Transcript dg.o Slides - Columbia University

Interactive Dynamic
Aggregate Queries
Kenneth A. Ross
Junyan Ding
Columbia University
Scenario Outline
User
Data Request
Dynamic Query
Graphical User
Interface
Mediator
Web
...
Unified
Results
Traditional
DBMS
Dynamic Query
Engine
Data Files
e.g., PUMS
Engine Decoupled from Interface
Can use a variety of interfaces
Multiple connections to one server
Can “do one thing well”
Client/Server parallelism
Abstract interaction via API
Engine Performance Goals
Interactive data exploration
Millions of records
Thousands of columns (but look at ten or so
at a time)
Aggregates and statistical measures
Fine adjustments at 30 answers/second.
Technical Details
Main Memory Implementation
Multidimensional tree structures
Cache consciousness
Branch Misprediction
SIMD
Asynchronous work
GlossIT
Internet
GetGloss
Xml
glossary info
.gov url
ParseGloss
Sensus
GlossIT
Automatic Ontologies
from Web Pages
Judith L. Klavans
Peter K. Davis
Samuel Popper
Columbia University
Where are Glossaries?
Internet
GetGLOSS
Web Crawling to Find Glossaries
GetGloss
ParseGLOSS
Building an Ontology
ParseGloss
Output for
SENSUS Ontology
SENSUS
Data Users
Social Science Research Data Component
Electronic Data Service – Columbia Univ
Librarians and Data Specialists
Steady stream of different user groups
Collect user logs and interview users
Coordinated by Walter Bourne
DGRC User Interface Testbed
Menu presented as
grid of alternating
rows and columns
– Top level items in left
column
Ontology entry shown
in beam for selected
item
– Located as
near as possible
DGRC User Interface Testbed
Color coding shows
parental and
semantic
relationships
DGRC User Interface Testbed
Fisheye
magnification of
region of interest
Magnified group
laid out to avoid
internal overlap
Goals of Evaluation
Optimize the effectiveness of the interface,
Identify usability problems,
Provide feedback on the overall functionality,
Anticipate changes in user need that might drive
future development,
Validate the design,
Indicate the extent to which the interface
improves on previous interfaces.
Methods of Evaluation
Interviews to Experts
Analysis of DataGate Interface
Design and Testing with Heuristic for
Database Interface
User and Task Analysis
Interview Findings
User Type Identification
– Novice and Power/Expert Users
User Goals
Kinds of Questions
Types of Searches
Related Terms for Searches
– Difficulty of Use of Alternative Terms
Selecting the database
Learning to Use the Interface
– Innovative Interface
– Need Orientation and Time to Familiarize with the Interface
Interview Findings
Searching Styles
Flexibility to Searching Styles
Helping the User Define the Search
– Help users to Visualize the Context and Structure of
Information
– Definition and Redefinition of Search
Standardization Problems
Suggestions for the Design
Census Characteristics and Interface Possibilities
• Variables
– Hierarchical Structure
– Massive Amount
• Terminology
– Definitions Change
– Obscure Terminology
– Census Question
Change
• Geographical References
– Boundaries Change
– Unique Boundaries
– Codes for Areas
– Various Meanings for
Same Names
• Content Visualization
– Display Information
Organization
• Dynamic Menu
– Magnification on Selected
Items with Full Content
• Zoom In, Zoom Out
– Manipulate the Level of
Magnification
• Searchlight
–
–
–
–
Multiple Layers of Display
Alternative Terms
Definition of Terms
Alternative Pathways
• Create Dynamic Maps