cone trees: animated 3d visualizations of hierarchical information

CONE TREES: ANIMATED 3D VISUALIZATIONS
OF HIERARCHICAL INFORMATION
George
G. Robertson,
Jock
Xerox
Palo
3333
D.
A4ctckinlay,
Alto
Research
Coyote
Palo
Alto,
415-494-4755,
Hill
CA
ancl Stuart
Road
94304
rol)ertso]l.j)a,rc@
xerox.
colll
ABSTRACT
ing technologies
inlation
cognition”
tasks,
technologies
information
can
iion
offer
3D visualization
potential
solutions
when
the
structure
ized.
We
describe
iion techniques,
visualizing
erarchy
of
screen
whole
structure.
sonle
of the
tual
the
Tree,
enable
cognitive
be
gous
to Scientific
SenlNet
visual-
load
is used
for
The
The
hi-
ploitation
effective
visualization
to the
these
when
be visualized
an-
“large
scale
structure
thus
Information
techniques
interactive
these
the
ancl
of the
exploitecl.
We
V2suakzation,
analo-
Visualization.
[2-?]system
is an
early
of 3D visualization
use of
The
of the
l;llolvledge
structures
and
(,o shift
~esult,s
human
percep-
understanding
tended
in SenlNet
\Yt=re oftjeu
to be cluttered,
the
exanlple
of’ the
of infornlation
visualized
bases,
is used
In this
visual output strategies,
inierjace
metaphors,
graphic presentahlethods:
analysis of contions, screen layout. Analysis
Domain
Specific
Designs:
tenis of particular
domains.
inforntaiion
retrieval.
User-Interface
Design
Issues:
paper,
were
the
mostly
largr
graphs,
cognitive
!vas still
ex-
structut’es,
arl]itfrary
and
structure
structures
archies
are
quite
tures
to
Visualization
allows
of intellectually
exploiting
tion
the
and
hunlan
in the
we
deal
with
Infornlation
access
infortnation
spaces,
is there
and
how
systeill,
an
lnanagenlent
because
the
at
the
various
parts
3D
The
lask
of
diKicult.
using
I,i\”e Iotld
aninla-
Permission
to
copy
title
of the
that
copying
Machinery.
and/or
01991
without
fse
of
exists
Our
infor-
tures,
and its date
is by permission
To copy
otherwise,
of this
specific
permission.
ACM 0-89791-383-3/91
of the Association
or to republish,
/000410189...$1
to
aid
by exploi(il~g
aninlated
called
called
describing
what
this
Enlerg-
material
notice
Ml
in
(lie
graphs
can
links),
is further
en-
in addition
iufolnlatio]l
Ilnallag{:rlleat,
aninlation
hLIIIIa
difstruc-
auxiliary
animation
r(xll[ccs
II ])ercep(ual
to
paper,
is
for Computing
requires
e fee
.50
by
niques
contribute
the
Cone
look
coglli-
system
to
Visualizer
vlsualizatioli
driven
by
the
ancl
human
indicate
the
and
mechanisnls
for
3D
tech-
cognition
a f~anle\~orli
TIIP
of multiple
interactive
these
scale
FRAMEWORK
Coprocessor
managenlentl
in
nlanagement.
tecl~niqucs.
smooth
addi-
Trees
perceptions
where
large
provides
Cogntt/vc
In
of Cone
VISUALIZATION
lag
supports
ancl
struc-
in a prototype
feel
of the
solving
access
develol)
agents
hierarchical
Vtsua/i.~e7[10].
and
some
Trees
in infortnation
which
of
is inlplenlented
Info rmat~ on
we describe
evoked
cludes
1$39
Trees,
the
INFORMATION
Informat,iou
‘k
is given
visualization
Cone
tion
to visualize
and
arbitrary
visualization
hlterclctiv~
Hier-
nlany
infornlation
(with
use interactive
~ ‘1
L.ls,>
systenl,
rate.
are related.
appear,
of hierarchy
in large
astonishing
that the copies are not made or distributed
for
advantage,
the ACM copyright
notice end the
publication
cases,
in
recognition
volume
all or part
good
iafor-
graphs.
appearing
are
sonle
hierarchical
arbitrary
hierzwchies
visualization
access
task
granted
provided
diract commercial
In
I$re also
and
is difficult
it is hard
to
sense
set of problenls
where
expands
and
nlake
hence
into
utility
than
ubiquitous,
exploit.
