Interaction &
Dynamic Queries 1
CS 7450 - Information Visualization January 29, 2004
John Stasko
Interaction
• How do you define “interactive”?
Spring 2004 CS 7450 3
Response Times
• .1 sec - animation, visual continuity, sliders
• 1 sec - system response, conversation break
• 10 sec - cognitive response
Interaction - 3 Transformations
• Data (Primary Focus)
− Selecting cases to show
• Visual
− Selecting a different representation
− UI controls for chart/graph/vis
• View
Spring 2004 CS 7450 5
Data Transformation
• Selecting what you want to see
• What are some techniques?
Example
Spring 2004 CS 7450 7
Data Transformation
• Interaction forms:
− Pop-up tooltips (bubble help)
− Direct walk
− (General) Brushing
− Details-on-demand
− Dynamic query
− Histogram brushing
Pop-up tooltips
• Hovering mouse cursor brings up details of item
Spring 2004 CS 7450 9
Direct Walk
• Linkages between cases
• Exploring one may lead to another
• Example:
− Following hyperlinks on web pages
Brushing
• Applies when you have multiple views of the same data
• Selecting or highlighting a case in one view generates highlighting the case in the other views
• Very common technique in InfoVis
Spring 2004 CS 7450 11
Brushing
Same item
Details-on-Demand
• May not be showing all the data due to scale problem
− May be showing some abstraction of groups of elements
• Expand set of data to show more details, perhaps individual cases
Spring 2004 CS 7450 13
Dynamic Query
• Probably most common technique
• Let’s explore more details...
DB Queries
• Query language
− Select house-address From atl-realty-db
Where price >= 200,000 and price <= 400,000 and bathrooms >= 3 and garage == 2 and bedrooms >= 4
Spring 2004 CS 7450 15
DB Queries
• Pluses?
• Minuses?
Typical Query Response
• 124 hits found
− 1. 748 Oak St. - a beautiful …
− 2. 623 Pine Ave. -
− …
• 0 hits found
Spring 2004 CS 7450 17
Problems
• Must learn language
• Only shows exact matches
• Don’t know magnitude of results
• No helpful context is shown
• Reformulating to a new query can be slow
• ...
Dynamic Query
• Specifying a query brings immediate display of results
• Responsive interaction (< .1 sec) with data, concurrent presentation of solution
• “Fly through the data”, promote
exploration, make it a much more “live”
experience
− Timesharing vs. batch
Spring 2004 CS 7450 19
Dynamic Query Constituents
• Visual representation of world of action including both the objects and actions
• Rapid, incremental and reversible actions
• Selection by pointing (not typing)
• Immediate and continuous display of results
From: Ahlberg &
Shneiderman, CHI ‘94
Imperfection
• Idea at heart of Dynamic Query
− There often simply isn’t one perfect response to a query
− Want to understand a set of tradeoffs and choose some “best” compromise
− You may learn more about your problem as you explore
Spring 2004 CS 7450 21
Software Demo
• HomeFinder - Univ. of Maryland
What Did We See?
• Interface
− buttons
− sliders (nominal --> ordinal)
− alphasliders
Spring 2004 CS 7450 23
Query Controls
• Variable types
− Binary nominal - Buttons
− Nominal with low cardinality - Radio buttons
− Ordinal, quantitative - sliders
Alphaslider
A B C DEF G HI J K L M N OP QR S T UVW XYZ Goldfinger
Current selection
Spring 2004 CS 7450 25
Rangeslider
100
Low selection thumb
Real data range
High selection thumb
0
FilmFinder
VideoSpring 2004 CS 7450 27
Spotfire
Current assignment
Spotfire Features
• Starfield display
• Tight coupling
− features to guide the user
− rapid, incremental, reversible interactions
− display invariants
− continuous display
Spring 2004 CS 7450 29
Nice Application
www.myrateplan.com/cellphones
DQ Strengths
• ?
