Rather than the user population embracing these efforts, however, complaints about usability in the human-computer interaction aspect of information retrieval have intensified. This trend owes its existence to the inordinate number and variety of interfaces out and about these days. Or perhaps the trend lies in simple human nature: We love to find fault with things. One alternative explanation that might be worth considering is that the conflict between users and information interfaces has, at its root, a biological component. 
 

SO MUCH DEPENDS ON RED

With the exception of those who are visually impaired and must rely upon their other senses, humans are visual creatures who can process an image and interpret it much more quickly than reading text. Indeed, people will more quickly perceive and process a photograph of a red rubber ball sitting in a field of green grass than they will the sentence, "There is a red rubber ball sitting in a field of green grass." When looking at the photograph, understanding is almost instantaneous. With the sentence, it takes a longer period of time to read the words, interpret their meaning, and create a visual image in the mind's eye. 

This fact has not been lost on those responsible for interface design and information organization. What has resulted is a growing interest in the area of information visualization and its application to the information industry in terms of information structures that provide users with an alternative method of retrieving data using visual cues rather than text, or a combination of an underlying graphical structure with a textual component. 
 

DEFINING INFORMATION VISUALIZATION

The User Interface Research Group at PARC (Palo Alto Research Center Inc.) [http://www.parc.xerox.com/istl/projects/uir/projects/InformationVisualization.html] defines information visualization as "the use of computer-supported interactive visual representations of abstract data to amplify cognition." Got that? In more mundane language, it means that computer generated images are used to represent abstract concepts in a search result set. What's more, users can move these images around in various arrangements to gain a better understanding of the abstract concept being represented. 

Information visualization became a hot topic in the early 1990s—not surprising since it corresponds roughly with the rollout of the first graphical Web browser, Mosaic. Shortly thereafter, libraries began offering Web-based resources such as online catalogs and article databases. As general users began interacting more frequently and independently with online interfaces, concerns over usability and human-computer interaction with respect to data organization became of primary importance. 
 

ONLINE LIBRARY OF INFORMATION VISUALIZATION ENVIRONMENTS

This is reflected in the vast amount of literature addressing the issue of visual representations of data. Simply running a search in your favorite Web browser (I used Google) will yield enough reading to keep any interested novice busy for a week. One of the leaders in this field, Dr. Ben Schneiderman of the University of Maryland, is responsible (both directly and indirectly) for a great deal of the research that has been reported. The Online Library of Information Visualization Environments (OLIVE) [http://www.otal.umd.edu/Olive/] is a wonderful resource for obtaining background information on the types of visual structures and their uses. Users not only have the opportunity to read about the eight structure types listed, but can also follow links to current projects utilizing the technology. 

Schneiderman's research was the driving force behind Spotfire [http://www.spotfire.com], a commercial venture that provides data solutions using visualization technologies. Pacific Northwest National Laboratories [http://www.pnl.gov/infoviz/] has also contributed a great deal to the body of work, as has the aforementioned User Interface Research Group at PARC. 

In all of these instances, it is important to bear in mind that the data itself often takes a back seat to the effort to render this data on a graphical interface. Donald T. Hawkins ("Information Visualization: Don't Tell Me, Show Me!" ONLINE 23, No. 1, January/February 1999, pp. 88-90) effectively distills the issue to a simple sentence when he states, "It is important to note that what is being communicated in visualization is not the information itself, but its structure." This is a very valuable distinction to make, as the good intentions of information visualization can sometimes be lost in its implementation.

So how could libraries benefit from the field of information visualization? What type of visual interface would marry well with our organizational traditions? There are as many possibilities as there are ideas. However, consider the hyperbolic information structure—a powerful tool employing hierarchical trees superimposed over an atypical geometry—to offer the user a comprehensive view of a data structure.
 

HYPERBOLIC INFORMATION STRUCTURE

To appreciate the advantages that hyperbolic structures offer, it helps to have a brief explanation of the methodology used to create a hyperbolic display. It is not as frightening as it sounds, and those readers who have more than a fleeting interest in geometry should excuse this over-simplified explanation. I will concentrate on one particular model of hyperbolic plane mapping known as the Poincare disk model. In a paper published by John Lamping, Ramana Rao, and Peter Pirolli ("Visualizing Large Trees Using the Hyperbolic Browser," Proceedings of the Conference on Human Factors in Computing Systems, April 13-18, 1996, Vancouver, British Columbia, Canada [http://www.acm.org/sigchi/chi96/proceedings/video/Lamping/hb-video.html]), the authors explain the mapping of a hierarchical tree structure to a hyperbolic display citing two significant qualities of the structure:

• The nodes or components of the tree diminish in size the farther away they are from the center of the display.

 
• The number of nodes or components grows exponentially from parent to child. 

When mapped to a disk, the hyperbolic structure provides an effect that resembles that of the fisheye lens—the amount of room on the display devoted to a point at the center is much greater than the amount of room devoted to points around the periphery. When using this model, there are always several generations of nodes visible, allowing the user to extract a fuller understanding of the hierarchy's structure without getting lost in it. A hyperbolic display contains much more space than a simple Euclidean plane because the circumference and area of the circle it's mapped upon grows exponentially with the length of its radius. The larger the circle, the greater the amount of usable space.
 

REVOLUTIONALIZING SUBJECT SEARCHING

For libraries, this approach could revolutionize subject searching. Take, for example, the results most of us obtain when we perform a subject search in our regular online catalogs. I executed a subject search for "history" on a Web-based interface running over a LUIS system. The results, of course, were overwhelming and consisted of an extremely long list of subject headings and subheadings (and subheadings of the subheadings) spread across approximately 190 pages, with 25 headings or subheadings per page. 

Similar results arose when I performed the same search in other large academic library catalogs using other management systems such as ExLibris and Endeavor. The results from a basic subject search using the term "history" yielded pages and pages of subject headings, with 10 to 25 per page, depending upon the system's settings. Obviously, the term "history" is much too broad to be effective, but based upon empirical data, I believe it accurately represents the searching behavior of the average unassisted library patron. For the average user being inundated with thousands of possibilities, arranged in a linear fashion and distributed among hundreds of screens, information overload is almost guaranteed. 

It would be more effective to present the patron with a graphical representation of all pertinent subject headings arranged in a hierarchical tree. Unfortunately, it would be impossible to represent this tree structure in its entirety on a regular graphical interface because such a tree would be much too large to fit on the typical computer screen. 

One solution would be to place the structure on a hyperbolic plane and then map it to a two-dimensional circular area. This would allow the user to browse each heading and see how it relates to the other headings in the tree. Because the entire structure is mapped at one time and processed by the computer as a whole, the transitional animation between nodes being brought into focus and those getting dropped off when manipulated by the user is smooth and unaffected by delay. Lamping and colleagues describe this animation technique as a geometric translation of the structure on the hyperbolic plane, which simply means that the entire tree (or structure) is moved along a straight line in whichever direction the user decides to go. 
 

REAL-WORLD APPLICATIONS

To view and interact with an actual example of a hyperbolic browser, take a look at InXight Software's Star Tree navigator [http://www.inxight.com]. This is a relatively simple example of a hyperbolic tree structure being used as a site map. To interact with the display, the user can either click once on the node to make it jump to the center, or click and drag the structure around. Double-clicking on any node will open a new browser window that will display the page(s) associated with that area.

Another good example, also by InXight, is The Universal Library [http://www.ulib.org/webRoot/_hTree/] hosted by Carnegie Mellon University. This project of the School of Computer Science, which promotes its lofty goal of "Access to all human knowledge, anytime, anywhere," is a much better illustration of the technology not only because of the current magnitude of the structure (which is great), but also because of the expectation that it will continue to grow as more material is added. However, while this particular case serves as an example of what hyperbolic structures can do for the organization of knowledge, it also effectively demonstrates some of the limitations of the technology in its current state. 

