about whom

If you love it when the elite pundits are proved wrong and the instincts of the common man — and common blogger — are proved right, Wonkosphere.com can plug you into a higher state of political awareness. The Web site, created by Kevin Dooley from the W. P. Carey School of Business and Steven Corman from ASU's Hugh Downs School of Communication, monitors feeds from 1,200 liberal, conservative and independent blogs, keeping a finger on the pulse of the 2008 election.

If you love it when the elite pundits are proved wrong and the instincts of the common man — and common blogger — are proved right, Wonkosphere.com can plug you into a higher state of political awareness. Wonkosphere is a Web site created by Kevin Dooley from the W. P. Carey School of Business and Steven Corman from ASU's Hugh Downs School of Communication.

It monitors feeds from 1,200 liberal, conservative and independent blogs, keeping a finger on the pulse of the 2008 election. The site measures how much candidates are being blogged about and what is being said about them. At its core, Wonkosphere uses Dooley and Corman's network text analysis technology called Centering Resonance Analysis (CRA). Developed at Arizona State University, the technology is licensed to their company, Crawdad Technologies, which hosts Wonkosphere.

How it works

Subscribed blog feeds are automatically pushed to the Wonkosphere computer. Then, CRA scans the text to reveal the tone of what is being said. Dooley writes about the results at wonkoblog.com. According to the Crawdad Web site, CRA "applies natural language processing techniques to convert blog texts into networks, then analyzes these networks to determine the relative importance of all the words in the text."

As a result, the site automatically determines "how much any given post is about a particular candidate, and how positive or negative the writing is." Centering Resonance Analysis has been shown in research to represent the "collective consciousness" of how an audience interprets a text, and has been shown to be more accurate than simple "word counting" methods. Crawdad's development of CRA has been funded, in part, by the Air Force Office of Scientific Research.

"We strive to keep a completely objective standpoint on the candidates," says Dooley, a professor of supply chain management at the W. P. Carey School of Business. "The blogs are opinion leaders. Once primaries start, voters pay less attention to experts and opinion leaders. They pay more attention to what their neighbors are saying in face-to-face conversations."

Wonkosphere features

There are four components to, Dooley explains.

• Buzz share: This bar graph, which greets viewers on the Wonksphere main page, features numbers that are similar to polling results, except the data are from the blogosphere, not from the general public. Buzz share shows who is being blogged about the most.

• Tone: This feature tells how positively or negatively people are talking about the candidate. For example, people liked former Arkansas Gov. Mike Huckabee's personality, but once they learned how the Republican stood on issues, after his Iowa Caucus victory, he lost popularity, Dooley notes.

  Conservatives ignored him for a month, until he won a series of states on Super Tuesday because of his enduring qualities as a warm speaker and a prototypical Southern gentleman. Former U.S. Sen. John Edwards had the most positive tone (although apparently not enough buzz among Democrats) throughout the campaign season; bloggers used terms associated with goodness or happiness.

• Buzz columns: This feature lists the most representative posts of the day. According to the site, "these are the posts that best represent the blog posts in general and are not necessarily the 'hottest' single issues being discussed at a given time." This differentiates Wonkosphere from aggregator site, like Digg.com, which uses readers votes to determine the most-read stories. "We don't go by popularity, we go by representativeness," he said.

  Using this feature, a campaign watcher could keep an eye on what people from other camps are thinking. Reading the same writers "constrains the universe of what you're exposed to," Dooley says. "There's a certain level of comfort in that, but it's constraining." Wonkosphere gives a "broader perspective on what's being said," Dooley says. It lets less visible blogs pop up on a given day, based on merit.

• One Step Ahead: This feature highlights blogs from previous days' posts that forecast what will turn up in other blogs or news reports the next day. A blog that makes "One Step Ahead" drives the conversation or stakes out a whole new ground.

A leading indicator

"Up until the primaries started, we were able to show changes in a candidate's buzz share numbers that would show up about three weeks later," Dooley says. For example, even before former Massachusetts Gov. Mitt Romney dropped out of the Republican race Feb. 7, Wonkosphere indicated that Arizona Sen. John McCain was in the driver's seat.

"There's almost no scenario for McCain to lose [the GOP nomination]," Dooley observed in an interview before Romney suspended his campaign and made McCain the presumptive nominee. "They've already thrown everything they can at him. The blogosphere and mainstream talk radio have already gone all-out this week [around Super Tuesday] and it didn't make a difference."

