A new method for implementing fuzzy retrieval from a spatial database

Abstract

An algorithm has been defined to implement a new type of fuzzy query (i.e., iterative, ranked retrieval) from an exclusive, partitional, spatial database. First, a working definition of the type of fuzzy query is developed. This type of query does not depend on the traditional definition of a fuzzy set membership function. This membership function allows only a predetermined number ( N) of the highest ranking items to belong to the set. Membership in the set is determined not just by an item's facets, but also by the distribution of other items which are retrieved as a result of the requirements established by the query. Next, a solution technique for implementing this type of fuzzy query is developed. The algorithm works by first converting a fuzzy query to a range query, and then retrieving and ranking all items which fall within the range query's boundaries. The algorithm proceeds by iteratively lowering the desired target rank until the N “best” items are found. When the target rank is lowered, range query limits are implicitly calculated from the lowered target rank. A prototype system implementing this technique for retrieval from a PLOP hashing database is described. A simple example is presented. Finally, conclusions are drawn and future research directions discussed.