Users who select particularly populated sections of the tree and try to move them to the center of the display will notice that the labels crowd each other out, making them impossible to read. And that is not the only drawback associated with these structures. Hyperbolic browsers require a JavaScript-enabled browser such as Netscape or Internet Explorer to operate. At present, sites using hyperbolic trees as navigational tools are really tools that facilitate subject browsing more than true search engines. The processing demands that these applications place on the user's machine can be a serious obstacle for those still using older/alternate operating systems or outdated chip architectures. 

Such structures must necessarily include an option for traditional keyword searching. Indeed, hyperbolic trees are wonderful for browsing collections, but quickly become cumbersome to the point of being useless when trying to locate specific documents or pages when the user does not know where in the hierarchy the pages lie. 

These applications are also not very compatible with the text readers employed by sight-impaired users, so the blind would find the barriers to using hyperbolic browsers extremely trying, if not insurmountable. These problems can only be solved by the eventual replacement of older systems as they fail with age and the continued development of technologies to assist the disabled. The issue of node crowding, however, has been addressed by another technology that, while employing many of the same navigational conventions as hyperbolic trees, goes a few steps further.
 

RELATIONAL DATABASES AND HYPERBOLIC TREES

TheBrain Technologies [http://www.thebrain.com] is a relational database that uses a similar presentation to hyperbolic trees with some important distinctions. Firstly, the structure, while it is comprised of nodes in a hierarchical tree, is not a hyperbolic structure mapped to a circular display area. Rather, the application displays the primary node in the center (like a hyperbolic browser), with several branches leading to other related nodes. Secondly, each node has the capability of having links to other nodes outside of its particular branch, characteristic of a relational database. Whenever a user selects a particular node, it automatically moves to the center and a new set of nodes relating to the selected one fans out from there. This differs significantly from a hyperbolic browser in that the user is able to see not only the nodes that are members of the same branch, but also any other nodes that might be related. For example, using the Web engine developed by TheBrain Technologies, called WebBrain [http://www.webbrain.com], users can select from sixteen primary subject areas. As an example, I selected computers. Clicking on the "Computers" node moves it to the center and a wide variety of other subjects relating to "Computers" fan out around the display. At all times, the node clicked on to move to the next level down remains situated above the node currently being explored to allow the user to backtrack without getting lost. 

Another navigational aid involves the use of mouseovers to highlight the connections between nodes. Choosing "Computer Science" as the next subject area down from "Computers" provides 23 more navigational options. Moving the mouse over each node highlights its connection with the originating "Computers" node as well as any other relationship with which the selection is associated.

WebBrain also incorporates a more traditional keyword search option on a split screen in recognition of the limitations that subject tree browsing entails. Ultimately, the technology demonstrated by TheBrain Technologies is more sophisticated than the hyperbolic browser.
 

THE FUTURE OF VISUAL INFORMATION STRUCTURES

Unfortunately, I predict that it is unlikely we will see hyperbolic browsers or visually orientated relational databases replacing traditional search engines any time soon. Current conventions for searching and information retrieval are so ingrained that will take an enormous cultural shift among information professionals to pave the way for graphically oriented search tools. 

Studies have shown that users favor retrieval engines employing information visualization techniques over their textual counterparts even though user searching success rates between the two are very similar. And yet, the effort to design interfaces that cater to users' desires rather than to those of the information professional is a constant struggle, usually with the user on the losing side. Indeed, my informal research into the opinions of my colleagues regarding both they hyperbolic structure and the relational database revealed an overwhelming disapproval of both. The most oft-cited complaint with each example was the absence of a sophisticated text retrieval engine with Boolean search capabilities. 

On the bright side, information visualization techniques, aided by improvements in processing speed and graphics, are gaining a foothold. Slowly, users are becoming turned on to the advantages that visualization can offer.
 
 

Comments? Email letters to the editor to marydee@infotoday.com.