Determining parameters for efficient retrieval in index structures for hybrid data spaces
Carsten Kropf · 2014
Different kinds of access methods supporting boolean retrieval in hybrid data spaces exist. We inspect a class of these index structures using a categorization of low and high frequently occurring keywords. This access method uses a basic R*-Tree augmented with bitlists for the representation of a set of terms. Two limits are given for these access methods in realistic environments: the length of the bitlist B Length and the limit separating the set of low and high frequently occurring terms H Limit. This paper presents a theoretical analysis of the setup of H Limit as well as an empirical analysis of the bitlist length for two different corpora in a typical database environment. The final target of this paper is the determination of the free parameters to provide efficient retrieval of data in realistic application domains.