Tune relevance factors (FAST Search Server 2010 for SharePoint)
Published: May 12, 2010
FAST Search Server 2010 for SharePoint lets you tune relevance in several ways. Predefined relevance (delivered out of the box) was tested and tuned against large sets of different data. However, relevance is not absolute, and because perceived relevance differs greatly among users, organizations, and locations, it is important to tune relevancy according to your specific needs.
The relevance score is quantified in a rank score for each searchable item. This section introduces different tuning possibilities, discusses tuning details, and provides usage examples. Since some tuning is performed through programming and Windows PowerShell cmdlets, this section also describes some technical aspects related to relevance.
The following are involved when tuning relevance:
Static (quality) rank Rank points added to a search item before it is indexed. The number of points given is based on boost values that are stored in four predefined static/quality rank components. In addition, any custom integer managed property can also be used to tune static rank. You can decide which of these components to use at search time so that you only get the rank points from the components that you specify. You can control the effect of the static rank points by modifying the weight of the different components. Static rank points assigned to an indexed item are independent of the search words used, so static rank only indicates something about the general importance of an item.
Dynamic rank: Rank points added to search items at search time. The number of points given is based on a combination of the search words used and boost values that are stored in dynamic rank components. You can control the effect of the dynamic rank points by modifying the weight of the different components.
Keyword rank: Rank points added to an item when a search word matches a specified keyword. Recall is improved by searching for the associated synonyms. Also, the search experience is improved by displaying additional relevant information as Best Bets and Visual Best Bets and incorporating document- and site promotions to move relevant search results higher up in the search results list.
Linguistic features: like stemming, spell checking, stop words and tokenization effect both relevance and recall.
Fast Query Language (FQL): a query language intended for creating queries programmatically. FQL can be used only for solutions developed for FAST Search Server 2010 for SharePoint.
In this section: