The search continues

The problem with search technology is that it involves too much searching.

That's the conclusion of industry executives who are trying to take information retrieval to a higher level. "The state of the art for access to information on the Web is really terrible," said Eric Sall, senior vice president of marketing for iPhrase Technologies Inc., a natural language search vendor. Citing a Forrester Research Inc. study, Sall said up to 95 percent of users will abandon a search if it takes more than five steps.

Indeed, traditional index and search technology looks rather old-fashioned to those pushing the technology envelope. They contend that government agencies can't rely on staples such as keyword searching if they expect to create appealing, citizen-centric Web portals or build effective employee intranets.

Instead, software providers are pursuing advances that will take some of the effort out of searching, expand tools in new directions and make the new technology easier to deploy.

The Personal Search

New developments include personalizing search results and using social networks.

Ashok Chandra, a senior vice president at Verity Inc., describes the evolution of search technology in three tiers: information discovery, organization and personalization/social networks. "As we go up the tiers, it's really about enlarging the context in which the document or information resides," he said.

Personalization relies on user profiles to serve up the most relevant information. Smartlogik Group PLC's Insight, for example, builds a personal user profile based on information the individual has published, browsed or searched for, according to Carla O'Brien Malley, product marketing manager at Smartlogik.

Inktomi Corp.'s profiling approach, meanwhile, is to integrate with Lightweight Directory Access Protocol databases, which are popular for organizing information about individuals or computer resources on a network. This summer, the company plans to release group-level personalization, which will analyze trends associated with groups rather than individual users, according to Andrew Littlefield, chief strategist in Inktomi's search solutions group.

In addition to boosting document relevance, such profiling can also locate human sources of expertise. This social network technology adds another dimension to search results. For example, a company intranet with personalized search technology can direct a user to an in-house expert on a given subject.

"Social network technology is of great interest," said Arun Baheti, California's e-government director. "Our own experience is that personalization has resulted in a high level of customer satisfaction and increased user visits."

California's Web portal aims to provide a quick and easy way for citizens to obtain information, Baheti said. "We will never be totally successful using simple methods of information retrieval like standard text searches," he noted. Instead, the portal relies on Verity's K2 software, which started shipping with social network capabilities last year.

Federated and Peer-to-Peer Searching

Vendors are broadening the scope of search technology in other ways. Federated search, for example, allows users to send queries across multiple sources. This could mean culling search results from different search engines or various databases and applications.

Verity began shipping tools with federated search capabilities last year, although Chandra said the technology is "just at the beginning."

Inktomi plans to add a federated search capability that will pull in content from databases and, potentially, other search engines at query time, Littlefield said. This summer, the company will provide the ability to pull in content directly from Extensible Markup Language repositories.

Expanding searches into peer-to-peer networks is another new development. Peer-to-peer systems allow all workstations and computers in a network to act as servers to all other users on the network.

But customers may not be ready for this search development. Inktomi deployed a peer-to-peer prototype early last year, but Littlefield said the experiment generated little interest. n

Moore is a freelance writer based in Chantilly, Va.

***

Keywords in search technology

* Personalization: Relies on user profiles — based on documents created, browsed or accessed — to serve up the most relevant search results.

* Social networks: Identifies subject experts in an organization and automatically links users to them as part of a search result.

* Federated search: Allow users to execute queries across multiple sources, which could mean culling search results from different search engines or various databases and applications.