1. Introduction

There has been a great deal of research recently on techniques for supporting the exploration and manipulation of information. This research has produced new ways to visualize information, new techniques for interacting with visualizations to manipulate information, and new tools for supporting the creation of visualizations. Much of this work has been quite general in producing visualization techniques applicable to diverse data (e.g. quantitative, symbolic/relational, geographic, temporal) and interaction techniques that can be combined and applied across many different styles of visualization (e.g. 1,4,6,7,8,9,10,11,12).

This work has begun to produce several general purpose analysis tools, each with its own strengths. The Table Lens [11], for example, is a dynamic spreadsheet environment for exploring large, multidimensional data sets with techniques for focusing attention on subsets while viewing the rest as context. Its strengths include techniques for rapidly creating and viewing the relations among new data attributes. Another evolving analysis package is IVEE [2], which provides the ability to rapidly create multiple dynamic query sliders to filter data. A third example is the SAGE system [12], whose central feature is rapid design of visualizations that integrate multiple attributes.

Taken together, these systems illustrate a fundamental user interface design question: how can we use the complementary features of different visualization and analysis tools in a coordinated way. Even for just these three systems, how can we create new attributes in one, filter the same data with another, and visualize the resulting subsets with a third? More generally, what user interface approach would enable people to easily move and combine interesting subsets of information across the isolating boundaries imposed by different applications?

Of course, the coordination problem is not unique to these tools. Most people who work with large amounts of information also use custom applications. For example, in domains like transportation scheduling and tracking (which we have been using as a test case), analysts use one system to generate and display airplane schedules, another for tracking the location of cargo in transit, and a third for managing warehouse inventory and requisition handling. The interfaces to these applications each have useful visualizations but no mechanism to explore relationships among the different data they portray. For example, there is no way to explore the relations among the locations where supplies are stored, the people who order them, and when they are scheduled to be shipped by air.

These problems suggest the need for a user interface environment for people who work in information-intensive domains - an electronic workspace for people who explore and analyze large amounts of data daily. Such a workspace must provide several key capabilities.

First, it requires user interface techniques that enable information to be selected and combined from multiple application interfaces, visualizations, and analysis tools.

Second, it must enable rapid generation of visualizations to integrate information from these diverse sources. The value of integrative visualizations is obvious. However, because the combinations of information that people will create are often unpredictable, it is not possible for software developers to create every visualization in advance. Therefore, an effective workspace must provide tools by which users can create new visualizations as needed without great effort or skill.

Third, consistent user interface techniques are needed with which people can filter, control level of detail, navigate, and create new information wherever it is displayed.

Fourth, an effective environment should make it easy for people to share and communicate their results in collaborative settings, where they must iterate between analysis and presentation activities frequently.


In order to address these needs, we are developing an approach within a software environment called Visage. Our goal is to incorporate basic information exploration components within a new user interface paradigm.

This paper describes several key elements of Visage:


1. A consistent information-centric user interface paradigm.
As the name implies, this paradigm strives to provide users with greater direct contact with objects that represent information they need to view and manipulate to perform their work. In this paradigm, information is represented as first-class objects that can reside and be manipulated in visualizations, application user interfaces, on desktops, in briefing materials, or anywhere else people elect to place it. It is ultimately concerned with usability (i.e. it is user-centered), in that it seeks to reduce the complexity and restrictions created when people cannot access information directly and instead must face the mechanics of running and coordinating applications and working with file system metaphors.

2. Dynamic visualization generation.
In order to provide integrative views of information, we are incorporating work on SAGE, a knowledge-based automatic graphic design tool [12]. This approach provides rapid generation of visualizations customized to users' immediate data exploration tasks.

3. Interactive information manipulation.
These include tools for:

  • Finding and interactively partitioning, filtering, and selectively combining subsets of data on which to focus.
  • Controlling the level of detail with which this information is viewed using drill-down and roll-up techniques (drill-down commonly refers to the process of segmenting or breaking down aggregated data along different dimensions to create a larger number of smaller aggregates; roll-up commonly refers to the process of merging detailed data into aggregates that summarize their attributes).
  • Assembling, laying out, and interactively presenting information to others.


Next Section: An Example

Paper Index

Back to the Visage Home Page