It, does so by
cognitive
A sinlilar
world,
to
collections,
to stinlulate
data.
information
mation
scientists
data
perceptual
visualization
of patterns
the
large
almost
be transforxnecl
hanced.
on visualizing
rather
applications,
so the
Scientific
we focus
nmtion
ferent
INTRODUCTION
in
especially
to
system,
KEYWORDS:
out
and
solutions
Visualtza-
which
aninlation
potential
call
structures.
to nlaxilnize
and
Interactive
user’s
can
for 3D visualization
ofler
especially
Information
Cone
in 3D
aninla-
problenl,
information
space
interactive
information
of these
called
is presented
to this
the
one
hierarchical
available
and
Card
(lenter
The task of managing
and accessing large information
Enwrging
spaces is a problem in large scale cognition.
for
K.
fralll
is
.4rc/)21ect7~7e[ 10],
asynchronolls
animation.
navigation[7,
for
F\YOL’li
6] and
1t
in-
object
manipulation[7],
Interucttue
visualizers,
and
3D
Objects
Roonzs
for
for
building
managing
rotates
the
so that
path
information
to
workspaces,
from
the
the
the
front
and
substlruc~Llre
Tllissystelll
fact,
issilllilar
could
it works
enters
into
navigation
system
nature
glove),
fice
workers,
The
for
than
wear
user
is drawn
3D
with
information
makes
walks
The
of a physical
to learn
management
a visualization
substructure
The
11~-
display
of the
(of-
to the
replaces
of 2D and
rooms
aucl
information
or
workspace
is a
familiar
a collection
trieval
access is made
called
mechanisms
Level
systems
call
this
in the
make
a Retrieval
supports
Top
retrieval,
back
search
The
prototype
desktop,
editing
often
through
much
operations
Plate
4 shows
sider
Level
system.
form
of relevance
text
databases[l].
the
does
2 shows
not
text
the
in
Cone
The
the
tree
Llser under-
Tree,
fit
the
aspect
is shown
Plate
ratio
only
for
the
3 SI-LOIW an alterll:Lti\’e
which
for
appearance
is done
animation
load
what
to the
would
is horizontally
each
nocle.
after
with
or~-
Robertson
selection
llocle
are
Top
However,
I\’e
track
visualizer
complex,
perceptual
if node
the
the
rotation,
color
rather
tball
some
of the
user’s
constancy
Since
take
allo!vs
the
enables
the
is colnpleted,
Llser
rota-
seconcls
substructures
percept~lal
s}stelll
to
track
about
it
no time
is neecled
to
Of Ok-
ph(,UCJlllellOU
thinking
the
the
several
between
‘~h(? peL’[. epfLla]
Com
chsplayecl
animation.
Llser woulcl
~vithout
system,
selection
relationships
animation
relationships
imation
to shift
without
the rotatiom
ject
feecl-
is used
human
happen
structure
to reassimilate
re-
as Eclitor
in the
underlying
cau
IS clone
to help
chsplayed
highlighting
rotated
ubiquitous.
Each
sees
is completed.
hence
Canl
has text
cognitive
except
by embedding
the
and
Interactive
of windows.
ubicluitous
text
well,
Robertson
ented
tions
Information
user
system
text.
Workspace
system[5],
very
path,
where
metaphor
system.
