December 30, 2015

atlasify_ice_hockey_med.png
Recently, Department of Computer Science and Engineering Assistant Professor Brent Hecht and a team of researchers announced the beta release of Atlasify, an “exploratory search” engine project that is a collaboration GroupLens at the University of Minnesota and the WebSAIL group at Northwestern University.

In Hecht’s announcement on the GroupLens blog, he describes Atlasify as “a system that lets you make a map of almost anything.”

And when Hecht says “anything,” he means it.  The examples he demonstrated during a recent interview with CS&E ranged from “badminton” and “country music” to “World War II” and “Star Trek.”  Additionally, he showed many more sample queries that he and his team have set up on the Atlasify homepage so users can see the search engine in action.

Unlike many popular search engines that generate a list of hyperlinks, each Atlasify query generates its own, unique interactive heat map for users to explore.

“Atlasify allows you to create maps there are no atlases for,” said Hecht.  “For example, there isn’t an atlas for ‘ice hockey,’ but we can make one—in fact, we made one.”

Atlasify is not intended to compete with Google and Bing. Rather, Hecht says that Atlasify is a search engine targeted at the large minority of information needs that are exploratory in nature, needs for which existing search engines are not optimized.

Exploratory information needs occur when people are seeking to broadly learn about a subject, not look for a single answer to a question or locate a specific webpage. Because of the limitations of current search engines with regard to exploratory queries, building technologies to support them are an active area of research.

“Curiosity is a very important use case,” Hecht said.

One key innovation Hecht and his team discovered while building the framework underlying Atlasify was that everything they learned using geography to explore new concepts also applied to a number of other non-geographic systems as well.  This has allowed the team to create “maps” as they apply to a historical timeline, the Periodic Table of Elements, and the US Senate seating chart.

“There’s literally no difference behind the scenes between what’s going on for the ‘map’ on the Periodic Table of Elements and the geographic map,” said Hecht.

For more information about Atlasify, be sure to read Hecht’s post on GroupLens or the team’s paper published at ACM SIGIR, one of the top publication venues in information retrieval.