USC Annenberg Online Journalism ReviewUSC


Kosmix.com paints Web searches in shades of gray

2006-06-19

By Micah Ailetcher: New search engine Kosmix.com is taking a different approach to the typical Internet search by categorizing its results.

Rather then presenting search results in a standard list form that can be thousands of answers long, the results are organized and sorted into different categories.

The goal, according to product manager Mark Johnson, is to provide better answers. By presenting the search results in categories, Johnson feels that they have an advantage by allowing the user "to see the data in a lot of different ways." In this way, they see themselves more as a "compliment to something like Google or Yahoo" rather than a replacement.

For example, Kosmix’s U.S. Politics search engine categorizes its results as "Conservative," "Liberal," "Libertarian," or "Political News." The site currently hosts five different search engines: Health, Video Games, Finance, Travel and U.S. Politics.

Kosmix started with its Health search engine. A query in this engine, for example, for breast cancer produces results that are identified as "Basic Information," "Expert Information," "Message Boards," "Blogs," "Alternative Medicine" and over a dozen other categories.

The right side of the results page presents Websites in the familiar search engine format, which includes the individual Website’s category in Kosmix in addition to the Website’s summary so users will know before they navigate away to a web page whether it is classified as a "Quiz" about breast cancer or as the findings from a "Clinical Trial."

On the left side of the results page is a list of links that allows the user to display the search results by a specific category. For example, this can be helpful if a user only needs to view the results for breast cancer that are categorized as "Symptoms."

With its patent-pending algorithms, or as Johnson calls it, their "secret juice," Kosmix crawls the web and classifies different sites by the different "signals" that they provide. There is an amount of subjectivity involved in classifying sites as "Liberal" or "Expert Information," but once a classification is made, the algorithm moves on to classify new sites by looking at things such as what classified sites the unclassified site is often linked to.

"When you are dealing with a very nuanced subject, like politics, there are always going to be gray areas," Johnson said.

These "gray areas" are exactly where Kosmix sees itself coming into play by helping the user to sort out the type of Website that they are looking for in the search stage. "In some cases, the algorithm can do a much better job than we can in figuring out [a Website’s] subtlety," Johnson explained.

Kosmix hopes to add more categories to its search results. For the U.S. Politics engine, one consideration is a "Green Party" category. The Video Game and Finance engines are the latest additions to the site, but Kosmix is hoping to add more diversity to their search engines.

"The endgame is clear," Johnson said. "The goal is to categorize the entire Internet. Eventually, no matter what query you type in, we will have the categories that are appropriate for it."

Links to this article: Technorati, Yahoo

This article has been archived and is no longer accepting comments.