Search Startups Target Clustering

Two smaller search players are unveiling new sites that use linguistic techniques to refine searches and categorize results.

As the major players in Web search duke it out, a growing number of startups are embracing a different way of retrieving and sorting search results.

Called clustering, the technology dynamically groups search results into categories as a way of solving one of the perennial problems of Web search: figuring out what the searcher is really looking for.

A new search engine named Clush this week plans to announce its arrival on the clustering scene. It follows the launch last week of a search site called Clusty from clustering-focused search company Vivisimo Inc.

While both search engines categorize search results in clusters, they are taking different approaches. Clush is built atop its own search index, which, for now, is tiny in comparison to those of Google Inc. and Yahoo Inc. Clusty, on the other hand, is a meta-search site, drawing its results from multiple engines and other information sources.

Clush, first publicly demonstrated during the Search Engine Strategies Conference in August, is the brainchild of search-engine optimization experts and is part of parent company InfoSpider Inc., of Milpitas, Calif.

Clush President Frank Mattox said he and others at the company were frustrated with the search experience on other engines, most of which base results on various concepts of link popularity.

Clush instead bases results on more of a linguistic analysis. Its clustering technology delves into heuristics in order to group and categorize various Web pages, Mattox said.

"What we were trying to do was to come up with a better idea and search experience," Mattox said.

/zimages/1/28571.gifClick here to read more about another startup trying to move beyond keyword-based search.

While it originally included a single grouping of results, Clushs latest update delivers keyword results with four boxes of clusters surrounding the query.

The clusters represent different types of categorizations for a keyword, letting a user refine the search. For example, searching for the term "cars" will return clusters targeted around the most common car searches, alternative words for cars, parts of cars and types of cars.

"What we find is a lot of people just type in generic keywords like cars, but they might be looking for something more specific," Mattox said. "What this does is to help you think and provides expanded ideas of what you might be looking for."

Clushs Web index sits at millions of pages so far, and Mattox said the company is working to grow it exponentially. Part of Clushs indexing includes images of Web pages themselves, which the search engine displays to the right of its results as a quick preview of a site.

Along with crawling the Web, Clush offers a paid inclusion program to sites. It also lets users rate results, using the feedback to tweak relevancy and discover spam-oriented sites. Clush also is working to license its technology to other sites and companies, Mattox said.

Next page: Clustering in search of a broader base of users.