Jan 25, 2006 (03:01 PM EST)
New Search Engine Aims For More Targeted Results

Read the Original Article at InformationWeek

A new search engine called Dumbfind.com that launched Tuesday aims to compete in the already-competitive search market.

Dumbfind.com entered onto the scene with a two-box search engine approach. Their idea is to combine traditional keywords with topics or categories. Before query results are served up to users, the search engine sorts its findings into related tag clusters to closely match the searcher's intent. When the search query is returned a list of tag clusters also appears. "We want to push the envelope, innovate, and offer something new," said Chris Seline, chief executive officer.

Aside from the main search engine, an interesting feature in Dumbfind News offers query results by day. About two-weeks are gathered and identified in separate folders on the left of the Web page.

Dumbfind.com, which took two years to design and build, is supported by five employees and an undisclosed private investor. The site runs on Linux and indexes searches with help from proprietary software, applications written in Java, sophisticated algorithms, and 50 Penguin Computing servers housed in a Washington D.C. office.

The algorithms build "large relationship indexes" to determine how search-word concepts relate to each other. The first search box returns queries on literal words and the second search box returns terms related to the word. For example, when searching for the word "music," the search engine also looks for related terms. The algorithms compare words when queried, and knows that musician maybe related to guitar or vocalist.

The downside, for now at least, when searching for queries with the two-box search engine is the time it takes to return results nearly doubles. The upside is it delivers more relevant Web pages. The company is working to increase the speed.

There are many new tools on the roadmap. Next week, Dumbfind.com will launch a "random" feature search. "If you've searched for a site and you want to look for related content, you can hit the 'random' button and it will return sites very related to the word," Seline said.

Other tools that will integrate with Dunbfind.com based on social search are scheduled to debut in March or April. A full site will is planned for June or July. To offer e-mail as a service is not in the plan, but revenue will come from advertising.

And Dumbfind.com hasn't been deterred by U.S. government lawyers asking search engines for data relating to search terms. The Department of Justice is looking to defend the 1998 Child Online Protection Act (COPA), which was blocked in court, by gathering data from the major search engines to determine the amount of pornography accessed online and the chances a minor might come across it as they search for other content on the Web. "It brings up the point that Google and other search engines do know a lot about their searchers," Seline said. "In a way it's sending a message to search engines to stop storing so much user data and they may decide to comply with consumers wishes."

It's tempting to track your users to target advertising, Seline said, but Dumbfind.com will refrain from the practice for now. The site already has keyword-based advertisements that are targeted to the search query. Soon it will have tag-based advertising where the ads are determined on the query results.

Dumbfind.com is not Seline's first search engine. The electrical engineer in 1999 launched 2wrongs.com, a search engine based on user bookmarks he sold in 2001 to Cyveillance, online security company.