Home Research publications academics about members partnerships contact

Dynamic Queries and Query Previews
for networked information systems:
the case of NASA EOSDIS

Jump to:

Related HCIL pages:      Summary of Dynamic Queries : Predecessor of Query Previews
Related HCIL pages:      Census Project : For other examples of Query Previews
Related HCIL pages:      NSF Project : For more general examples of Query Previews

Query Previews allows users to rapidly gain an understanding of the content and scope of a digital collection. It uses overviews and previews of abstracted metadata that allows users to perform rapid and dynamic elimination of undesired data. We developed prototype query previews for a variety of NASA EOSDIS situations. Later, operational prototypes were developed at the Global Change Master Directory and more recently at Raytheon for Terra data.

Research on Query Previews also continues in the context of our projects with the Census Bureau and the NSF.

What is EOSDIS?
The NASA EOSDIS (Earth Observing System Data and Information System) project is attempting to provide online access to a rapidly growing archive of scientific earth data about the earth's land, water, and air. Data is collected from satellites, ships, aircraft,and ground crews, and stored in designated archive centers. Scientists, teachers, and the general public will then access this data over the internet. Our efforts at HCIL center on developing user interfaces to facilitate the browsing and retrieval of data from the very large archive. Our research integrates two methods: Dynamic Queries and Query Previews.

Problems with the traditional query systems?
The traditional approach to querying is to use a form fill-in interface, but such an approach leads to user frustration when the query returns either zero hits or a very large number of hits. Often, users cannot even estimate the total number of hits their query would have returned as the system only returns the first 25 - 50 hits. It is difficult to estimate how much data is available on a given topic and how to increase or reduce result set sizes.

Our approach
Our approach to overcome these challenges is to present overviews and previews of abstracted metadata that allows users to perform rapid and dynamic elimination of undesired data. The reduced volume of the abstracted metadata allows queries to be previewed and refined locally by the user, before they are submitted over the network. We developed the concept of query previews that allows users to:

Dynamic Queries involve the interactive control by a user of visual query parameters that generate a rapid, animated, and visual displays of database search results. As users adjust sliders or buttons, results are updated (within 100 msec.) on the display. The enthusiasm users have for query previews emanates from the sense of control they gain over the query. Empirical results have shown that dynamic queries are effective for novice and expert users to find trends and spot exceptions.

Dynamic query user interfaces apply the principles of direct manipulation to query formulation and imply:

Query Previews take advantage of the fact that in most cases, users are only interested in a subset of the entire database. Users are presented with generalized previews of the entire database using only the most salient attributes, they select rough ranges, and they are immediately given the availability of the data for their proposed query.

Once the scope has been sufficiently narrowed, a second query phase, called Query Refinement, can start. This phase is a dynamic query interface with all attributes and all records represented individually, but only for those selected in the Query Preview.

Looking at Census Data Using Query Previews:

Current data exploration systems that use command languages or form fillin interfaces fail to give users an indication of the distribution of data items. This leads many users to waste time posing queries that have zero-hit or mega-hit result sets. Query previews form a novel visual approach for browsing large information warehouses. Query previews supply data distribution information about the database that is being searched and give continuous feedback about the size of the result set for the query as it is being formed. Empirical results show that query previews improve user performance.

        US Census Bureau maintains large amounts of tabular data. Browsing these tables poses similar problems that we aim to solve with the query preview approach. We believe query previews can improve access to Census data in terms of user performance and satisfaction. Below, we present a demo (1) built by using this approach and 1990 US Census Income Data. This demo shows the distribution of householders over three attributes, age, income level, and ethnic origin of the householder.  Each attribute is represented by a bar chart. Selecting bars from a chart immediately updates the distribution of data over the other bar charts. Hence, users can dynamically investigate a large space of charts by simply selecting/deselecting bars. A separate bar represents the overall count of householders that satisfy the current selection. In this example, we first select the householders that are older than a certain age (2). Then we select the high-income householders (3). Finally, we select a specific ethnic origin (4) and fetch only the selected portion of the data (5). Querying can continue over states and counties of this extracted portion. If requested, users can change their previous selections and start investigating a different portion of the data.
1 2 3 4 5

    (click on icon to view full screen)

Related HCIL Pages:

QUIS: Questionnaire for User Interface Satisfaction.
Dynamic Queries and Query Previews for Networked Information Systems: the case of NASA EOSDIS
Generalized Query Previews

The challenge of multi-valued attribute data
Our original simple technique works very well for data having a single value for each attribute (e.g. the case where each record only has one topic, and covers only one time period, and one area of the globe). However, EOSDIS data often has multi-valued attributes (e.g. datasets can have multiple topics, or cover multiple areas) that led to our current effort to develop special techniques to control the size of the preview tales that countain the counts of records.