Rooms
Plate
rotation
of node
cards
selectecl
to be willing
these
use the
Robertson
a selection
so the
perceptLlal
each
short-
by the
animation.
information
and
is similar
after
of
the
transformation
relationships.
likely
This
second.
entire
in the
rotations
animakcl
stand
represent
space.
the
node
are bL’oLIg]I(
follo\viug
the
(ahel-
world
a collection
which
are
at a rate
continuously
an audience
aroLmd
and
l-he
llarallel,
can also be rotated
Tree
layout,
artifacts
it easy
that
is not
in
reality
the
for
supports
user
structure.
simulation
that
system
The
interacts
is intended
top
engaged
equipment
into
each
to the
andis
equipment.
Rooms
roorm.
special
one
and
up
artificial
cues and interactive
example)
special
as in
path,
Typically,
about
The
done
transformation
track.
environ-
(mouse).
3D environment
visualization
to wear
3D
the
workstation
the
node
node
highlighted.
ale
est rotational
(in
I-iowever,
Dinput devices
manipulation
of perceptual
formation
3D
12
the
and
s, but rather
and
ealitysyste1lls[3]
ealitysystem).
conventional
hconventiona
user still
met
rtificialr
in more
nlentswit
in
toartificialr
mninana
selected
selected
~LlbStr LICtLIL’e
When
for
the
am
reassimilat-
ion.
mation:
has visualizers
unstructured,
linear
cal structures.
Linear
ing a 2D layout
onto
called
the
with
classes
and
of infor-
three-seglnent
Wali[8].
The
hierarchi-
I-Hierarchical
hierarchy
use of available
3D wall,
the whole
using
structures
fit
Trees.
Cone
Robertson
Tree.
top
of
Nodes
the
Plate
are drawn
hierarchy
in
1 is a snapshot
like
is placed
3x5
index
near
the
three
of a simple
cards.
ceiling
to be broad
tencl
aspect
).
A 3D
reformation.
drawn
below
the first,
To see this
aspect
ratio
of the
layer
has
divided
by
each
level
that
the
The
body
the
cone
of cones
Ifrheu
cones
of
the
tree
are
of each
is easily
behind
a node
the
same
in
children
to fit
height
depth).
reduced
bottom
their
Cone
the
(the
base
a progression
layer
fits
cone
is shaded
perceived
in
the
yet
width
in cones.
The
room.
Each
room
which
does
not, block
for
1 levels
base
ratio
room
much
so that
the
\ie\v
ancl
fixed
with
the
mouse,
the
Cone
for
Tree
190
the
2D
and
trees
b =
3),
to fit
Ill
the
may
the
diameters
to fit,
The
a t~pical
aspect
ratio
screeu
of the
lhe
ratio
thete
is b, the
is b~–l/l,
factor
and
is
Figure
1
faclors
‘~rce
line
l>ottoln
the
to three,
(b = 2
asi Jcc( rat)lo
Ie\el
near
of
Aspect
gets larger.
1>)’ adjusting
of four
wldt,h
exponentially,
l]laj]chmg
Cone
~vith
If
nodes
ratio
uearly
little
tlLe aspect
factor
aspect
fo~ slnal]
contra st,,
Ivlth
the
comic]er
branching
hierarc-
2 D layo L~ts
a line
to till
size
of the
applica-
t~rpical
for
notl
scroll
image
in real
‘1’his
1001: hke
the
increases
room
to either
to be problematic
Llses clepth
of
structure
s[!l] \vould
have
encountered
~rorse as the branching
happem
same
algorithm
analytlcallyj
and
of the
would
effective
visualization
Llse a size-reduced
ignoring
is b~-l
S11O!VSlvhat
it.
is selected
are
layout
effect
tree,
the
insLwes
of the
transparently,
of a 2D
height
diameters
image
cletail
enable
allcl shalloiv.
tencls
(a size-reclucecl
the
layout
hierarchies
ratio
to maximize
and
user
or
tions
more
is fixed
layout
hlostl
and is the apex of a cone with its children
placec]
spaced along its base. The next layer of nodes is
with
The
structure.
room,
evenly
tree
graph
screen.
