Flamenco Makes Info Search Easier

Contact Omeed Elboudwarej at [email protected]

  • Printer Friendly Printer Friendly
  • Comments Comments (0)

UC Berkeley researchers have developed a search interface framework called Flamenco that could potentially revolutionize the way we search for digital information.

The research project is headed by UC Berkeley Associate Professor Marti Hearst along with students at the UC Berkeley School of Information Management and Systems and the Computer Sciences Division.

Flamenco, financed in part by the National Science Foundation, employs an interface framework that allows users to easily search through large amounts of information without feeling overwhelmed.

"Our goal was to allow people to have a ‘browsing the shelves' experience with online information while at the same time also allowing for directed search," Hearst said. "We noticed that for many collections, there is information associated with each item that is useful for browsing, but most search systems do not make that information visible to users."

Hearst's group has applied the Flamenco interface to several image collections, including a subset of the collection of the San Francisco Fine Arts Museum. They made use of categories that had been assigned by the museum curators, such as artist, media, and date, but augmented these with non-standard categories that reflect the subject matter of the images. Subject categories include objects depicted in the works of art, such as buildings or the moon, as well as properties of the images, such as shape and color.

The categories were then cross-linked so that searching for an image, like sailboats, would lead to related subcategories such as a sailboat painting done by Paul Signac in the 19th century

"We wanted to allow people to browse along several different dimensions and in any order they like, which is something that is not possible with shelves in the bookstore or library," Hearst said.

The search interface is integrated with a keyword search function designed to guide the user towards possible choices by allowing searches to be expanded and refined while "maintaining a consistent representation of the collection's structure," according to Hearst.

For example, a user who is browsing the art history collection for bridges can first view thumbnail-size images of all the images that contain bridges, then group them into a given century or decade. From there, the images can be narrowed down according to the artist, type of media used and colors that are present.

The search can then be modified while surfing the different subcategories by issuing keywords such as "sunset" to find images containing both bridges and a sunset, and then group the images in any order that the user wants.

Hearst said that current search engines are poor in their search results and in their top-level directory because the interface framework they use is not always systematically integrated with searching or within the information architecture in general.

A usability study using a prototype interface was done in which 32 art history students explored a collection of 35,000 fine arts images and compared the search experience to a standard image search interface.

The study results showed that 97 percent of the participants found that the prototype helped them learn more about the collection and 90 percent preferred it overall.

"The interface allows a lot of flexibility, but our usability studies show that people for the most part do not find it confusing," Hearst said. "Rather, they find it intuitive, interesting, and more helpful than an image search interface modeled after those we see in web search engines."

The finished product should be available for use in December 2004 when the project wraps up after five years of research. Hearst hopes that it will eventually be used for library catalogs such as the California Digital Library.

"Our usability studies suggest to me that our framework is among the best available for image browsing and search," Hearst said.


Comments (0) »

Comment Policy
The Daily Cal encourages readers to voice their opinions respectfully in regards to both the readers and writers of The Daily Californian. Comments are not pre-moderated, but may be removed if deemed to be in violation of this policy. Comments should remain on topic, concerning the article or blog post to which they are connected. Brevity is encouraged. Posting under a pseudonym is discouraged, but permitted. Click here to read the full comment policy.
White space
Left Arrow
Image Study: Video Games May Improve Vision
Video games may not be as bad as your mother told you, according to two new...Read More»
Image Sea Urchin Skeletons Help Researchers Bone Up on B...
By examining the processes in a sea urchin's early ske...Read More»
Image Theory Links Supernovae To Smaller Set of Outburst...
A new discovery about the biggest and brightest star in the gala...Read More»
Right Arrow

Job Postings

White Space