Beyond the Inverted Index

TL;DR

This paper introduces the group-list, a new data structure that improves the efficiency of set operations in document retrieval systems compared to traditional inverted indexes.

Contribution

The paper proposes the group-list, a novel data structure that enhances intersection and union operations in document retrieval, outperforming inverted indexes in experiments.

Findings

Group-list outperforms inverted index in synthetic dataset tests.

Group-list simplifies implementation similar to inverted index.

Experimental results demonstrate efficiency gains in set operations.

Abstract

In this paper, a new data structure named group-list is proposed. The group-list is as simple as the inverted index. However, the group-list divides document identifiers in an inverted index into groups, which makes it more efficient when it is used to perform the intersection or union operation on document identifiers. The experimental results on a synthetic dataset show that the group-list outperforms the inverted index.