TOWARDS "VIRTUAL REALTY":
EXPLORATIONS IN SPATIALIZING WEB CONTENT

BY BRYGG ULLMER

The following is a paper describing my first-semester work in spatial info environments done in conjunction with my Fall 1995 Software Design Studio project. Early studio presentations, screenshots, and other supporting materials are accessible here. This was originally a CHI '96 short paper submission (limit two pages), but was not published.

You're now at <http://www.media.mit.edu/~ullmer/papers/urbcyber/>. The officially-formatted CHI version is here. My December 3 draft of substantially different form, which included an Implementation section introducing 3wish (cut from the final draft for lack of space), is here.

Any comments would be appreciated -- let me know what you think! Thanks!

Brygg Ullmer (ullmer@media.mit.edu), 12/17/95

Abstract

The paper presents exploratory work developing interactive spatializations of Web content grounded in an urban metaphor. Special attention is devoted to content representation, for which we employ symbolic web-iconics; and to spatial navigation, where we use a monorail metaphor to advantage over purely walk-through or fly-through approaches.

Keywords: Information visualization, WWW content representation, 3D interactive graphics, urban metaphors

Introduction

The page-based paper metaphor for Web documents popularized by Mosaic and Netscape is fairly effective at presenting content in detail. However, the Web's page metaphor has shortcomings for large information spaces -- spaces of hundreds, thousands, or more elements -- surrounding navigation, visualization of context, distributed authoring, and the representation of human presence. Spatial interfaces have the potential to support richer, more connected navigation of large infospaces by presenting complex, symbolically-legible information representations -- in short, information landscapes. At the same time, compelling spatializations of abstract information such as delivered by the Web are few and far between. Even with a standard for representing networked 3D geometries, the current state of VRML net-content -- often a space of bulky geometries with little informational content -- reflects the difficulty in deriving compelling spatializations for abstract information.

We have pursued a concentrated, short-term effort to develop a spatialization of abstract Web content which begins to address the above-referenced page metaphor shortcomings. We have adopted an urban metaphor for this effort, and have focused design attention on the symbolic representation of web-content and on navigational affordances for large spaces of such information.

Urban Metaphor

There are many reasons why an urban metaphor for the Web, as suggested in True Names, Neuromancer, Snow Crash, et al., is compelling as a design metaphor. Physical urban spaces embody sufficient richness to envelop a dynamic, distributed space of many millions of entities without resorting to spatially-discontinuous hyperlinks. They are characterized by emergent form and content over time; are inhabited by many live people as well as dead geometries; and while textually annotated, are largely symbolically legible and navigable. Of special import to the Web, urban spaces are characterized by distributed design and implementation; urban designers are primarily concerned with macroscale form, while legions of architects, builders, and "end-users" shape the finer forms of individual structures. Equally important, urban spaces are defined and critically shaped by the spatial scarcity and consequent value of land, which emergently both supports the mediation of (scarce) user attention and the establishment of a foundation for spatially-situated commerce. Lastly and central to our work, urban spaces have the potential for substantial visible clarity and "imageability," as discussed compellingly by Lynch [Lynch] and many others.

We have constructed online urban spaces inspired by the works of Lynch, the MIT Media Lab VLW [VLW], and the science fiction cyberspace literature. A snapshot is crudely visible as Figure 1. Our current spaces consist of "Jersey barrier" information-bearing "buildings" texture-mapped for distinction, capped by text, and surfaced with the "web patches" introduced under Urban Signage. We employ distinctive geometries such as giant chickens and spinning bananas as visually distinctive landmarks. Districts/zones are represented with texture-mapped groundplane polygons, and loosely corresponds to host spatial Web servers. We employ text in the space partially in the style of [VLW]. And finally, we realize a path/node "transportation infrastructure" in our monorails, described under Urban Transport. The environment is implemented in our Tcl VRML language extension, 3wish, which compactly supports geometry synthesis, behaviors, and Web content-integration in a net-savvy fashion. [3wish]

Figure 1: Urban cyberspaces overview shot

Figure 2: Web patches example

Urban Signage: Content Representation

While textual annotations such as street and building signs are widely present in physical urban spaces, human navigation in these spaces can be seen as largely symbolic; with familiarity, label text may be subsumed altogether into the landscape. We wished to support a similar process in our urban spaces with web-content, partially following the example of [Nygren], by iconographically representing the salient compositional features of Web pages. Our present representation automatically integrates the first few words of Web page titles as bold text heading the "web patch." The rest of the page is symbolically summarized with black/red lines for normal/hyperlinked text, and (uniform size) black/red rectangles for normal/hyperlinked images. As depicted in Figure 2, this allows (for instance) documents which are primarily textual, primarily fragmented hypertext, and primarily hyperlinked images to be readily differentiated. In a larger sense, this potentially enables a visual space bearing hundreds of such patches to evolve from a largely explicit textual space into a largely implicit symbolic space. Web patches are employed in this light in our urban cyberspace, mapped to our "Jersey barriers" as visible in Figures 1 and 2. Clicking on a patch currently invokes a remotely-controlled Netscape child viewer.

Urban Transport: Spatial Navigation

The presence of lucid, intuitive navigational affordances is critical for realizing functional information landscapes. In our opinion, traditional walk-through or fly-through viewers do not provide satisfactory affordances for general cyberspace navigation, as they both occupy the user more with the mechanics than semantics of navigation, as well as lacking support for quickly reaching highly salient views. We have developed the monorail metaphor introduced by Snow Crash as a mechanism for quickly locomoting viewers to distinct destination/viewpoints while preserving the spatial metaphor and offering overview vistas of the space. A view of this device in our environment is visible in Figure 1. Our implementation consists of a series of post-like nodes, positioned at salient locii and orientations in the space; and a series of interconnecting pathways between these nodes. Clicking on a pathway/"rail" causes the viewing camera to be rapidly transformed to an overview viewpoint above the more distant node of the rail. Currently we use both 45° pitch "birds-eye" views, and 90° (vertical) pitch for map-like views, chosen to maximize visual salience at individual locii. Once in an overview position, clicking on a cone located above each monorail post will bring the user from the overview into an immersive view, where a walk-through viewer supports local navigation. A single key-click will rapidly transform the camera back from walk-through mode to the overhead overview camera position, thus reducing the "lost in cyberspace" phenomena.

Summary and Future Work

We have introduced a continuing effort towards interactive spatializations of Web content, with new progress in representing Web content, navigating information landscapes, and developing an urban metaphor for cyberspace. Our continuing work attempts to further develop these features, as well as exploring richer architectural affordances, semantic zooming, naming strategies supporting variable latency/bandwidth operation, and avatar user-representation issues.

Acknowledgements

We would like to thank Terry Winograd, Paul Freiberger, Diane Schiano, and many others at Interval Research for supporting motivating explorations during the summers of 1993 and 1994; to Bill Mitchell and Mitch Kapor for the motivating Fall '95 course project; to Ron MacNeil and the students of the MIT Media Lab VLW for their inspiration and support of the present work; and to AT&T for supporting our 1995-96 studies.