InfoVis CyberInfrastructure

1

InfoVis CyberInfrastructure

Katy Börner

School of Library and Information Science

[email protected]

Presentation at InfoVis 2004, London, UK

July 15th, 2004

http://iv.slis.indiana.edu/db

http://iv.slis.indiana.edu/cr

http://iv.slis.indiana.edu/sw

http://iv.slis.indiana.edu/lm

2

Katy Börner: InfoVis CyberInfrastructure, InfoVis Conference 2004.

Motivation

Support developers and programmers in the comparison and distribution of new algorithms.Support (non-programmer) users in the utilization of advanced InfoVis algorithms.Interconnect algorithm developers and users. What algorithms do users need/want?

Provide a unique resource for InfoVis research and education.

Support ‘Knowledge Domain Visualization’ research.


Related Research

InfoVis Toolkit by J.-D. FeketePrefuse by J. Heer, S. Card, James A. LandauUniversal Visualization Platform by G. GrinsteinPiccolo/Jazz by B. Bederson

GraphViz by S. NorthJUNG by S. White Tulip by D. Auber

GeoVista by M. Gahegan, Alan McEachrenCommon GIS by Natalia and Gennady Andrienko

VTK ToolkitIBM OpenDX

What focus should a InfoVis Toolkit have?Fekete’s Toolkit is well suited for the design and customization of complex visualizations (multiple windows, search, etc.)IVC supports the combination/pipelining of data analysis & visualization algorithms.

3






IVC Oracle DB and Search Engine

The Team

Design and ImplementationJay AskrenSaiful BahariAndrew BangertChristopher FriendStephanie GatoTodd Holloway (Lead)

Ruchi KapoorKetan ManeLalitha VisvanathQian Wang

Data ExpertsDaniel O. KutzElijah Wright

Graphic DesignCaroline Courtney

Project Start

September 2003

4


IVC DB System Overview

Oracle/Apache/Tomcat/Java—Well understood and reliable tools

DB featuresPotentially one terabyte of data in flat filesRelational designAllow for more collections to be added

Search engineSearch on abstract, author, title, journal, date published, and moreUser login for both Indiana University (IU) and non-IU usersUser historiesAdministration of data and user accountsCompressed downloading of resultsTerm-by-document and co-author matrices of results


IVC DB Data Sets

5









6


IVC Software Framework

The Team

Master Minds/ProgrammersJason Baumgartner, SLISNathan James Deckard, CSNihar Sheth, InformaticsBruce William Herr, CS

Shashikant Penumarthy, SLIS

Graphic DesignCaroline Courtney, Fine Art

Project Start

2001

Algorithm Development and IntegrationVivek Agrawal, Summer Intern Renee LeBeau, SLISJosh Bonner, CSTodd Holloway, CS Jeegar Maru, CSLaura Northrup, CSSriram Raghuraman, CS Nihar Sanghvi, Informatics Hardik Sheth, InformaticsSidharth Thakur, CS Ning Yu, SLISYuezheng Zhou, CS

Students taking K. R. Subramanian’s (UNC Charlotte) InfoVis class integrated diverse algorithms into the IVC.


IVC Software Framework

Web Site: http://iv.slis.indiana.edu/sw

Jason Baumgartner, Katy Börner, Nathan J. Deckard, Nihar Sheth. An XML Toolkit for an Information Visualization Software Repository. Poster Compendium, IEEE Information Visualization Conference, pp. 72-73, 2003.

7



8



9



10



11


IVC Software FrameworkDocu is at http://iv.slis.indiana.edu/sw

ANALYSIS ALGORITHMS generate models from parsing other data structures and/or processing on the data

LAYOUT ALGORITHMSrun graphical processes on the appropriate model

DATA MODEL PERSISTENCEfactory to persist a model to a particular data store (i.e. XML format, database)

STANDARD MODEL INTERFACES based on Java 2 Swing standard models

CODE INTEGRATIONnew algorithms can be integrated easily by supporting one or more of the models


Framework can run different data analysis and IV algorithms on a standard set of input data formats (tree, matrix, network, table, list).

