Answering Label-Constrained Reachability Queries via Reduction Techniques

Abstract

Many real-world graphs contain edge labels for representing different types of relations between nodes, such as social networks, biological networks and knowledge graphs. As a fundamental task, the label-constrained reachability (LCR) query asks whether a given vertex s can reach another vertex t, only using a restricted set of given edge labels. However, existing works build a heavy index while taking too much time for answering queries online, exhibiting a poor performance on large graphs. To further reduce the index size and accelerate both index construction and query processing, we introduce two novel pruning techniques including degree-one reduction and unreachable query filter. Extensive experiments demonstrate that our proposed techniques significantly boost state-of-the-art methods.