Efficient semantic search over structured web data: A GPU approach

Abstract

Semantic search is an advanced topic in information retrieval which has attracted increasing attention in recent years. The growing availability of structured semantic data offers opportunities for semantic search engines, which can support more expressive queries able to address complex information needs. However, due to the fact that many new concepts (mined from the Web or learned through crowd-sourcing) are continuously integrated into knowledge bases, those search engines face the challenging performance issue of scalability. In this paper, we present a parallel method, termed gSparql, which utilizes the massive computation power of general-purpose GPUs to accelerate the performance of query processing and inference. Our method is based on the backward-chaining approach which makes inferences at query time. Experimental results show that gSparql outperforms the state-of-the-art algorithm and efficiently answers structured queries on large datasets.