Microsoft Tests Search Clustering
Microsoft Research is testing a search toolbar that utilizes an experimental search result clustering (SRC) technique to automatically categorize search results into groups by topic. But Microsoft isn't the only one who sees potential in the feature; America Online this week implemented clustering into its revamped AOL Search.
Clustering brings a semblance of order to search results, which are traditionally arranged by their ordinal ranking without much emphasis on making navigation more convenient.
When toolbar users enable clustering, queries are passed through a back-end clustering server, sent to a search engine to retrieve results, and subsequently re-formatted for display. Results are displayed in two columns: The first lists topics; the other details the contents of each heading.
Microsoft Research Asia's Web Search and Mining Group has engineered the SRC algorithm to efficiently scan for distinct and independent keywords so that clusters can be formed "on-the-fly."
Rival America Online has already put clustering to work in a production environment. The newly redesigned AOL Search clusters search results from Google in a side pane in order to help users find what they are looking for with minimal effort. Unlike Microsoft, AOL has outsourced the job of developing its clustering capabilities to Vivisimo.
Microsoft refused to say whether it plans to add search result clustering to its MSN Search, although the company's research arm has been developing the technology for quite some time.
Information on the SRC technique was first presented at a United Kingdom trade show in July 2004 in an essay titled Learning to Cluster Web Search Results.
The Microsoft Research SRC Toolbar is available for download via FileForum.