Multiset Representation of Objects in Information Retrieval Systems

Abstract

In this paper we present multiset-trie -a novel data structure which operates on objects represented as multisets. The multiset-trie is a search-tree-based data structure with properties similar to those of a trie. In particular, we efficiently implement the standard search tree operations together with the special set containment operations, i.e. subset and superset queries in the context of multisets. These are called submultiset and supermultiset, respectively, and are used for implementation of various queries that can be performed on multisets in a multiset-trie. The corresponding running times of the developed functions are mathematically and experimentally analyzed. One of the most important queries is the search of the nearest neighbor given an input object. The nearest neighbor search of a multiset-trie makes it a good alternative for the index data structures that are used in information retrieval systems. In particular, our research is focused on the application of the multiset-trie to full-text search systems.