And Huckabee's endurance has not surprised Wonkosphere. He notes that the blogosphere — and the mainsteam media — both are less connected to the South than to the rest of the United States. "We picked up on Huckabee early on in August. I thought he had the best chance of any secondary candidate in both parties of breaking out of the pack," Dooley says. "I think he's still got a future ahead of him."

Dooley says that early in the primary process, McCain folks started to assume — incorrectly — that Wonkosphere was favoring Romney. "The bloggers for McCain were being too quiet," he says. The reaction: McCain supporters let it be known that it was their patriotic duty to get their blog signed up to Wonkosphere. McCain's buzz share increased somewhat.

As far as who the nominees will be, Dooley says, "The raw data is not any different from anybody else's perception, that McCain has it locked up and no one sees who will win for the Democrats. It depends on what happens day by day." Dooley is confident that Wonkosphere.com and Wonkoblog will be more accurate and timely than the mainstream media. "The data can only speak for itself through November," he says. "It will do that more quickly than we'll see public opinion change."

New territory

Wonkosphere is an outgrowth of Crawdad Technologies, which Dooley and Corman started in 2002. New features are on the way. Wonkosphere is planning blog postings by topic, in the spring. "We're already collecting some of that data," Dooley says. Already he knows that education is near the bottom of concerns in the blogosphere, where there is more talk about crime, although the two issues are interrelated.

Dooley also plans to reinstate the "Mud Meter" for 2008 to monitor negative language from the two campaigns' official Web sites. Releases from the Bush and Kerry campaigns were measured on a site attached to the Crawdad Technologies Web page. Wonkosphere was launched two months after Dooley and Corman started collecting data in June 2007.

"We get about 1,000 to 1,500 page-views a day," Dooley says. "We started off with a quarter or a third of that." Half of the page-viewers are regulars, half are people who happen upon the site. There are 10 to 20 blogs that get 100,000 or more page views a day. Examples are the Huffington Post and Townhall. The Wonkosphere partners hope eventually to get 10,000 page hits — only about 50 to 100 blogs can claim traffic at that level.

Blogger nation

Like the country, the community of bloggers is divided into camps. "They fall into two big clusters, liberal and conservative," Dooley says. "There's not a lot of crosstalk between the two groups. You don't see red-and-blue blogs." "We developed this Web site to show off the capabilities of the [Crawdad] technology to the marketplace," Dooley says.

No one else can use the technology without a proper contract with Crawdad Technology, which sells a desktop product and a service that listens to blogs for companies that want to monitor bloggers' opinions of them and their products. Crawdad Technologies and ASU share in the revenues. Crawdad is not the first business venture that is linked to a university.

Dooley cites the "Gatorade model," a business-academic arrangement that made the University of Florida very, very rich after the sports drink was invented on campus decades ago. Likewise, the University of Wisconsin benefited from dairy product patents. Such deals are high-risk, high-payoff ventures. "Many fail, but when one or two do succeed, it's a huge payoff to the university," Dooley says.

Impact of blogs

Dooley says blogs offer unprecedented speed in putting a finger on the pulse of public opinion: "Compared to newspaper, TV and radio, there's no comparison." Sometimes a seemingly minor sidebar re-emerges as big news. For example, a storm swept the blogosphere after New York Sen. Hillary Clinton remarked that the civil rights movement was inspired by Martin Luther King Jr. — a leader like Illinois Sen. Barack Obama — but became law thanks to a can-do president, Lyndon Johnson.

"It had come up on the net and died a week before it popped back up in the mainstream media," Dooley observes. Because blogs react so quickly, more reporters have come to use it to decide what to write about and what questions to ask, Dooley says. It amounts to market research for free.

The Crawdad Web site says its technology provides software and services to analysts and research professionals who need to transform unstructured text into insight about politicians, issues, celebrities, brands, products or customer service. Cold Stone Creamery is one of many companies that use Crawdad's Listening Post to monitor blogs and MySpace for buzz about their products.

Bottom Line:

• Text-analysis technology delivers a quick and accurate take on what people are thinking.
• Its applications are many in the political and commercial worlds.
• It is the kind of technological-academic arrangement that benefits innovators and the universities with which they are affiliated.
• Wonkosphere.com, even before Super Tuesday, indicated that the November general election would pit Republican John McCain against Hillary Clinton or Barack Obama and that Republican Mike Huckabee has a promising future.