Concept-based search

Concept-based search identifies and suggests alternative search queries that are closely related to the user’s search query. The idea is to focus a user's search activity from the general, where lots of results are returned to the more specific with fewer, better matching results.

One way for search engines to implement concept based search is to examine how closely the results match those obtained from other searches. If there is a close match it is likely that the two search queries are related and the second query can be suggested as an alternative. Analyzing clicks can also reveal relationships. If two different queries both result in a large number of clicks on the link there queries may be considered as related.

The popularity of searches can also be used to match independent queries. Microsoft has filed a patent application (Method for finding semantically related search engine queries; 20060248068; 2nd November, 2006) based on this concept. As an example the change in popularity for searches about the “winter olympics” might match those for “curling” or “Bode Miller” (a downhill skier). Microsoft's invention analyzes the density of a given query at various points in time. That is how many searches are there for “winter olympics” compared to the overall number of searches. This removes global effects such as a rise in overall popularity of the search engine affecting results. A mathematical process called Fourier analysis can be used to make rapid comparisons between the various results.