Austin, Texas
May 18-20, 1994

The Research Library the Day After Tomorrow

Beyond Stacks: Navigating The Library Without Walls

Catherine C. Marshall
Xerox PARC

I am here today to talk to you about navigation; in specific, navigating the library without walls. When I tried to represent a library without walls and navigating it, I came up with several different images. Here are two.

First, the cartoon character, Dilbert. He is the quintessential nerd. In one recent cartoon, he is sucked into cyberspace, and flies around with his arms widespread as he navigates through this abstract maze of links and nodes. He says, "It's beautiful." He is interacting with the minds of brilliant people from around the globe; their ideas and knowledge fits together, and it is exhilarating. Of course, when he describes it to his co-worker, she just says, "Oh, so you're kind of a nerd. Right?"

The other thing I thought of when I thought of the image of navigating the library without walls comes from this comment by Nicholson Baker in a recent issue of The New Yorker:

We will be able to tour mind rooms full of 3-Dimensional representations of catalogue cabinets by gesticulating with our data gloves like armchair Shivas, so that we will have some intuitive sense of the size of the collection we are interrogating, just as we do now, especially when we walk into the lobby of a library we haven't visited before and size up its rows of cabinets; we will have ways to maintain a sense of where we are in the database, as we do now through drawer and cabinets labels, through guide cards, and through our subconscious feel for where the restrooms and the circulation desk are... (Nicholson Baker, "Discards," The New Yorker, 4/4/94)

I noticed that there was a letter of rebuttal in the next issue of The New Yorker commenting, "So what does the inside of Mr. Baker's refrigerator look like?"

These two perspectives suggest that there are different kinds of technologies developed for "navigating" a "virtual library." You can look at libraries as physical artifacts, places that hold collections and that have already existing means of access, and you can duplicate that in a digital world. One of the sections on the videotapes I am going show you does just that--the physical world as a digital world.

The second thing you can do is adopt the technological perspective of the Internet as a model for large-scale information spaces, and think of library services as distributed information services for a network. We all know there are some problems with that model.

The third perspective, which is largely under appreciated, is to take a practice-oriented perspective, look at the work that librarians and library users do, and start from there to develop technology, anticipating that technology will change these practices over time.

At this point I want to talk about three technologies and scales for navigation. The first one is 3-D visualization. This provides us a sense of coarse-grained access, where you can get an overview of a large information base compactly on a screen and see indices into places you might want to go in the collection. It also has that feeling of flying around at home. I am going to use Xerox PARC's Information Visualizer -- which is the work of my colleagues Stuart Hart, Doc McKinley, and Joey Robinson.

The second technology I am going to talk about is hypertext. We had some nice examples last night of navigating in a very local, medium-grain sense, where you can do a lot of local traversal and browsing following references using links. Those were really nice ways of navigating over materials you want to know about. I am going to use an example from NCSA Mosaic, since that is the most ubiquitous form of hypertext right now. I never thought I would be talking about World Wide Web--I did not predict the growth very well at all. It looks to me like everyone has already seen NCSA Mosaic, so I will review it only briefly.

The third topic I am going to talk about is information workspaces. This really has to do with the prime way to get any list of materials--what you do when you are handling materials to understand them. It is on the level of a "scholar's carrel" or a community memory.

This first example is about visualizing the coarse structure of the library and the materials in it. I am going to show you a videotape of the Information Visualizer, which uses 3-D color graphics and imaging to use the human perceptual system's ability to process, shape, color, and movement. It helps people to understand relationships among semi-structured materials, and it brings scientific visualization techniques to large information spaces.

The first excerpt on this tape is about something called 3-D Learning. It is really navigation through flying around in a 3-D space, mapping the visual to the virtual, working directly from the artifacts.

A prototype system called the Information Visualizer is used to explore the use of 3-D interactive animation for information access applications. The information visualizer runs in a 3-D version of a room's work space manager. Users walk around in a collection of 2-D and 3-D rooms and interact with artifacts that are abstract representations of information or information structure. There are visualizers for structured and unstructured information.

This system is similar to an artificial reality system, but it is made to work with a conventional 3-D work station and mouse. Office workers are not likely to wear special equipment such as gloves and helmets. Even so, users are drawn into the 3-D environment by perceptual cues and interactive animation. (Xerox PARC Information Visualizer videotape)

Next on the videotape is cone trees. Those are a 3-D graphic representation of hierarchical relationships among a topic structure, using, for example, a card catalog would show topic relationships. It uses visual cues again like smooth animation and drop shadows so you can see what is happening when you are interacting with it. It gives abstract information a physical shape. This is a little different than the 3-D rooms concept I just showed you in that it takes the abstract and maps it to something you can see. Following that is a brief presentation on a perspective wall, which is a fish-eye view of something like a time warp or any kind of data that is indexed over two dimensions.

