GeoPhrase Tree: An Efficient Index for Frequent Phrase Query over Spatio-Temporal Ranges

The rapid growth of social media has led to massive volumes of posts tagged with locations and timestamps, offering valuable insights into local events and trends. Identifying frequently mentioned phrases within specific spatial regions and time periods is key to uncovering such localized topics and user interests. In this paper, we study the problem of querying frequent phrases within user-defined spatio-temporal ranges over large text collections. The main challenge lies in efficiently matching and aggregating phrases across vast and various documents. Existing approaches either incur high computational overhead or yield incomplete results. We propose a novel hybrid index, GeoPhrase Tree, that accelerates phrase matching by pre-constructing Frequency Suffix Trees for spatio-temporal textual objects, indexing them with R-Tree, and applying bound-based pruning strategies during online traversal. To further optimize performance in large-scale and memory-constrained scenarios, we introduce a partial storage strategy. Experiments on real-world datasets demonstrate that our method achieves superior efficiency, scalability, and semantic expressiveness compared with state-of-the-art baselines.