Information Visualization: Ten Years in Review

1
Information VisualizationTen Years in Review

• Xia Lin
• Drexel University

2
Before 1990

• Static graphical representation
• Graphics are made, not generated
• Graphics do not support interactions
• Graphics illustrate the organization of
  information
• Graphics are used to help the analysis of
  information structures
• Examples
• Maps based on citation analysis
• Semantic term relationships
• Semantic Net representation

3
Around 1990

• Scientific data visualization
• Popularity of Macintosh and Windows
• Availability of computational power

4
Motivations

• For data analysis
• Visual inspection of data properties
• Dimensional deduction
• For graphical representation of large amount of
  data
• Clustering and grouping
• Discovery of hidden internal structures
• For visual interaction with the data
• interactive online searching
• browse large amount of information

5
Motivations

• To utilize human perception for information
  seeking
• Human can apprehend relationships on graphics
  fast and sometime intuitively
• Human can understand graphical relationships that
  otherwise difficult to represent
• To understand/reveal information structures
• Understand information structures help online
  searching and retrieval
• Reveal semantic structures through graphical
  representation

6
Around 1995

• IV for IR starts to get popular before of some
  web applications
• HotSause
• SemioMap
• WebCutter (Mapuciino)
• AltaVistas LiveTopic
• Xerox PARCs research prototypes

7
Expectations

• Most of these systems did not live up to their
  expectations
• Limited success
• No clear advantages over other approaches
• Many are for demonstration only
• not practical
• No instant mapping and visualizing
• Not easy to be understood by the user

8
Lessons

• Applications that Look great do not guarantee
  to have users.
• Visualization tools should reduce, rather than
  adding cognitive loads to the searcher.
• No one feels that he/she has to use these
  visualization tools yet.

9
Problems

• Precision and Clarity
• If all details are shown, the result is confusion
  
• If only selected details are shown, it may be
  lack of precision needed.
• Graphics are often not conclusive
• subject to interpretation
• subject to the cognition of the viewer.

10
Problems

• Structures
• Structures help people understand.
• Structures also disorient people easily.
• Usefulness
• For what purposes is an application created?
• For what purposes do people use the application?
• How usefulness can be demonstrated?
• No theories
• No experimental results
• No practical applications

11
A Successful Story

• Spotfire
• Completed in 5 years from research prototypes to
  commercial products.
• Focused on data presentation for data analysis
• Deterministic, rather than fuzziness
• Usefulness, not just pretty pictures.
• Utilized simple functionalities
• Not the most advanced features
• Practical

12
A Developing Story

• Kohonen Mapping for Data Analysis
• A banking report example
• A drug treatment/development example
• Marie Synnestvedts data

13
Baking Cabling Messages

• FINCEN (Financial Crimes Enforcement Network)
  receives thousands and thousands of messages each
  day from banks all over the world, which one
  deserves more attention?
• Solution
• cluster messages
• identify trends
• Interact with the data
• Samples
• Kohonen net input 243 dimensions, 130 input
  vectors
• Kohonen net output 14 by 14
• Index parameters words appear in at least four
  messages and no more than half of the total
  input.

14
Map of the Suspicious Activity Reporting (SARS)
15
Drug Treatment Data

• WAR (Wyeth Ayerst Research)
• Desired to have a visualization tool for data
  exploration on experimental dr
• Complained about the limited exploration power of
  Spotfire.
• Sent me a sample data for mapping
• When the mapping was completed, the director was
  gone.

16

• Data
• 8624 cases (patients)
• 120 independent variables (treatments)
• Kohonen output 20 by 20

17
(No Transcript)
18
Maries Data

• 488 cases
• 12 Variables used for mapping
• SiteExtrem
• SiteHead
• SiteTrunk
• SiteSubVol

• ThickGroup1
• ThickGroup2
• ThickGroup3
• ThickGroup4

• Level2s
• Level3
• Level4
• Level5

19
SiteExtrem
Level 4
Thick G3
Level 5
Thick Group4
SiteSubVol
Level 3
SiteHead
Thick Group1
Site Trunk
Thick G2
20
Our Current Projects AuthorLink/ConceptLink

• Make it practical
• Make it simple
• Make it useful
• The purpose of visualization is INSIGHT, not
  pretty pictures.

21
Design Objective 1

• Develop visualization tools that work on real
  world data.
• Working with data that have meaningful structures
• Thesaurus
• Citations
• Document collections with good semantic
  structures
• Real time mapping
• Large databases, small visualization areas

22
Design Objective 2

• Develop tools for associative mapping
• Analyze co-occurrence data
• Co-citation counts
• Co-occurrence of terms in documents
• View the invisible
• Reveal "the meaning of associations"
• Without visualization, the meaning could be
  hidden in the data.

23
Design Objective 3

• Develop practical visualization interfaces for
  information access.
• Simple and Practical
• Everyone can use it without much learning
• Useful
• Connecting to good resources
• Focus on contents
• Not pretty graphics
• No additional cognitive interpretations

24
Design Objective 4

• Develop real time interaction for information
  visualization
• drag-and-drop from visual mapping to search
  engines
• real-time feed-back from search engines.
• Mixed-initiative interaction
• The search engine responses to what the user asks
  for.
• The search engine may also conduct searches
  before the user asks it to to do.

25
Design Objective 5

• Develop a flexible system architecture for system
  integration and future expansion.
• Design in Java
• Develop a middle solution that might be ported to
  other databases/search engines.

26
System Architecture
27
Concept Mapping
28
(No Transcript)
29
Author Mapping
30
(No Transcript)
31
Future Research

• Beyond interaction
• moving from interaction to cooperation and to
  collaboration.
• Creating a culture and the environment for
  information visualization
• user education
• hardware and software improvement
• Encouragement of graphical thinking