Efficient Spatial Keyword Search in Trajectory Databases

TL;DR

This paper introduces a novel hybrid indexing method for efficient top-$k$ spatial keyword queries on trajectory databases, combining text relevance and location proximity to improve query performance.

Contribution

It proposes a new cell-keyword conscious B$^+$-tree index for processing complex trajectory queries involving both spatial and textual data.

Findings

The proposed method achieves high performance in empirical tests.

The hybrid index effectively balances text relevance and spatial proximity.

Scalability is demonstrated through extensive experiments.

Abstract

An increasing amount of trajectory data is being annotated with text descriptions to better capture the semantics associated with locations. The fusion of spatial locations and text descriptions in trajectories engenders a new type of top-$k$ queries that take into account both aspects. Each trajectory in consideration consists of a sequence of geo-spatial locations associated with text descriptions. Given a user location $\lambda$ and a keyword set $\psi$, a top-$k$ query returns $k$ trajectories whose text descriptions cover the keywords $\psi$ and that have the shortest match distance. To the best of our knowledge, previous research on querying trajectory databases has focused on trajectory data without any text description, and no existing work has studied such kind of top-$k$ queries on trajectories. This paper proposes one novel method for efficiently computing top-$k$ trajectories. The method is developed based on a new hybrid index, cell-keyword conscious B$^+$-tree, denoted by \cellbtree, which enables us to exploit both text relevance and location proximity to facilitate efficient and effective query processing. The results of our extensive empirical studies with an implementation of the proposed algorithms on BerkeleyDB demonstrate that our proposed methods are capable of achieving excellent performance and good scalability.