Skip to main content



Distinguishing Forests from Trees in Search Engine Results

By Gabe Goldberg, HCIL Media Fellow

Categorized Search

The overview for this "leonardo da vinci" search is organized by topic, geography, government agency, and when the searcher last visited the pages. Categorized overviews help users organize and assess search results, providing a more satisfying Web experience.

We all know people who cherish tiny details but never quite grasp the big picture. They're figuratively unable to see the forest for the trees, missing the landscape's glory while obsessing over whatever grows in front of their nose.

Sometimes searching the Web feels like this. Google or any favorite search engine can cheerfully deliver a thousand -- or two million -- search result hits yet not reveal patterns, groupings, or gaps in what it quickly but mindlessly displays. Making things worse, hardly anyone looks at search results beyond the first screen or two; we either settle for one of the first few links, or add search words to prune our results. But this runs two risks.

First, we may miss a key Web site that for some reason isn't highly ranked by our search engine. Not everyone knows that search engines rank results using proprietary criteria; even worse, ranking methods often change without notice, so identical searches days or weeks apart may yield very different results. Second, there's no clue or cue about search result patterns. And the human mind can't grasp a thousand -- let alone two million -- links to see what they might collectively reveal.

Enter Bill Kules and Ben Shneiderman, respectively Graduate Research Assistant and Computer Science Professor at the University of Maryland. They're investigating how organizing the display of search results provides contextual and visual cues that make searches more powerful.

Their technology, partially supported by an AOL Fellowship in Human-Computer Interaction, is ideal when searchers are unsure of the target or goal. This is a variation on the famous Supreme Court quote: searchers may not know what they're looking for, but they recognize it when they see it.

Results, arranged in meaningful and stable categories using structures created by Kules' SERVICE program (as opposed to the ad hoc clustering used by some commercial search engines), are shown in a compact listing in the left side navigation bar. Important text (title, snippet, URL) is arranged for efficient scanning and skimming. SERVICE retains benefits of the traditional ranked results list, while adding an overview.

The list allows efficiently scanning and skimming title/snippet/URL -- which remains a critical task. The categorized overview adds another perspective on results, showing their distribution across categories.The overview also lets users explore results, narrowing them to a single category or subcategory.

Categorizing results is proving to change peoples' search style. For some searchers, the categorized overview simplified formulating queries. They issued a somewhat broad query and then browsed the appropriate category. Others used the overview to organize exploration of results, first perusing results in the Business category, then Science, Health, etc. Other users only used categories when frustrated by normal searching.

An interesting surprise is that empty categories -- which might have been expected to include results -- are in fact meaningful in some searches. For more information visit: http://www.cs.umd.edu/hcil/categorizedsearch


Tech Reports
Video Reports
Annual Symposium

News
Seminars + Events
Calendar
HCIL Seminar Series
Annual Symposium
HCIL Service Grants
Events Archives
Awards
HCIL Conference Travel Award
Job Openings
For the Press
HCIL Overview
Become a Member
Collaborators
Collaborating Groups + People
Academic Visitors
Join our Mailing List
Contact Us
Visit Us
HCIL Store
Give the HCIL a Hand
HCIL T-shirts for Sale
Our Lighter Side
HCIL Memories Page
Faculty/ Staff
Students
Ph.D. Alumni
Past Members
Research Areas
Communities
Design Process
Digital Libraries
Education
Physical Devices
Public Access
Visualization
Research Histories
Faculty Listed by Research
Project Highlights
Project Screenshots
Publications and TRs
Videos
Books
Products
Presentations
Studying HCI
Masters in HCI
PhD in HCI
Visiting Scholars
Class Websites
Sponsor our Research
Sponsor our Annual Symposium
Active Sponsorship
Industrial Visitors