NaviQue: Advanced Environments for Information Access
Space-Scale Diagrams
Social Filtering of Information
Prosection: Hi-Dim Visualization
Purely Graphical Reasoning
Older stuff:
| The NAVIQUE Information Gathering Environment | |
|
|
What are basic desiderata of information gathering environments? |
| We developed design considerations for the
construction of advanced information environments, and a prototype interface
that attempts to respond to them. The design considerations came from task
analyses of information gathering activities, from changes in the global
information environment, and from advances in human-computer interaction.
These led to a number of desired design properties that are guiding our
prototyping efforts, including a system called NaviQue. It is a visually
rich environment for information gathering and organizing, based on a navigable,
fractal structure of
information, ubiquitous queriability, lightweight interaction with ad hoc sets, and information visualization. The resulting interaction paradigm smoothly integrates more than a half dozen synergies between querying, navigation and organization. gathering environments. |
|
REF:
|
|
| Social Filtering of Information | |
|
|
This was some of the earliest published work on what is now know widely as Collaborative Filtering , used, for example, by Amazon.com to make recommendations. |
| Together with Will Hill, Mark Rosenstein and Larry Stead, we focused on electronic streamlined support for allowing a virtual community of users to help each other find things they want. Our test case was a Recommender system that was startlingly effective in helping people find videos they would like, based on finding people with similar tastes and combining their opinions in a customized way. | |
SAMPLE REF: Hill, W., Stead, L., Rosenstein, M. and Furnas, G., Recommending and evaluating choices in a virtual community of use. In Human Factors in Computing Systems CHI '95 Conference Proceedings, ACM, 1995, 194-201. *** . [PDF] |
|
| Multitrees | |
 |
|
| Multitrees are a new class of structures for information access and reuse. They are between trees and DAG's in generality, and have the property that although they themselves are not trees, the sets of ancestors and the sets of descendents of any node are both trees. In an information setting, these structures can be used to represent both varied contents (descendants) and varied contexts (ancestors) of a node. Since both of sets of descendants and ancestors are trees, they support many familiar tree interface strategies for graphical viewing and browsing. Multitrees also serve as a model for reuse of hierarchical structure, since they can be conceived of as unions of trees that share subtrees: Each user picks a set of subtrees of large hierarchical resource and then spins a customized hierarchy above those pieces for his/her own purposes (e.g., various professors putting together course syllabi from the materials in the library). | |
SAMPLE REF: Furnas, George W. and Zacks, Jeff. Multitrees: enriching and reusing hierarchical structure. Human Factors in Computing Systems CHI '94 Conference Proceedings, Boston Ma, ACM, 1994, 330-336 [PDF] |
|
| Prosection - Hi-Dim Visualization | |
|
|
Question:
Answer:
|
| This theoretical work is aimed at making fundamentally
more powerful visualizations in high dimensions (e.g., 4-10 dimensions).
Such visualizations are useful in multivariate statistical data analysis
and certain graphical explorations in mathematics. Research has involved
carefully defining and understanding the strengths and limitations of two
graphical devices, in particular: using projection to see structure
of low dimensionality and section (slicing) to see structure of
nearly full dimensionality ("low co-dimension"). The most important contribution
is noting how compositions of these two,
prosections, can show structure
of any intermediate dimensionality. We have completed a fairly extensive
theoretical treatment on the mathematical properties of the tools (Furnas
and Buja), and have work in progress with Dianne Cook on a visualization
program based on the theory.
In addition to its direct interest to visualization applications in statistics and mathematics, there are also some general insights into how logical analogs of sections and projections can be used in other interfaces to reveal structure. For example, the SuperBook(TM) interface posts search 'hits' against its Table Of Contents. This can be thought of as a tree "projection" of a section through a high dimensional term/document space. Such a view may be useful for reasons related to the theory developed here. |
|
REF: Furnas, George W. and Buja, A., Prosection Views: Dimensional Inference through Sections and Projections, Journal of Computational and Graphical Statistics, 3(4), 1994, 323-353. (Publisher: Am Stat Assoc, Inst of Math Stat, Interface foundation of N Am) (Special issue focusing on this work, with discussions). [PDF] (1MB) |
|
Space-Scale Diagrams |
|
|
|
|
| The problems of scale presented by very large
information worlds have lead in recent years to a number of strategies
in computer interfaces, including for example, fisheye viewers, the Perspective
Wall, multiple-zoom viewers, and infinite interactive zoom viewers. Space-scale
diagrams give a unifying framework for conceptualizing and visualizing
these interfaces. Roughly analogous to space-time diagrams, they represent
both space and magnification explicitly, have an interesting geometrical
structure that allows the clear formulation of certain scale-related interface
concepts, problems and solutions including: semantic zooming, multi-scale
grids, fisheye views, and an interesting and practical analysis of optimal
pan-zoom trajectories.
We subsequently applied this theoretical work to practical problems -- tools for authoring of and navigation in multiscale environments. |
|
REF:
|
|
| Effectively View-Navigable Structures | |
 |
|
| Certain fundamental properties are necessary for information structures to be effectively view-navigable as they get larger. (View-navigation is essentially the familiar method of moving from one locally visible part of an information structure to another until a target is found.) Fundamental properties needed include: views of the structure must be small, connecting paths must not be too long, and the paths to targets must be discoverable by a Navigator. This work explores concepts useful in understanding these constraints, including that of the viewing graph of a structure, as well as to-sets and residue. | |
REF:
|
|
| Purely Graphical Reasoning | |
|
|
Can we do rigorous, formal reasoning with picture representations instead of strings of symbols? |
| The basic idea of this work has been to explore computation strictly in the "picture" domain. The older work concerned reasoning and deduction. Axioms and rules of inference map from images to images, instead of from strings to strings (as in usual, sentential logics). Deduction proceeds by chaining these rules. Research includes finding good demonstration areas for the concept, working out the necessary meta theory and mechanism (e.g., what are variables, quantification, or unification for images?), and writing a modest size virtual machine (15K lines of zeta-lisp code), called BITPICT, to try out the basic ideas. There are two particular relevances of this work to CHI. First is the understanding it might yield by providing a model of human reasoning about graphical interfaces (see CHI'91 paper, below). Second, the graphical reasoning machine itself can provide a substrate for graphical interactions, assisting a user with certain graphical tasks. (see current research). | |
REF:
|
|
The most well known paper in this set is the CHI'86 paper:
Two 1991 tech memos on some prototype fisheye software:
Furnas, G. W., The Fisheye List-Sampler System, Bellcore TM-ARH-020557 , 1991. 12pps, [PDF]