the
in 3D
space
A 2D layout
conventional
hy
The
of
screen
structure.
on the
through
CONE TREES: BASIC APPROACH
laid out uniformly
Cone Trees are hierarchies
dimensions.
is presented
by fold-
are visualized
a malleable
Cone
three
structures,
structures
Perspective
are visualized
for
height
of Figure
Although
is
ancl cone
1 is
fkug
120,00
Interactive
1
1
1Woo
80.00
animation’s
scribed
above,
ploiting
the
f=2D, b-2
60.00
more
enjoyable.
i+2D,
—’
By
b.3
the
observing
user
systems,
insights
20.00
‘Ile
basic
Figure
1: Aspect
Ratio
of 2D
and
tion,
3D Trees.
the
aspect
makes
ratio
that
Cone
branching
introduces
can
a limitation
be effectively
Trees
independent
factor,
and
on the
displayed
of the
number
number
of levels
10),
their
the
limit
USER PERCEPTIONS
eral
other
Cone
user
Trees.
Cone
tiou,
perceptions
For
Trees
selected
path
paths,
both
because
spective.
Our
rotation,
object
reconfigure
The
user
fisheye
of
tion
results
and
Finally,
These
depth
cue
is
cues,
also
Robertson
formation
shadows
mation
also
the
different
projection.
While
the extra
information
in understanding
nodes
shadows,
Plate
Trees
to focus
provide
operations
by
a
shadow
in the
do not
structure,
from
top
of
The
seem
onto
it still
pe~haps
of their
In
this
sec-
that
can
techniques
trees
fo~
clynamically,
information.
ancl
These
manipulate
the
visua.lizecl.
appears
prune
visible.
nocle
back
away
from
to manage
the
tab
(lro\\~-
vie\v
is
top
node
clone
or
on one particular
menu
node,
f’lom
a qrmu
Trees),
into
To focus
tows.rcl
hiclc[eu
Can]
of
Prune
or by ges-
so that
These
view
a nocle
are
for
others
We
the
by menu
is mocli[ied
tab.
of hidiug
operations.
clescendants
of the selected
applications
operation
leaving
prunes
only
gardening
and understand
where
of the
for
In
the
de-
operatiom
large,
in-
nocle
to a new
infor-
mentioned
complex
to
to help
continue
statts,
subliminally.
191
the
target
earlier,
rooted
by highgrasped
new
position.
takes
\vorlf
all nodes
place
into
ill another
allimat, iou
during
are made
the
aucl
search.
invisible.
is
provided
Search is initiated
by menu
to currently
visible
nodes,
il~ter active
to
a~ that
the
retrieval
typed
The
When
are
or parameters
are
Tfee.
is provicled
information
parameters
the
in [6] are med
cone.
to its
change
Cone
substructure
Feeclback
of the
to
techniques
described
the
it is moved
typically
degrading
on
drag
Search
selection
wishes
rearranging
position.
is released,
eration
‘2D
aud
the bocly
erty sheet.
is restricted
user
clynamically,
techniques
a nocle
node
the
hierarchy
directly
grasp
available.
Tree
directly
the
it easier
growing
If a wser flicks
node
grow
make
As
infornla-
direct
that
ways
substructure.
and
to its right
on the
substructure
need
gardening
of a pruuecl
sired
often
pruning
node
by flicking
node
a 3D
conveys
users
its
it (or
the chilclren
lighting
the
Trees.
of
p~uuecl
3D manipulation
infor-
of a simple
to focus
gain
SEARCH
inducting
are clone either
all
belotv
needecl
closer
Cam
more
form
to
operations
being
at a node.
node,
structure
the
provide
hierarchy,
CAD/CAM
hierarchies.
perception.
Cam
in
clescribed.
explore
called
operations
directecl
all the siblings
in a nunl-
additional
and
sLlb-
aucl some
on a particular
for
collectively
substructure,
quickly
structural
3) convey
in shadows,
the
be-
select
cast
which
This
hierarchy
users
are
additional
1, the Cone Tree
clustering
in the
the
hierarchy,
parts
by clicking
transfornla-
depth
Trees,
information
For a complex
either
also
enhanced
coloring
3D
and
hierarchy.
(Robertson
about
and
easily
distance
lighter
been
further
selected
ing
use of 3D per-
perspective
the
between
AND
additional
Cone
to
of the
to
structure,
objects
Trees
unclerlyiug
appears
other
structure
to reflect
like
Cone
view
is enhanced
enhances
convey
for
Plate
about
the
have
several
wser
the
user
structure.
through
searching
the
used
rotate
and feel of Cone
parts
view.
The
SemNet~2]
can
to
restructuring
For
3D
are idealized
about
have
changes
of cones
and
user
of
interest
that
is further
the
Obviously
of
larger
their
of 3D depth
links,
floor.
from
effect
it.
shadows
mation
view
and
Lighting
nodes
and
is also
ways
of trees,
the
infornm-
mechanisms.
lighting.
effect
of interest
in size
the
3D perspective
to highlight
ways.
viewer.
closer,
because
perception
ber
tion
view
of
of the
relationships
viewing
tures
of
view
a degree
view
and simulated
a fisheye
effectiveness
view[4]
the
on
and grow
sev-
perspective
describe
of the
mentioned,
the
3D
fisheye
is brighter,
selection
new
to
general
cause of coloring
reported
the
a fisheye
having
as in
already
enhance
example,
provides
without
function,
effects
completely.
rotations
properties
the tree,
to perceptual
more
structure
making
of nodes,
(until
is reached).
In addition
the
MANIPULATION,
look
structure
it
helps
addition,
be pe~formed
allow
number
(about
it
provicle
we describe
and
of levels
brings
In
technique
perceptual
also
It
life,
This
GARDENING,
0,00
by ex-
to
into
into
as de-
load
system.
information
which
effect,
information
insights
structures.
40.00
the
continuous
gains
perceptuzd
of cognitive
perceptual
ancl
understand
COIMTW:
reduction
human
interface
tasks
~—
,*
the
primary
is the
always
either
a pop-Lip
command,
Tl]e search
process
LO alloiv
lVhen
During
by
propand
op-
~c) avoid
(11c
Lumr
a search
the
search,
a progress
mains,
indicator
and
visible.
whose
size indicates
Plate
A
5 shows
with
search
node
of the
the
node
a red
score.
file
name.
with
the
highest
for
reassimilate
the
formation
chy.
are numerous
Trees.
will
EXAMPLES
structures
We
have
that
large
can take
developed
use to illustrate
with
three
these
hierarchical
advantage
of Cone
applications
which
we
tem
techniques.
rotation
first
used
to present
example
representing
mation
access
We
have
hierarchy,
also
file
visualized
files.
To
has
ever
our
file
and
Infor-
Unix
is the
Unix
surprisingly
The
plex
added
structures
time
Using
system.
shallow
our
size
and
current
of the
The
browser.
Search
each person
is
is done
(e.g.,
title
phy.
We have
The
largest
executives.
is the
first
in
Users
biographies
organizational
similar
implemented
several
the
Since
requires
time
the
search
and
for
a selected
contained
this
of facts
location)
can
to
structure
a database
or office
of autobiographies.
with
a
top
other
Xerox
people
chart
operating
plan.
program,
and
highlights
all
from
projects
ing
previous
could
the
to
plan
portfolio,
project
of milestones
and
achieve-
a selected
definition
each
with
A
are
typical
search
project.
used
to reorganize
early
the
Tree
browsers,
of potential
module
plan
document
structure),
applications,
management,
feet
of screen
the
aspect
the
2D
and
local
area
(library
structure
net work
browsers.
there
are several
we can make,
and
observations
several
and
and
issues
the
effectiveness
It
is
10 layers,
also
appears
unbalanced
is hard
these
hier-
to track
be-
In pracshallolv
and
limits.
conclusions
the
analysis
must
argued
another
form
will
It
in
displayed
possible
graph.
in another
the resulting
tree.
For
technique
with
the
user
of’
ancl
it is sizeHo\vever,
Cone
formal
and
and
a depth-first
the
a Cone
graphs
of
Tree
is
evaluation
reveal
Many
to mind,
then
this
(unless
l?uture
question,
is also
through
cycles
The
hierarchy,
obscured).
rotation
three
considering
the
actual
layout.
come
here.
and,
this
the
ignores
about
earlier.
get
are hierarchies?
hierarchies
the
that
which
surprising
at once
details
has 292 nodes.
take
through
of scrolling.
a 3D
ubiquitous
is not
of
organization
layout
would
structure
resolve
of using
layer
2D
described
the
2D layout
650 node
largest
scroll
on demand),
for
192
ratio
be
could
cles,
raise.
an optimized
In the
This
see the whole
graph
use experience,
tentative
we can
to
displayecl.
are broad,
structure
space!
layout
scribed
and
this
it
and
demonstration,
limits
of substructures.
does
above,
then
How
DISCUSSION
on analysis,
COm-
CAD/CANl
nodes,
for
within
take?
but
ple
class
space
reduced,
studies
to a
includ-
object-oriented
management
fall
its
of 30.
tree
an over-optimized
placement,
cut
Based
screen
structure
link
gains
a number
factor
effective
hierarchies
thus
are
1000
balanced
many
mentioned
just
stages
structure.
are
much
cannot
finds
Cone
during
watching
in the
and
about
are more
and
same
chart
be seen in
a company’s
describe
from
it
relation-
be effectively
appearance
however,
How
this
are augnlented
year.
mechanisms
software
book
and
Anaid
charts.
on paper,
to visualize
narratives
related
of operating
desired
Trees
statements
the
manipulation
There
Text
project,
(brief
ments)
Cone
than
A rotating
Assuming
used
than
is the
structural
there
can
branching
Trees
unbalanced,
one visualization.
We also have
that
times
animation.
known
cluttered
more
of uniform
tice,
biogra-
Corporation
80 pages
organization
cause
a database
person’s
Cone
archies.
about
organization
650
that
that
becomes
with
or a maximum
example
sys-
optimal
it is clear
shorter
complex
techniques,
hierarchy
display
reduced
second
the
animation
gained
slow
community.
unbalanced.
The
While
without
is well
Cogn-
perceptual
is unknown,
insights
rotate
the
in much
interactive
trans-
in the
We typically
movement.
time
with
rotational
yovernor
[lO]).
result
tens
hierar-
and
first
file
win
of the
the
the
trans-
a simple
and perhaps
so that
in understanding
ships.
directory
600 clinxtories
this
provides
hierar-
well
a tree
from
complexity
(using
reassimilation
clear
after
to animate
the
to cognitively
is clear
seconds,
animation
work
cognitive
other
contents.
an entire
an entire
was
file
is
nocle
system.
directory
knowledge,
hierarchy
Tree
each
user’s
about
visualized
Cone
with
names
one
contained
10,000
directory
a Unix
1-5 show
which
anyone
The
on
The
structure,
in
is done
Plates
browser.
directory
a directory
Robertson
chy.
is a file
the
While
it
one second
with
which
system.
it takes
Architecture
track
rate
rates
The
easily
perceptual
time
on the
it takes
to about
can
shifts
it is many
can be fixed
ani-
animation
relationships
that
time
is interactive
that
animation,
Coprocessor
it down
the
depending
The
itive
technology
human
structural
without
formations
applications
this
to demonstrate
to the
demonstration
search
in
to quantify
of seconds
APPLICATION
information
load
it is difficult
is selected.
There
win
It is easy
cognitive
bar,
When
clearest
mation.
