Abstract:Similarity searches on heterogeneous information networks (HINs) have attracted wide attention from both industrial and academic areas in recent years; for example, they have been used for friend detection in social networks and collaborator recommendation in coauthor networks. The structural information on the HIN can be captured by multiple metapaths, and people usually utilize metapaths to design methods for similarity search. The rich semantics in HINs are not only structural information but also content stored in nodes. However, the content similarity of nodes was usually not valued in the existing methods. Although some researchers have recently considered both types of information in machine learning-based methods for similarity search, they have used structure and content information separately. To address this issue by balancing the influence of structure and content information flexibly in the process of searching, we propose a double channel convolutional neural network model for top-k similarity search, which uses path instances as model inputs and generates structure and content embeddings for nodes based on different metapaths. We design an attention mechanism to enhance the differences in metapaths for each node. Another attention mechanism is used to combine the content and structure information of nodes. Finally, an importance evaluation function is designed to improve the accuracy and make the model more explainable. The experimental results show that our search algorithm can effectively support top-k similarity search in HINs and achieve higher performance than existing approaches.

Top-k similarity search in heterogeneous information networks with x-star network schema.

Semantic Enhanced Top-k Similarity Search on Heterogeneous Information Networks

A Meta Graph-Based Top-k Similarity Measure for Heterogeneous Information Networks

HeteRank: A general similarity measure in heterogeneous information networks by integrating multi-type relationships.

Top-k Similarity Join in Heterogeneous Information Networks

Top-K Similarity Matching In Large Graphs With Attributes

Semantic Enhanced Top-k Similarity Search on Weighted HIN

On Perspective-Aware Top-K Similarity Search In Multi-Relational Networks

A distributed approach for top-k star queries on massive information networks

Panther: Fast Top-K Similarity Search on Large Networks

Top-K interesting subgraph discovery in information networks

Fast and Flexible Top-k Similarity Search on Large Networks

Effective Community Search over Large Star-Schema Heterogeneous Information Networks

Two birds with one stone: An efficient hierarchical framework for top-k and threshold-based string similarity search

Exploiting Transitive Similarity and Temporal Dynamics for Similarity Search in Heterogeneous Information Networks

Effective Similarity Search on Heterogeneous Networks: A Meta-path Free Approach

Scalable Top-k Query on Information Networks with Hierarchical Inheritance Relations

Top-k Socially Constrained Spatial Keyword Search in Large SIoT Networks

Empirical Research on Similarity of the Global Based on Complex Technology Trading Network

Integrating meta-path selection with user-preference for top-k relevant search in heterogeneous information networks

Approximate top-k structural similarity search over XML documents