• by Laurie Sullivan , Staff Writer @lauriesullivan, August 21, 2008

Misplaced contextual ads are hilarious, right? Not when they're your ads. Microsoft researchers are developing technology that scans the words in online articles to detect positive or negative connotations related to contextual ads placed nearby.

The technology being tested by the Redmond, Wash., company's incubator alerts the ad server to positive or negative words or statements in an article and where they appear on the page. Lee Zen, applied researcher, U.S. adCenter Research and Mining at Microsoft, calls the technology "content detection in sub-documents."

Although not in commercial use, the first version of the technology identifies the words. The second version pinpoints the block of type on the page. The technology relies on natural language processing to understand the meaning of words and the way they combine to form negative opinions or feeling. It can determine whether the online article discusses murder or crime prevention. Machine learning lets it build on knowledge from past experiences.

Identifying specific words in blocks of type would also let Google and Yahoo, which place contextual ads across a variety of news and sports sites, insert them on parts of Web pages that contain positive information rather than negative. The technology communicates the information to the ad-delivery engine before the decision is made as to where to place the ad. Not satisfied with not putting ads in the wrong place, Zen says the next step is to have an ad engine that places the ad in the correct place.