Periscope – A System for Adaptive 3D Visualization of Search Results

1
Periscope A System for Adaptive 3D
Visualization of Search Results

• By
• Wojciech Wiza, Krzysztof Walczak
• Wojciech Cellary

Presented by Gautam Nimmagadda
2
Authors
3
Wojciech Wiza, Ph.D.

• Poznan University of Economics, Poland
• Department of Information Technology
• Research Areas
• Virtual Reality
• Computer 3D Graphics
• Information Visualization
• Internet Intranet Systems
• Search Systems
• XML Applications

4
Krzysztof Walczak, Ph.D.

• Asst. Professor, Poznan University of Economics,
  Poland
• Department of Information Technology
• Research Areas
• Virtual Reality
• Multimedia Systems
• Interactive Television
• Internet and Intranet Systems
• Databases

5
Wojciech Cellary, Ph.D.

• Poznan University of Economics, Poland
• Head of Department of Information Technology
• Research Areas
• Internet Technologies
• Multimedia Electronic Business
• Information Society
• Database modeling of Virtual Reality

6
What is the problem under study?

• Presentation interface of the search results
• Lists of resources found with a short description

What is the problem with the current interface?
The current interface is not the best for
Information Search Tasks as a user is looking for
holistic presentation of the search results
7
How else can we present the results?

• Graphical representation of data
• 2D map or a 3D scene

The authors choose 3D scene to visualize the
search results as they believe it provides better
user interaction, user cognition greater
information capacity
8
Related Work in 3D Visualization

• Visualization projects of Large Data at HCIL
• Visualization of Massive Newsfeed in 3D by Rossi
  and Varga
• NIRVE project at NIST
• VR-VIBE at University of Nottingham
• ViOS by Julian Lombardi
• Starlight System at PNNL

9
NIRVE (NIST Information Retrieval Engine)
Visualization of Search Results A Comparative
Evaluation of Text, 2D, and 3D Interfaces(1999)
by Cugini et al.
10
NIRVE (NIST Information Retrieval Engine)

• Visualization of Search Results A Comparative
  Evaluation of Text, 2D, and 3D Interfaces by the
  NIRVE team
• Study showed that the available features and
  the required tasks seemed to matter more than the
  dimensionality of the visualization

11
VR-VIBE
http//www.emptiness.org/vr/vrvibe.html
12
ViOS (Visual Internet Operating System)
http//www.howstuffworks.com/vios.htm
13
Problems in the earlier systems

• Only one interface environment
• Limited information capacity

14
AVE Adaptive Visualization Environments

• Automatically selects an interface, from a
  collection of available interfaces, that best
  fits the search results
• Allows user to influence the selection
• Presents a global view by visualizing all the
  results returned by search engine
• Three types of interfaces depending on search
  result volume holistic, analytic and hybrid
  interfaces

15
Periscope System

• Prototype system based on AVE
• Acts as an intermediary layer between a user and
  indexing search engines
• Based on X-VRML (Extensive Virtual Reality
  Modeling Language)

16
Periscope System
Visualization engine coverts the text results to
3D scene Metamodels choose an appropriate
Interface Model for the Visualization Engine to
automatically generate a 3D scene
17
Periscope System

• Metamodels
• An X-VRML program used by Periscope system to
  select an appropriate Interface Model when a
  request for Interface Model change is received
• Uses static and dynamic routes
• Static routes are hardwired
• Dynamic routes chose an Interface Model depending
  upon algorithm in Metamodel, data from DB and
  parameters provided by the user

18
Periscope System

• Interface Models an abstract collection of
  interface models
• Composed of three parts 3D Scene Model,
  Interaction Interface and 2D User Interface

19
Periscope System

• Interface Model Selection
• Based on interface readability paradigm by Wiza
  et al. (2003)
• Facet F is a pair (i, ci)
• i is a visualization dimension
• ci represents visualization dimension capacity,
  maximum values i can have in an interface model

• n is the maximum number of objects

20
Periscope System

• Aggregation tables created to determine which
  holistic interface is to be used
• Table created based on document attributes
• Each aggregation table is built on unique subset
  of the set of attributes of the documents

21
Periscope System

• Interface is selected based on best possible fit
  between aggregation table, interface model and
  mapping of attributes
• Each of the aggregation tables is matched against
  all Interface Models based on some rules using
  the readability properties F (i, ci) n.

22
Sample Interfaces

• Holistic interface
• Two visualization dimensions
• Each sphere represents language of documents
• Each slice represents distribution of documents
  within DNS domains

23
Sample Interfaces

• Analytic interface
• 7 visualization dimensions
• More details of each object
• Useful for detailed analysis of smaller data set

24
Sample Interfaces

• Hybrid Interface
• Holistic view of all documents
• Additional visualization dimension color provides
  details of the files

25
Conclusions

• Periscope
• Provides global view
• Provides novel user interface
• Provides multiple interactive interfaces
• Allows model switching
• Customized attribute mapping
• Supports comparative searches

26
Future Work

• Improvements in interface selection method
• Development of new interfaces

27
Usefulness of Periscope

• Feedback from users that have used Periscope
  confirms that the idea of three-dimensional
  visualization of search results is accepted
  especially as a tool complementary to the
  traditional search engines
• A Method of Holistic 3D Visualization of
  Arbitrarily Large Datasets (2005) by the same
  team

28
Usefulness of Periscope

• No study evaluating the system is presented
• Previous study of NIRVE showed contrasting
  results to the authors claims
• Value of Periscope remains unclear

More useful in data warehousing data mining?
29
Assumptions

• There is a need of the global view of results in
  a search task
• Little justification, maybe more illustrative
  examples?
• Document properties like size and language useful
  in a holistic view of the search result
• Are they properties that users look for?

30
Questions