• by Steve Smith , Staff Writer @popeyesm, June 12, 2009

Quick note before getting into the meat of today's column. The agenda for OMMA Behavioral, July 30 in San Francisco is posted here. We welcome your input and advice and hope to see many of you there, especially many Left-coasters who couldn't make the New York show in February.

 

Voicing opinions online may be one of the core behaviors that most of us online either practice or pursue and watch. In fact, we spend an awful lot of time hunting for opinions, across blogs, scrolling into article comments sections, hunting for ratings and reviews, etc. When most of us read impartial news articles about a topic, we also wonder what the reigning sentiment is around the issue, out there and apart from the voices quoted in the piece before us. This is the subtle itch that an upcoming product from New York start-up Jodange is promising to scratch, for the benefit of publishers, marketers, and perhaps even stock-pickers.

"Thoughts, feelings and sentiments coming off the Web -- that is what the technology is about," says co-founder Larry Levin. The technology uses linguistic analysis to extract opinions from text, identify the sentiments expressed, the opinion holder and the topic. When combined together they produce, for example, the "Top of Mind Recovery Pulse," which analyzes everything from news articles to blog posts and even Twitter tweets to surface and quantify attitudes towards the bailout, deficits, economic growth and many other sub-topics.

Available as a Google Gadget, the widget can drill into opinions expressed by Timothy Geithner or The Wall Street Journal on any of these topics and quickly show whether the source skews positive or negative. The technology becomes especially interesting when it is matched contextually with any content page. If I am reading a piece about the new iPhone, the widget would pull in current sentiments about the device being voiced elsewhere in blogs, tweets and articles, ranked by sub-topic or by the strength of the opinion. The model moves beyond the typical recommendation engine. It culls from across the Web opinions of others about the topic you are engaged in at the moment and even finds and quotes the pertinent opinion point in their prose.

It is an interesting model because it aggregates an important activity we rarely discuss as a discrete kind of behavior or content -- opining. The applications are multi-faceted. For publishers, a box like this could be leveraged in many ways. It could be a link farm out to the rest of the Web from an article, a kind of aggregation service. It could also be used to re-aggregate the user-generated content that heretofore has been siloed in message bases or scattered across comments sections attached to specific articles. Now, any opinion relevant to the article at hand can be surfaced regardless where it was first planted.

For marketers, the same technology could be focused on brand management. The sentiment engine not only can follow the attitudes and conversations around the brand, but identify the key influencers who are driving the conversation. "You start to see the people who are the instigators and the followers, those who just amplify or initiate," says Levy.

You may also begin to see the trends in thinking and sentiment that ultimately foreshadow behavior. Jodange has already been experimenting with predictive analysis on stocks. The technology can track the news and opinion on a given company and map the changing tone to changes in market value. Ultimately the algorithms may be able to detect the leading indicators of market or pricing shifts. Further, you could identify the opinion makers whose expressions have the most effect, positively and negatively, on company reputation or stick movements.

Once you have a handle on aggregated opinions, there are so many ways the content can be spun, and Jodange is exploring some of them already. In-text tagging could pop up opinions tied to specific topics as you're reading, or aggregate sentiments about or from a person named in an article. As a freak about the possibilities of mobile, I am especially interested in how the engine could be turned into a mobile app. In essence, running a sentiment search on a company or topic could apply to the Web at large in much the same way some of us are searching Twitter right now.

I am fascinated by the concept of opinion and sentiment as types of online content and even behavior that form part of the digital media experience. How Jodange itself fares in the market is anyone's guess. The technology needs to spider the content broadly and ID the material effectively. I can't vouch that either element is there yet. For instance, running topic searches in the Google "Top of Mind" gadget returned diverse results: either limited returns, or some false positives. The more refined topic-specific tools for tracking subjects, like the economic recovery and S&P 500, were genuinely enlightening, however. Graphs can register visually how positive and negative sentiment is tracking over time, related to a topic, or voiced by or associated with specific opinion maker like President Obama or Geithner.

We often say light-heartedly that the Internet is a large collective "brain." This sort of technology is fun and fascinating because it applies to the Web a CAT scan, a mind reading and mood ring all in one. It could tell us how people are thinking and feeling and perhaps are even poised to act.