Participated in this project:

Catherine Plaisant, Assistant Research Scientist (Project lead)
Ben Shneiderman, Professor Computer Science
Egemen Tanin, Graduate Research Assistant
Maya Ventkatraman, Faculty Research Assistant
Stephan Greene, Faculty Research Assistant
Richard Beigel, Associate Visiting Professor
Tom Bruns, Faculty Research Assistant
Khoa Doan, Post-Doctoral Researcher
Kawin Ngamkajornwiwat, Graduate Research Assistant
Laurent Cailleteau, visiting undergraduate student

and collaborators at Raytheon (Bob Harberts and Wenlan Feng) and at the Global Change Master Directory - GCMD (Lola Olsen, Gene Major, Steve Johns, Chris Gokey)

Related Workshop

Publications: ( * marks the suggested first readings)

Prototyping history: (Please use recently upgraded browsers only)


see NSF project

Query Preview work continues in the context of our NSF project.

see Census project

Query Preview work also continues in the context of our Census project.
1999 Latest NASA demos not here

The latest prototypes were developed in collaboration with Raytheon ( Bob Harberts). Please contact him directly for an update.

Spring 99Multivalued attribute
data techniques
Those techniques scale to any number of records, and therefore could be used for EOSDIS inventory searches
FALL 98 SAR technique:
Single Attribute, Range only
This technique has a more restrictive interface but preserves the counts. Warning: all three attribute applets work, but the preview data is not updated when users change attribute yet (need connection to real data server)

. Binary Previews:
older version
Alternative interface
Query Preview without counts, only tells if there is data or not(Work in progress)
GCMD hybrid technique

This technique deals with multivalued data but could not scale to more than about 10,000 record, therefore only useful for EOSDIS directory searches
January 1998: GCMD demo (at GCMD)  Beta version running at NASA Global Change Master Directory (GCMD) 
September 1997: Earlier version Older version of the GCMD prototype at HCIL
Summer 1997: Initial Java Prototype with GCMD data  First operational prototype implemented by Stephen Greene and Egemen Tanin with NASA's Global Change Master Directory (GCMD). Results are returned from the GCMD database. A second level of preview is used instead of our refinement design. 
Summer 1997: Alternative Prototype with science-classified GCMD data  This prototype is similar to the above version, though here the topic attributes of the data are classified by science (abandoned for now)
Earlier prototypes

Earlier experiments and designs of Query Previews and algorithm test for dynamic queries
Fall 1996: DQ with 100,000 points New algorithms now allow Dynamic Queries updates in less than 100 msec. with 100,000 points and 10 sliders. 
Fall 1996: Pyramid Pyramid is an experimental year range selector written in Java. This small demo is not placed within its proper context and was never completed. Other version:: Pyramid-2
Spring 1996: Java Prototype A first prototype was developed in Java. The data used is hypothetical. No connection to real database. The refinement phase is only a rough mockup. The results of the preview are not really passed to the refinement. PC browsers gave the best results.
Fall 1995: TCL/TK Prototype (Screen Prints) Khoa went on to develop a prototype in Tcl/tk. Tom Bruns later refined the prototype, and created this series of screenshots. A video is also available that shows this prototype in action.
Fall 95 and Summer 98  TCL/TK Prototype (Working Version)
Download Plugin
This is a version of the initial prototype that has been revised so that it can run on a browser. 
Summer 1995: Visual Basic Prototype Khoa Doan developed the first prototype illustrating the Query Preview/Quer Refinement approach, based on a design by Catherine Plaisant and Ben Shneiderman. 

Sponsors: This work is supported in part by NASA ESDIS (NAG 52895 and NAGW 2777). The recent work was done in collaboration with the Global Change Master Directory and with the V0-IMS team at Raytheon.

Related Links: