Knowledge Pyramid: A Novel Hierarchical Reasoning Structure for Generalized Knowledge Augmentation and Inference
This work addresses a domain-specific problem in knowledge graph reasoning for applications like information retrieval and recommendation, but it appears incremental as it builds on existing knowledge augmentation strategies.
The paper tackled the problem of insufficient generalization and poor robustness in knowledge graph reasoning by proposing a knowledge pyramid framework that extracts high-level pyramidal knowledge from low-level knowledge, resulting in improved inference performance and generalization, especially with fewer training samples.
Knowledge graph (KG) based reasoning has been regarded as an effective means for the analysis of semantic networks and is of great usefulness in areas of information retrieval, recommendation, decision-making, and man-machine interaction. It is widely used in recommendation, decision-making, question-answering, search, and other fields. However, previous studies mainly used low-level knowledge in the KG for reasoning, which may result in insufficient generalization and poor robustness of reasoning. To this end, this paper proposes a new inference approach using a novel knowledge augmentation strategy to improve the generalization capability of KG. This framework extracts high-level pyramidal knowledge from low-level knowledge and applies it to reasoning in a multi-level hierarchical KG, called knowledge pyramid in this paper. We tested some medical data sets using the proposed approach, and the experimental results show that the proposed knowledge pyramid has improved the knowledge inference performance with better generalization. Especially, when there are fewer training samples, the inference accuracy can be significantly improved.