Models from the algorithms can be serialized through the persistence layer; and it is generic enough for plugging in various persistence options (XML, SQL database, etc).

Based on Model-View-Controller (MVC) by focusing on standard data model interfaces for data exchange.

12


Demo IVC Software Framework

Download from http://iv.slis.indiana.edu/sw/ (code is hosted at Sourceforge.net)

CD containsReadme.txt04-iv-ivc.ppt (this slide show)IVC-www (web pages available via http://iv.slis.indiana.edu/) IVC-source (IVC source code, needs to be compiled using ant)IVC-build (contains an executable jar file and sample data)


Run the IVC Software Framework

Copy CD content to harddriveIn ‘/IVC-build’ double click ‘ivc.jar’ to open the window below

Use menu system to Load or simulate a data set.Analyze, visualize, or interact with a data set.To start a toolkit.To access code reference pages, learning modules, javadoc, get updates, etc.

13


Whenever an algorithm is selected, the user receives feedback on what algorithm was run, what parameters were used, any textual results, who developed this algorithm, etc.The amount of feedback can be customized. In addition, a log file is generated as a permanent track of all user actions.

Each loaded/simulated data set is internally stored as a data model. All data models of a session are listed on the right hand side. Right click a model to rename it.


Only algorithms that can be used with a selected data model are selectable (all others are grayed out).*

*The black shown network visualizations are prefuse demo’s that do not (yet) use the IVC data modules but require their own data format.

14


Currently Available Algorithms

Data ModelingP2P network modeling and search algorithms

- Structured P2P Systems (Content-Addressable Network Model (CAN) , Chord Model)- Unstructured P2P System (PRU Model, Hypergrid Model)

TARL model for the co-evolution of author-paper networks

Data AnalysisGeneral Search Algorithm – Breadth First Search, k-Random Walk SearchCAN Search Algorithm, Chord Search Algorithm

VisualizationJUNG network layout algorithmsPrefuse demos

InteractionPrefuse demos

ToolkitsAW toolkit

In Sept 2004, most of the algorithms at http://iv.slis.indiana.edu will be integrated, see menu.ini.


Sample Exercises

Compare search time for structured and unstructured P2P networksGenerate P2P network models of similar size, run search algorithms, compare results.

Compare different network simulation/layout algorithmsLoad/generate a network data set, visualize it using JUNG network layouts.

Run prefuse visualizationsLearn about different interaction algorithms (prefuse code, data is pregiven).Use the IVC to run a toolkit.

15


Prefuse Visualizations

Ballon Tree: load /sampledata/prefuse/chitest.hdirTreeMap Demo: load /sampledata/prefuse/chitest.hdirZoom -> right mouse button

Radial Graph: load /sampledata/prefuse/friendster.xml or/sampledata/prefuse/terror.xml

Force Demo (data is pregiven)


Interaction (prefuse code)

DistortionFisheye ViewZooming Pan

16


ActiveWorld Toolkit

Fromsampledata/activeworldtoolkit

Load mapper.ini(toolkit initialization file)Load Propdump culture.dump(‘Culture’ is a world in the ‘Quest Atlantis’ universe of virtual worlds)Load Registry questatlantis.reg(the Quest Antlantis universe has one registry)

Checkmark Objects, Links, Teleports to display those.


