Overview
The visual search engine
is a system that allows users to search for media items (animation, image,
text) is a visual, associated way. The idea is to mimic a brainstorming
process or any other creative process, where the end result is different
from what you thought you are looking for at the beginning.
The Visual Search Engine
was developed at the Zapa Digital Arts development center in Tel-Aviv at
1996. It implemented in the Print Shop LiveMail product, and was released
by Broderbund Inc. at 1997.
Design
Issues
The design guidelines we
defined for ourselves were:
-
An easy-to-use, WYSIWYG Interface.
-
Non-standard GUI components.
-
Support for the following media
types: Animation, Image, Background and Border.
The
Concept
Direct keyword search that
leads to an associated search.
 |
Direct Keyword Search
The VSE works in two passes.
First - a direct keyword search is done, to identify an item based on the
keyword the user enters. In this case, the user looks for the keyword -
'fly' (note - an ‘auto-complete’ feature completes the user typed words,
assuring that the keyword exist in the database). The results are presented.
The item in the center is the selected item. The user can rollover with
the mouse above the item to see the animation running. Then the associated
pass is done. |
| Associated
Search
The algorithm scans all the
database for items that are ‘associated’ to the center item. ‘Associated’
means that they have more then 1 keyword in common with the selected item's
keywords. The items that has more keywords in common with the selected
item are presented near the center item, items with less common keywords
are presented far from the center. Those items does not necessarily have
the keyword 'fly' in their keyword list.
The user can rollover with
the mouse above any item to see its animation running. One click
on any of the items, and the screen re-arranges with the selected item
in the middle, and new items around it based on the association algorithm.
This way the user sees a large variety of items every time, and can look
for items in an associative way rather then by searching directly. |
Examples
1. An animation item search
with the keyword: Walk
2. An image item search with
the keyword: Red
|