Robertson
in a file browser
in the
The
re-
by making
with
search
of a search
“cone”
search
are shown
is highlighted
result
string
is completed,
score
much
results
its relative
the
the
how
search
nodes
files
shows
partial
only
Tree
arbit~ary
may
not
a few
to
have
take
(like
cut
links
dotted
can be used
graphs
of simbeen
any
traversal,
If the
form
examples
some
with
be satisfactory.
cycles, this may
de-
directecl
detect
and
are saved
lines
shown
to visualize
a lot
of cy-
IIowever,
work cluite
A
better
adaptable
well.
for
characterization
Cone
Tree
of the
display
class
[4]
of graphs
is a topic
for
future
work .
To
summarize,
mation,
we believe
the emerging
animation,
the human
fectively
exploited
to
of large
information
clear
for
that
processing
seems
plausible
to the
to maximize
user
studies
not
needed
It
can effectively
use of screen
to verify
shift
that
and
3D
space.
expand
16-23.
Generalized
Human
New
Henderson,
D.
243,
of
York:
.4.,
fisheye
Factors
1986.
Card,
S. K.
workspaces
in a window-based
ACM
Transactions
July,
SyS -
Rooms:
the
to reduce
graphical
on
Proceed-
Computing
ACM,
and
virtual
views.
in
space
user
Graphics,
5,
inter3, 211-
1986.
seems
system.
proven)
iems,
face.
access
class
work.
perceptual
yet
effective
are
is a large
of CHI’86
contention
can be efand
G. W.
ings
use of multiple
interactive
system
techniques
animation
load
3D and
There
these
[5]
of infor-
management
spaces.
(but
used
structure
perceptual
improve
which
interactive
nitive
the
technologiesof
and
applications
that
Furnas,
[6]
cog-
And
it
can
be
J. D.,
ings
S. K.,
Texas,
21,
and
movement
SIG GRAPH
(Dallas,
Graphics,
these
Card,
controlled
workspace.
Formal
on
Mackinlay,
G. Rapid
’90
1990).
1990),
G,
a virtual
Conference
August
4 (August
Robertson,
through
3cI
Proceed-
In
Computer
171-176.
conclusions.
[7]
hlackinlay,
J. D.,
G. A semantic
devices.
190,
References
[8]
[1]Cutting,
for
D. R., and
dynamic
ings
Pedersen,
inverted
of SIGIR
.$eptember,
’90,
index
Brussels,
J. O. Optimiza.tions
maintenance.
Belgium,
ProceedACM
Human-Computer
Fairchild,
W.
Semnet:
tions
and
iion,
of large
its
M.,
integrated.
[9]
Poltrock,
S. E.,
three-dimensional
knowledge
applications
Guindon,
graphic
bases.
for
R. (cd),
and
Furnas,
human-computer
Lawrence
G.
Erlbaum,
Foley,
entific
J. D. Interfaces
American,
for advanced
October,
computing.
space
G.
of input
5, 2-3,
145-
G.
detail
G.,
and
and
context
C;ard,
S.
smoothly
Proceedings.
ton,
E. B. Automatic
Technical
of Computer
Seattle,
layout
of large
Report
No.
Science,
University
Washington,
directed
88-07-08,
July
Depart-
of Washing-
198s.
science
1o] Robertson,
i7~terac-
J. D. l’he
1988.
Interactive
[3]
Robertson,
Interaction,
Robertson,
wall:
CHI’91
Messinger,
ment
representa-
In Cognitive
and
Press,
1990.
K.
J. D.,
Perspective
graphs.
[2]
S. K.,
of the design
1990.
Mackinlay,
K.
Card,
analysis
ACM
Sci-
Williamsburg,
194
User
SIGGRA
Software
1987.
CT. G.,
Carcl,
Cognitive
and
S.
Interfaces,
PH
K.,
and
Coprocessor
Symposz
Teclt nologyj
Virginia,
Mackiulay,
.4rchitecture
In
P)oceedzn,gs
um
on
November,
10-18.
User
for
of the
Int erjace
1989,
AChf,