Cascade embedding model for knowledge graph inference and retrieval

Highlights：

• Translation-based models are popular tools for researching knowledge graphs, and have the potential to improve the performance of inference and retrieval tasks.

• Current translation-based models are limited in their ability to handle reciprocal and unbalanced relations.

• This paper introduces graph-based models which help to address these limitations. The authors found that the distance between the graph embedding of two entities is correlated to the entities’ relation. The correlation can help to refine the performance of translation-based models.

• A cascade model was designed to incorporate graph embedding and knowledge embedding into a unified framework. The objective of the framework is to optimize the learning process and to fuse the knowledge and graph embedding features from both local and global stages, leading to a more accurate embedding-based representation of relations.

• Different cascade structures were used to find the optimal solution to the problem of knowledge inference and retrieval. Experimental results show that the stepwise-cascade model significantly outperforms other baselines.

摘要

•Translation-based models are popular tools for researching knowledge graphs, and have the potential to improve the performance of inference and retrieval tasks.•Current translation-based models are limited in their ability to handle reciprocal and unbalanced relations.•This paper introduces graph-based models which help to address these limitations. The authors found that the distance between the graph embedding of two entities is correlated to the entities’ relation. The correlation can help to refine the performance of translation-based models.•A cascade model was designed to incorporate graph embedding and knowledge embedding into a unified framework. The objective of the framework is to optimize the learning process and to fuse the knowledge and graph embedding features from both local and global stages, leading to a more accurate embedding-based representation of relations.•Different cascade structures were used to find the optimal solution to the problem of knowledge inference and retrieval. Experimental results show that the stepwise-cascade model significantly outperforms other baselines.