GAME+: lossless knowledge-graph compression for improving the efficiency and effectiveness of rule mining

Abstract Knowledge graphs have been growing in popularity in recent years, due to their successful use in many data-analytics and knowledge-discovery tasks. As the sizes of domain graphs continue to grow, they can become too large to be processed efficiently by downstream applications. It is thus essential in many cases to address the efficiency issue by reducing knowledge-graph sizes via summarization. At the same time, to provide effective results on the graph summaries, many downstream analytics tasks require the summaries to retain as much information from the original graph as possible. It turns out that state-of-the-art data-reduction approaches may generate task-specific or lossy summaries, which may present challenges for the effectiveness of downstream analytics. Toward addressing the efficiency and effectiveness challenges, we introduce GAME+, a domain- and application-independent knowledge-graph summarization approach for generating size-reduced abstract knowledge graphs from a given graph. Through provenance maintenance, the abstract graphs generated by GAME+ can be used to reconstruct the original graphs, that is, the summarization in the proposed approach is lossless . We explore the performance of one type of knowledge-graph analytics, inference rule mining, on the abstract graphs as compared with the original graphs. Through various experiments on large-scale, real-world knowledge graphs, we provide evidence that our approach can substantially reduce knowledge-graph sizes, thereby improving rule mining efficiency on the graph data. At the same time, the abstract knowledge graphs can still enable effective rule mining due to the lossless nature of the approach. We posit that the knowledge-graph reduction provided by the proposed GAME+ approach can enable efficient, yet effective, results for diverse healthcare-related applications and use cases.