• Load users culture.log (via Users > Load User Data)

• Select users• Check Trails

17


Extend the IVC Software Framework

Read the ‘IVC Software Framework Programmer Manual 0.1’in /IVC-www/sw/papers/ivc-framework.doc

Table of Contents1 Introduction2 Target Audiences3 Major Design Decisions4 Plug-In Based Software Architecture4.1 Core4.2 Data Models4.3 Persistence4.4 Graphical User Interface4.5 Plug-Ins5. Extending the IVC5.1 Integrating New Algorithms5.2 Writing New Persisters5.3 Integrating Toolkits6. Conclusions


Steps to integrate a new algorithm1. Clean up the code and document it well. 2. Check if the IVC persistence layer satisfies your data storing and loading requirements

(if not goto 5.2). 3. Implement the Plugin interface (visualization, java code)

Two choices:(I) Build visualization window, return it to the IVC using getView method of the Plugin. (II) If your visualization uses a top-level GUI Container then start your visualization as a separate window and return null to the IVC.

4. Use java jar utility and archive your files including your plugin implementation and place it in the plugins folder. Put only java .class files in this jar. Put the other files at the same or lower level in the directory structure and make sure this is how you refer to your files within the code. Test your plugin to see that everything works.

5. Zip up or make a tar ball of all your files including non-java code. Now your algorithm is available to anyone who has the IVC. Anyone can download your zip file, unzip it in their plugins folder and the algorithm is immediately available.

18






IVC Learning Modules

19


Visualizing

Tree Data

http://iv.slis.indiana.edu/lm/lm-trees.html


Student’s Project Results

User & Task Analysis for Visualizing Tree Data

Visualizing the structure of IU’s Decision Support SystemVisualizing the co-occurrences of keywords in DLib Magazine articles. Visualization of the Java APIVisualizing the Library of Congress Classification System to retrieve legal materials in a library.

See Handin pages athttp://ella.slis.indiana.edu/~katy/handin/L579-S04/cgi/handinlogin.cgi

Image by Peter Hook and Rongke Gao

20


Validation: Teaching InfoVis using the Learning Modules

Time Series

Analysis &

Visualization

http://iv.slis.indiana.edu/lm/lm-time-series.html


Student’s Project Results

Time Series Analysis & Visualization

Using Timesearcher and the Burst Detection Algorithm to Analyze the Stock Market from 1925 to 1945Applying Burst and TimeSearcher to Chat DataLab Access TrendsQuest Atlantis Chat Log Data

See Handin pages athttp://ella.slis.indiana.edu/~katy/handin/L579-S04/cgi/handinlogin.cgi

21


Top fifteen most occurring topics from 1944 to 2004 in Timesearcher

Visualizing the Work of the United States Supreme Court Based on Time Data and Top Level West Topics

by Peter A. Hook & Rongke Gao

All topics by West Category and Sub-Category grouped corresponding to the five chief justices

All topics grouped by West Category and Sub-Category grouped over the entire lengths of the data set


Visualizing niches of the blog universe.

Visualizing Niches of the Blog Universe BY Mike Tyworth and Elijah Wright

22


Future Work

This summer, more data modeling, data analysis and visualization algorithms will be integrated into the IVC. Hope many of you will contribute to the IVC via Sourceforge http://sourceforge.net/projects/ivc.

A programmer-friendly Java API that allows researchers to pipeline data between analysis algorithms and visualization tools within and outside the IVC will be implemented.

Algorithm documentation and the learning modules set will be updated/expanded.

We plan to have demos/tutorials on the InfoVis CyberInfrastructure at the - InfoVis Conference 2004 in London, UK, July 14-16.- IEEE InfoVis Conference 2004 in Austin, Texas, Oct 10-12.- Visualization and Data Analysis Conf. 2005 in San Jose, CA, Jan 17-18.


Acknowledgements

Craig A. Stewart, Mary Papakhian, Anurag Shankar all UITS generously made the Research Database Complex available for this project and provided very insightful comments.

Stephanie Burks, Principal Unix Systems Administrator, Research and Technical Services, UITS has been instrumental in setting up the computing infrastructure and administration of the Oracle database.

Algorithm developers and integrators are acknowledged in the code documentation.

23

Documents

InfoVis CyberInfrastructure