Ranking Associative Entities in Knowledge Graph by Graphical Modeling of Frequent Patterns

Abstract

Ranking associative entities in Knowledge Graph (KG) is critical for entity-oriented tasks like entity recommendation and associative inference. Existing methods benefit from explicit linkages in KG w.r.t. exactly two query entities via the closely appearing co-occurrences. Given a query including one or more entities in KG, it is necessary to obtain the implicit associative entities and uncover the strength of associations from data. To this end, we leverage KG with Web resources and propose an approach to ranking associative entities based on frequent pattern mining and graph embedding. First, we construct an entity dependency graph from the frequent patterns of entities generated from both KG and Web resources. Thus, the existence and strength of associations between entities could be depicted effectively in a holistic way. Second, we embed the dependency graph into a lower-dimensional space and consequently fulfill entity ranking on the embedding. Finally, we conduct an extensive experimental study on real-life datasets, and verify the effectiveness of our proposed approach compared to competitive baselines.

Recommendation method of teaching resources for professional music courses based on knowledge graph

“…This type of recommendation places a heavy emphasis on feature engineering and is prone to excessive specialization concerns; however, it has the potential to remove reliance on scoring data and reduce the severity of the cold start problem by matching learner and course pictures on the basis of domain knowledge [13]. Traditional techniques to course suggestion, while their ease of use, provide just a surface-level understanding of each individual learner and their preferences in terms of the course content that they would like to study [14].…”