Wednesday, September 24, 2008
2008 R&D 100 Winner
Today’s search engines operate under the familiar keyword model. But sifting through mountains of data for the most relevant results differs from the human brain. We retrieve information by associating it with related information in memories. In an effort to mimic the learning, association and recall powers of people, Lawrence Berkeley National Laboratory, Berkeley, Calif., and SeeqPod, Inc., Emeryville, Calif., have built the Biomimetic Search Engine. Working from an unsupervised algorithm, this software solves search queries by linking relevant information automatically from each analyzed source of information. “Sky” and “cloud” produce meteorology, while “sky” and “dive” produce free fall or parachutes. The engine builds these contextual frameworks on the fly, generating a vector-based module that continues to grow as more is learned about the relationship between information. The Biomimetic Search Engine is now used in Lawrence Berkeley’s GenoPharm genomics search tool to uncover gene relationships, and it anchors SeeqPod’s music search and recommendation website, which indexes uploaded music from around the Internet. It has also been used successfully within Wikipedia.
Technology
Search engine software
Developers
Lawrence Berkeley National Laboratory
SeeqPod, Inc.
