Peer similarity search is a deceptively complex problem in information network analysis. Past research has primarily focused on similarity search in homogeneous networks, but real world data is often best represented using heterogeneous information networks, with multiple node and relation types carrying real-world semantics. Recent work addresses similarity search in heterogeneous networks by introducing the concept of meta paths, or paths that connect object types via a sequence of relations. These meta path-based similarity measures can capture the subtlety of peer similarity for paths containing symmetric edges, but real data contains asymmetric relations that play a significant role in peer similarity semantics, for instance citations in bibliographic networks. In this paper, we revisit the problem of peer similarity search among objects of the same type in heterogeneous information networks. We present an efficient meta path-based peer similarity measure, AsymSim, which both captures the semantics of peer similarity and remains sensitive to asymmetric relations in the network, allowing us to extract deeper peer semantics. We discuss how to efficiently handle AsymSim queries online and perform experiments on real DBLP data to verify the effectiveness of our proposed measure.
【 预 览 】
附件列表
Files
Size
Format
View
AsymSim: meta path-based similarity with asymmetric relations