Spring 2004 CS 7450 31
DQ Strengths
• Work is faster
• Promote reversing, undo, exploration
• Very natural interaction
• Shows the data
DQ Weaknesses
• ?
Spring 2004 CS 7450 33
DQ Weakness
• Can you formulate an arbitrary boolean expression?
− !(A1 V A2) ^ A3 V (A4 V A5 ^ A6) V …
• But do people really do this often?
DQ Weakness
• Controls take space!
− How much in Spotfire?
• Put data in controls...
Spring 2004 CS 7450 35
Data Visualization Sliders
Low selection thumb
Data distribution
High selection thumb
From: Eick, UIST ‘94
Drawing Queries
Query Sketch
You specify a timeline query by drawing a rough pattern for it, the system brings back near
matches User-drawn query
Spring 2004 CS 7450 37
DQ Weakness
• As data set gets larger, real-time
interaction becomes increasingly difficult
• Storage - Data structures
− linear array
− grid file
− quad, k-d trees
− bit vectors
Brushing Histograms
• Special case of brushing
• Data values represented in histograms that can be clicked on and selected (controls region)
• When items selected there, the
corresponding item(s) are highlighted in main view windows
Spring 2004 CS 7450 39
BH Example
DataMaps Maryland
&
Va Tech
DQ vs. BH
• Empirical Study
− Use DataMaps, a geographic (US states) data visualization tool
− Have participants do different tasks with both methods
How many states have pop between x and y in 1970?
Given 3 states, which has the lowest median income?
What’s the relationship between education and income?
Demo
Spring 2004 CS 7450 41
Findings
• Brushing histograms better and more highly rated for more complex discovery tasks
− Attribute correlation, compare, and trend evaluation
• Dynamic queries better for more simple range specification tasks
− Single range, multiple ranges, multiple criteria Functioned more as its own infovis tool
Functioned more as auxiliary control for other vizs
Case Study
Attribute Explorer
Video
Spring 2004 CS 7450 43
Fundamental Idea
• Show all the data, use different displays for different variables
• Use brushing so that interaction with one view highlights selections in others
DQ vs. BH
• Fundamental Differences:
− BH highlights data of interest;
DQ filters unwanted data
− DQ does single range query;
BH allows multiple ranges
− DQ users interact with the query (low,hi);
BH users interact directly with data
Spring 2004 CS 7450 45
Case Study - SDM
• Developed at CMU
• Limitations of static visuals
− Unable to focus on different objects with context
− Clutter and object occlusion
− Objects are dwarfed in the scale for the entire data set
− Fail to classify object sets
− Difficult to compare quantities which are not spatially contiguous
Chuah et al, UIST ‘95
System in Action
Video
Spring 2004 CS 7450 47
What We Saw
Object handles Maintain context: Body and shell
Viewing occluded objects Interactive augmentation
Interaction
• More to come...
Spring 2004 CS 7450 49
Upcoming
• Interaction and Dynamic Queries 2
− Reading
Chapter 5 again Fishkin & Stone
• Time Series data
• HW3 due in a week
Sources Used
• Spence, CMS books
• Shneiderman; Ahlberg et al; Chuah et al;
Tweedie et al articles
• Chiu & Wang F99 slides
Spring 2004 CS 7450 51
Extras
• Additional information
SDM Components
• Object-centered selection
• Dynamic and flexible operations
• SDM handles
• Objects constraints and feedback techniques
• Context persistence
• Set-wide operations
Spring 2004 CS 7450 53
Advantages
• ?
Advantages
• Focus and context
• View occluded objects
• View different object sets in different scales
• Visualization is not part of the underlying data
• Compare the widths and heights of objects at a distance
Spring 2004 CS 7450 55
Disadvantages
• ?
Disadvantages
• Object manipulation causes occlusion
• Users cannot operate on parts of objects
• With limited scope, certain objects are occluded by the focus of objects
• Changing the width/height can blur the linear relationships between objects