Jan. 17, 1996 Voyager 3, Endeavor's state-of-the-art library management system, is the first library automation system to offer a relevance search engine like those used on the Worldwide Web. And Voyager's creators believe that they're on the cutting edge of a paradigm shift, from traditional Boolean searching to an easier, more effective way to perform OPAC searches.

"It's no accident that Lycos, Netscape, Webcrawler, and many other Web search engines employ relevance searching, and it's no accident that we selected it for Voyager," said Endeavor CEO Jane Burke. "We designed Voyager to be able to access information wherever it resides, and relevance searching is the new standard in client/server systems like ours."

Cindy Edgington, the architect of Voyager, points out that students are far more likely to be familiar with Web search engines than with Boolean operators and the syntax of Boolean searching. "Where do they learn Boolean searching, except in library instruction programs, 'she asked. "And how many students actually learn Boolean well in those classes? Relevance is far easier to use: It requires only words, no operators that students and faculty may never remember anyway."

The essence of relevance searching is that it uses all of the information the searcher has at hand, and that information is natural language, not the user's best guess about a controlled vocabulary, said Edgington. "Boolean searching encourages using as little information as possible, and the process of narrowing and limiting parameters results in the loss of relevant documents." With Voyager's relevance engine, you simply type in words, everything you know about your subject, and let the computer sort it out"

"How many students have attempted to type a sentence into an OPAC search," Edgington asked. "With Voyager a sentence actually works."

"Keyword anywhere is the key," added Burke. "Voyager's relevance engine searches the entire record. Many systems don't even offer keyword except in the title and subject."

Burke offered a historical perspective on searching: "Boolean searching developed, in part, in response to a now-outmoded economic model where searchers paid by the 'hit' and sought to minimize connect time. The evolution of computing has changed all that. We no longer need to restrict access--we want to offer more. Relevance searching is cost efficient and far more effective. It expands results."

Relevance effectively does a fuzzy "and" search and then a series of fuzzy "or" searches. Through this process, which presents the user first with records that include all of the terms entered and then the records with some of the terms entered, relevance actually expands the scope of the search.

Boolean searching provides only records that exactly meet all of the conditions of the search. Relevance provides those records plus others that meet most of the conditions. In a large library collection, relevance actually suggests additional records the user would not have found in a Boolean search. Because of this, Endeavor believes that relevance is very appropriate for a library collection.

Voyager presents the results of a relevance search differently than most Web search engines, however. Instead of a numeric ranking, Voyager displays a small graph--a histograph (see illustration)--to show how relevant each record is. Users clearly see which records are most relevant to their search, and which are only somewhat relevant.

Edgington said she has done both Boolean and relevance searches of the records of academic libraries, with dramatic results. "A relevance search generated all the records the Boolean search pulled up, but it also produced useful records missed by the Boolean."

Librarians wedded to the paradigm of Boolean searching will immediately think of students printing out lists of tens of thousands of documents. That doesn't have to be the case, said Edgington. "With Voyager, librarians have the option of setting parameters. For example, if five search terms are entered, they can set a parameter that three of those terms must be there."

Jeff Petersen, the author of the Voyager relevance software, explained the system's weighting technique. "If a user enters five terms and all five match one record, that record will be at the top of the list of retrieved documents. Records with four terms will be listed next, and so on. The user can read through and make their own decision about how many documents are relevant and how many they need." Petersen added terms are weighted according to frequency of use in the entire database. "That means that a specific term is deemed more relevant the less frequently it appears in the entire database."

Burke acknowledged that being first in the library marketplace with relevance searching is a somewhat uncomfortable position. "Boolean searching truly is one of librarianship's paradigms, and paradigm shifts are difficult," she said. "But Endeavor is committed to relevance searching. We're evangelists for it. If it's right for the World Wide Web, it's right for libraries' OPACs."