SSPR: A Skyline-Based Semantic Place Retrieval Method

Abstract

With the introduction of spatial semantics in the knowledge base, semantic place retrieval on spatial RDF data has become a popular research topic in recent years. Most existing methods ignore the following two problems. First, exact matching leads to a large number of potential results being missed and ultimately returning limited results. Second, the Top-k linear ranking function transforms the multi-objective problem into the single-objective optimization, causing the results to be prone to extreme values. In this paper, we propose a new approach named SSPR, which replaces exact matching with fuzzy matching to make retrieval closer to the human experience of interpretation. In addition, inspired by skyline, we computation an efficient query algorithm to select places from both the semantic relevance and spatial distance, returning mutually non-dominated results. The experiments on different test sets demonstrate that our approach compared to the traditional kSP method balances spatial distance and semantic relevance while outperforming retrieval efficiency.