Ask Jeeves Search Brings Navigation Through Concepts

 
 
By Matthew Hicks  |  Posted 2005-03-18
 
 
 

Ask Jeeves Search Brings Navigation Through Concepts


SAN DIEGO—Ask Jeeves is preparing to release a related-search feature that it hopes will relieve Web searchers from the burden of repeatedly refining their queries.

Executives from the Oakland, Calif., company demonstrated a prototype of the feature this week during the OReilly Emerging Technology Conference here. It will expand on an existing related-categories function by offering categories of search-term suggestions that both narrow and broaden a query.

The expanded feature is expected to go live on the search site within the next few weeks, said Apostolos Gerasoulis, executive vice president of research and labs at Ask Jeeves. The company will first make it available for Web searches, but eventually plans to add the option to vertical search areas such as image search.

To deliver related queries, Ask Jeeves is tapping into the clustering technology built into its Teoma search engine.

"Were trying to improve the [search] technology and provide a differentiator from the rest of the industry," said Gerasoulis, a co-founder of Teoma. "We wanted to focus on quality that is very different from anyone else."

Teoma is able to discern clusters of communities on the Web, Gerasoulis said. By understanding the popularity of expert communities and relationships among them, Ask Jeeves can determine categories of related search terms.

Ask Jeeves, which bought Teoma in 2001, is a smaller competitor battling Google Inc., Yahoo Inc. and Microsoft Corps MSN division in the increasingly heated search-engine war.

Click here to read about Ask Jeeves recent update of its Teoma engine and release of personalized search.

Based on the prototype, Ask Jeeves will separate the navigation of related search terms into three categories. The first, under the heading "Narrow Your Search," provides refinements to a query, while the second, called "Related Topics," suggests more broadly related terms.

In a demonstration, a search for the rock band The Beatles returned narrower searches such as "Beatles lyrics" and "Beatles pictures" as well as broader suggestions such as "Rolling Stones" and "Beatlemania."

A third grouping focuses on peoples names that relate to a search. In his demo, Gerasoulis entered a query for "light bulb," hoping to find information about its inventor. In the "Related People" category alongside typical Web results, a link to a search on "Thomas Edison" appeared.

"This is navigation through concepts," Gerasoulis said.

With the upcoming feature, Ask Jeeves is exposing more of the capabilities of its clustering technology within the user interface of search. Clustering has gained the attention of both search startups and the larger engines. Startup Vivisimo Inc. has put clustering into action in its Clusty search site and as part of the revamped America Online Inc. search.

Google has demonstrated research work into clustering for grouping similar searches, while Microsoft Research is testing a search toolbar that clusters results.

Next Page: Adding research work into projects.

Projects


Ask Jeeves clustering feature is a result of work from its research labs. The search company eventually plans to give public previews of its research projects through a yet-to-be-launched Web site called Ask Jeeves Alpha.

Unlike its larger competitors, though, Ask Jeeves does not plan to keep any projects in alpha mode for long.

"The research work we do is very oriented to developing products," said Rahul Lahiri, Ask Jeeves vice president of product management for search. "At this point, were all about adding [research work] very quickly into the products."

To read an interview with Ask Jeeves CEO Steve Berkowitz, click here.

Teomas clustering technology also played a role in Ask Jeeves release of an improved relevancy algorithm for its image search, Gerasoulis said. Earlier this month, the company revamped the way it ranks results for images. It uses an image index from Picsearch.

During his presentation, Gerasoulis conducted before and after searches on Ask Jeeves image search, showing how a search for "sunset" changed from results where images of sunsets were interspersed with seemingly random images of people and logos to results more exclusively showing pictures of the setting sun.

Separately this week, Ask Jeeves took another step in supporting the Mozilla open-source project. Ask Jeeves launched a Mozilla Firefox version of its browser-based search toolbar.

Ask Jeeves previously has talked with the Mozilla Foundation about potentially creating an Ask Jeeves-branded version of Firefox and about contributing codes from Ask Jeeves desktop search product to the open-source community.

Check out eWEEK.coms for the latest news, views and analysis on enterprise search technology.

Rocket Fuel