This visualization, called the cone tree, shows parts of a unit's directory hierarchy. Each node is formed. When a directory is selected, the tree rotates through the shortest path to bring the selected path to the front. The selected path is highlighted with a directory name. 3-D is used to make more effective use of available screen space, but 2-D layout of the same tree would not fit on the screen. The user would have to scroll through the 2-D layout and would not have a view of the whole tree.

The 2-D layout could be reduced to fit, but details would be lost. Interactive animation is used to shift some of the user's cognitive load to the human perceptual system. You can see this by looking at a selection done without animation. It takes several seconds to reassimilate the relationships between the parts. With animation, the perceptual system tracks the relationships because of object constancy.

When the animation completes, no extra time is needed for reassimilation. If we walk into the next room, we see an alternative layout that allows text to be displayed for all nodes. This visualization is called the cam tree. As before, selection rotates the tree to move the selected path up to the front. Highlighting is done with color rather than text. Notice that the display provides a fish-eye view because of 3-D perspective and coloring. The objects of interest are larger, brighter, and closer than the other objects.

Search is done by popping up a property sheet and typing in the search parameter. The simplest search is to look for files with some string in the file name. The result of the search highlights the directories with matching files, showing the relative search scores with the red bar on the node. You can examine a file by selecting it from a menu of matched files. The screen rearranges itself so that the text of the file appears in an editor buffer.

Another way of looking at files is to walk into another room that uses a perspective wall to display the files. The perspective wall is used to lay out 2-D information onto a 3-D segmented wall. The front segment is used for viewing details, and the side segments provide context. In this case, each node represents a file with its horizontal position determined by its date of last modification. Vertical position is used to indicate the class of the file.

When a node is selected, the wall moves to bring the selected node to the center of the front section. As before, we have a fish-eye view, we use 3-D to make better use of available screen space, and we use interactive animation to exploit the perceptual system's ability to track relationships between moving objects. In regions of the wall where there is too much detail, we can stretch the wall like a sheet of rubber to reduce clutter. (Xerox PARC Information Visualizer videotape)

One claim that is made in this videotape--I did not believe it when I first heard it--is that the 3-D representation really makes the information a lot more compact. You can see it really does create a great deal more information in a smaller space. I also wanted to show a little bit more of that user interface that was on the videotape. This is what these rooms look like in overview, and it is another way of thinking of combining a physical metaphor with a metaphor for artifact representations. You can have a lot of rooms and interact with all of them in this miniaturized form.

I like to think about this as accurately finding where you want to be in the stacks. You can take advantage of the interlinkages between materials, between references, and between co-located materials to get some kind of serendipitous exploration of related data. So, as I said, hypertext is the technology in question, and Mosaic is the active standard right now for browsing the Internet. One of the reasons why is that it supplies seamless connections between different sources. I can be off in Lucerne and Indonesia and go to the University of Hawaii, without even realizing I had gone so many miles. It also provides seamless information in a standard data format so you get the seamless and transparent integration of many different data formats. I will put this in briefly for those who have not yet seen Mosaic.

A popular interface to the Internet is Mosaic. Available for work stations, personal computers, and Macintoshes, Mosaic gives direct access to the World Wide Web of information on the Internet. The workings of this system--where data is coming from or what protocol is being used to retrieve it--are invisible to the user, so navigating Mosaic information becomes quite simple. Mosaic information is displayed in documents; all of the underlined phrases are linked to other documents, which can be anyplace in the world. We can follow a linking by simply pointing at that item and can return to wherever we came from using the "back" button.

To initially explore the Net, we can start with an index of entering or starting points that branch off to the rest of the world's web. We can follow the link to the Honolulu Community College and look at an exhibit of a dinosaur fossil, or we can go instead to Switzerland, where a catalog of information is kept. From there we can jump to Japan to look at tourist information. Mosaic is only one of a number of browsers. (Video clip courtesy of Steve Putz, Larry Masinter, Dave Robson, and Jennifer Ernst.)

What I am going to talk about is visual pedagogy or visual and kinetic interpretation of information. It is also a navigable information work space that contains properties of physical work spaces and that works more on the idea of collaborative interpretation rather than individual use and access. It supports the evolution of a graphical language and structure for talking about data online. Some of it is based on empirical studies of analysts and the tools that they use. They are really good consumers of information, so it is sometimes helpful to look at them and how they apply these resources. The whole idea is to use visual symbols, give them properties, and create a hierarchy of information spaces.

[Ms. Marshall concluded her presentation with a demonstration of navigating among Web pages on the Internet.]