Jennifer Zaino
SemanticWeb.com Contributor

Twitter is the new American - make that global - water cooler. People are eager - and quick - to share their opinions about many things in 144 characters or less: Recent news stories have considered Tweet's influence on the Iranian election protests, the box office impact of negative Tweets about new movies, and GM's decision to cancel a planned Buick SUV in part thanks to unflattering tweets about the model.

Clearly, there's a lot being said in the Twittersphere - even if 40 percent of it is pointless babble, according to a recent study conducted by market research firm Pear Analytics. But that other 60 percent of unstructured Tweet text may be worth understanding better as they relate to your specific interests, and that's something that Expert System is planning to help users do with its as-yet unreleased semantically powered search site that aims to make hash-tag searches more accurate.

The developer of the COGITO semantic software engine for searching, discovering, classifying and interpreting unstructured text information wants to help people understand conversations categorically regardless of how different Twitterers may express a thought.

"Here's the problem with Twitter," says J. Brooke Aker, CEO of Expert's USA subsidiary. "The messages are so short, they are coming from so many people, they're about anyone's opinion and about anything under the sun. It's a mess and it's hard for me to follow unless I stay in a conversation on a near real-time basis."

Understanding conceptually

It's applying its technology to help users sort out not so much the minutia of everything being said but the nature of a hash tag debate - over, say, health care reform. It will categorize comments in order to provide a visual navigation tool for the debate (who said what most intensely, and over what period of time, etc.). It aims to save people the usual long-winded review process using the Twitter interface of who said what within a hash tag before deciding to read more or respond further, Expert says. Its semantic capabilities mean that it can go beyond keyword technology to distinguish between different concepts being expressed in a hash tag string even when the same words are used, or to tie together the same concept when Twitterers use different language.

"People may be talking about similar things but they may use different words to describe them, so this system can do things like identify the nature of the debate, how it's changing, if this is the hash group you want to join," he says. The graphical display will order related tweets from oldest to newest, and provide different-sized bubbles to indicate the number of people talking to the point.

So a small bubble would be two people talking back and forth and it would appear in the right-hand corner if the conversations were recent, while a big bubble in the right-hand corner would indicate a lot of people talking about a particular topic at hand. Users will be able to click on the bubble to get more details about where the debate is trending, perhaps the newest focus being less on the funding they may have thought it was about and more on, say, death panels.

Equally, Aker expects the tool can be of use to the parties invested in various debates - like the Obama administration in the case of health care. Had the site been available before the health care reform fight reached their current heated pitch, the administration might have used it to more efficiently take the pulse of reform in the Twittersphere. Perhaps they might have discovered through this that many average citizens with health insurance, tagging their tweets #health care reform or #health insurance or whatever else, still hadn't been convinced of the need for reform.

"Before it became an issue, [they could have realized they] were not doing a good enough job convincing the average voter why this is important," Aker says. "Something like this could have folded right into the nature of speeches the president gives, so he could have been out in front of being responsive to concerns before it ends up in the press."