Visualize Your Catalog

“Help, I have a paper due on global warming!” is an all-too-familiar reference question. But a search on “global warming” in an online catalog or database usually produces a long list of uncategorized results. Major B.J. Jansen and Amanda Spink found in their analysis of web document use that 80 percent of web searchers view only the first results page, and most rarely view more than ten to 20 citations. Students opt for shorter results lists because the longer ones lack any context information or categorization.

Two years ago, we surmised that visual displays of search results might remove some of the difficulties students face. Visualization presents large amounts of information in compact form and in several layers of detail. Search results are displayed according to easily discernible visual cues such as color, size, shape, and location in order to help users analyze results at a glance and decide what to examine and what to bypass. Patterns and connections between results become apparent. For beginning researchers, freshmen, and sophomores, a visual display might be particularly helpful.

Recent research supports our suppositions. Many people are visual learners. In the Net Generation and Libraries, J.K. Lippincott found high school students “relied heavily on information displayed in graphic form on web pages and often relied on graphics and visual clues to interpret the relevance of such pages.”

Visualizing search gives users the ability to browse in a virtual space, which is difficult in online search environments. When students browse titles on shelves in a physical library, they get a broader understanding of subject areas and related topics. Best of all, browsing opens up the possibility of a serendipitous discovery, one of the joys of doing research.

With these thoughts in mind, we began researching different types of information visualization, popularly known as “infovis.” There are many types of infovis, but libraries tend to use clustering and hyperbolic trees. Clustering is a text display of categorized results visible in Vivisimo's search engine, Clusty. Hyperbolic trees display information in a treelike formation showing connections among categories. Stanford University's HighWire Press is a good example.

After exploring several options, we decided Grokker was the best fit for us for several reasons. Grokker is a federated search product that provides both text clustering and a visual map, offering students a choice of how they want to view their results. It helps users focus their research by organizing concepts and revealing connections among topics. And the visual aspect makes research fun. In addition, Groxis, the company, had experience working with libraries and database vendors, including EBSCO and Thomson. It had also developed a desktop version of Grokker for Stanford that seemed to have great potential for student discovery.

At first, we planned to implement the desktop product in our Science Library as a pilot project. Then, Groxis informed us it was developing a web-based version—a better bet because it would be available to our patrons both on and off campus. We ended up with a web-based application that enhances the patron searching experience and interfaces with all of our EBSCO databases, IEEE, and our library catalog (Ex Libris's Aleph), as well as Google and the Internet Public Library.

We set up a Grokker Implementation Task Force with two librarians, an information technology specialist, and two consulting librarians. One librarian chaired the team, served as liaison among library administration, Groxis, and public services staff, and also helped administer usability studies. The other librarian organized the usability studies and provided feedback and suggestions about the product. The library systems specialist was crucial because he linked the catalog to display live data and configured EZ Proxy for off-campus Grokker patron access and authentication.

Many other library staffers contributed to the success of the project, including Sandy Card, head of cataloging, who assisted in selecting data fields from the MARC record, such as 245a, 650, 651, and 600v, x, y, z, to create a Grokker clustering plug-in for our online catalog. In addition, our library systems specialist (Tom Tran) helped Groxis design a record that displayed item availability at the top level in the results list.

Which databases to target for Grokker? We developed an initial list that covered a broad spectrum of subject areas. Since we wanted to focus on freshmen and sophomores, we included the large general databases we recommend for initial research. We modified this list many times based on input from our colleagues and because the API (application program interface) of some databases was not compatible with Grokker when we implemented it [see Richard Akerman's “Library Web Services,” p. 6–8].

We wanted to make sure users understood that Grokker was provided by the Binghamton University Libraries, so we customized it by adding our library logo and color scheme. We also worked with our web team to place a clickable image to Grokker prominently on the library homepage.

Library staff tested several beta versions of Grokker, providing suggestions to Groxis and the task force. As a result of the feedback, Groxis eliminated superfluous buttons, changed tab locations, and made sure “free resources” Google and Internet Public Library appeared at the bottom of the results list to encourage use of library subscription databases and to improve research. We're still receiving feedback. Groxis has been responsive and incorporates our suggestions into its monthly updates.

We wanted to avoid one hot-button issue: telling patrons items are available when they might already be checked out. To solve this problem, Grokker uses the SRU (Search/Retrieve via URL)/SRW (Search Retrieve Web Service) protocols developed by the Library of Congress and partners. SRU/SRW is a search protocol for Internet search queries that can use the Z39.50 standards.

Using these standards, Grokker harvests our data while the user searches in real time. No data is uploaded to Grokker, and all information presented during searching is done in real time. Grokker and Binghamton University Libraries collaborated to come up with this unique solution. We are the first Ex Libris Aleph site to use Grokker to show our catalog visually.

We conducted two rounds of usability studies this past spring [see “Real Simple,” LJ 11/1/06, p. 34–36] with five library student volunteers for each. We used the think aloud protocol for usability and the screencast program Camtasia to record audio and computer screen navigation (both cursor movement and clicks).

Our first usability study examined the beta version of Grokker, and we compared student database use in the native interface vs. Grokker. Students were asked to search a series of questions, first in the native interface, then in the Grokker interface. We analyzed the data subjectively based on answers found, time needed, and ease of searching and navigation. We gave Groxis suggestions about tab placement and tool management.

We were concerned about performance because it took sometimes up to ten seconds for Grokker to retrieve the data. Students didn't seem to mind the wait, however, since the search screen prominently stated, “This should take just a few moments.” Users loved the outline view, but the map view took a little getting used to. Yet as one student put it, “Oh, this is like a pie chart of articles” (left).

Our second round of usability studies examined the final implementation of Grokker. We evaluated the use of databases and our library catalog in the Grokker interface and started our study in the visual map view as opposed to the outline view as we had done in the first study. Again, the students loved the outline view. They felt both the map and outline view were “more organized” than other databases they'd used.

Some students did not understand that Grokker is an interface for library databases and our catalog, which will be a continuing challenge. But, overall, the feedback was very positive. As one freshman put it, “This will be insanely convenient.”