2023 KMWorld Media Kit Available Here 

Give your search a boost

Search is an ongoing endeavor in most knowledge-driven organizations. Workers spend a lot of time searching for information, and often don’t find what they want. If the search solutions available to an organization are not achieving the desired results, several options are available to enhance performance without the need for a complete overhaul.

At Sun Microsystems, engineers who need to acquire knowledge in connection with product development, skill development or technical questions often turn to the resources in Sun’s digital library. The library provides access to a wide variety of digital documents, including publications from IEEE and the Association for Computing Machinery (ACM), several collections of e-books, e-journals, market research, training sessions and social media tools for access to information.

"Our engineers not only had to work with numerous interfaces," says Christy Confetti Higgins, digital library program manager in Sun’s Learning Services organization, "but also had to go to each information source separately. It was not an efficient way to work."

To expedite the search process across those many repositories, Sun began using Grokker, a search aggregation solution from Groxis. Grokker sits on top of other search engines and is designed to make sense of results obtained by searching multiple data sources. The product name is derived from "grok," a term created by science fiction writer Robert Heinlein in Stranger in a Strange Land to convey a sense of profound understanding and the merging of multiple intelligences.

With Grokker in place, researchers can search multiple data sources from one interface and in one search operation. It also clusters the results by extracting metadata from each document found. Results can be viewed either in a hierarchical list or in a map view in which the results are presented as a series of circles, with larger circles representing topics containing a greater number of hits.

"We now have eight different data sources that our engineers can search using Grokker, and the results are aggregated into a single set of results," Higgins says. Results also can be viewed by source, and users can slide a marker along a scale to limit the date range to more recent documents.

Sun will soon be moving to the new version of Grokker, which offers some additional options. "We expect the next version to have more analytics on usage," Higgins explains. "That will give us some insights about what our engineers are finding or not finding."

In addition, Sun plans to use Grokker with the Google search appliance so it can search across more of the internal databases in departments throughout Sun. Currently, Grokker searches a few such databases, including archives of knowledgebases compiled from e-mail discussion groups.

"We have found Grokker to be exceptionally helpful in the research and learning process," says Higgins, "and we are looking forward to expanding its use."

The philosophy behind Grokker is that a long list of search results is not likely to be useful. "People often do not go past the first screen," says Randy Marcinko, CEO of Groxis. "They really benefit from a search solution that groups the results into meaningful categories."

Grokker maps the search results to a normalized set of metadata that makes it possible for users to do more analysis, because the data elements are made comparable.

"Other products can do a federated search but can cluster results only within their own collections," says Marcinko. "Grokker complements search engines tools by providing visualization, federated results and clustering across multiple data sources."

Better site search

One of the reasons that finding information on a particular Web site is so difficult is that users come to the site with a variety of objectives in mind. Some are looking for the answer to a question, some for background information and some for a particular product. Even the best Web designers cannot anticipate the needs of every user.

An innovative product from Baynote was developed to help connect people with the right content by observing the behaviors of visitors to the site. The Baynote Collective Intelligence Platform determines through a set of algorithms whether a visitor has been successful in obtaining the desired content, and uses that information to improve the search experience.

NetApp creates storage and data management solutions designed to produce innovation and cost efficiency. The company relies on its Web site to present product information to new and returning customers, but had received some complaints about the search function on the site. Through a serendipitous meeting with a Baynote executive, a NetApp employee became aware of the product and recommended that the company check it out. After a convincing demonstration in which search results were compared with and without the Baynote platform, NetApp decided to deploy it.

The Baynote solution works with NetApp’s existing search engine, which uses keywords and pattern matching to find content. "Our traditional search function was effective in locating documents containing the search terms, but not in anticipating what the user needed," says Michael Parks, Web technology manager at NetApp. "Someone looking for specs about a new product might end up reading a press release instead."

Baynote finds connections between pieces of content and the search terms that are being used. The algorithms look at information such as the length of time users spend on a page, whether they scroll down and whether they click on a link, to help determine whether a search was valuable. That information is fed back into Baynote in real time to allow for improved searching.

"One outcome of using Baynote is that the results list is shorter and more focused," Parks says. "We can also achieve a lot more control. For example, we can set the navigation so that the product page for a new introduction always comes up first if a user types that word."

KMWorld Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues