Knowledge Representation Learning Based on Neighborhood Semantics

Knowledge representation is a key technology for knowledge acquisition, application and construction of knowledge graphs. Aiming at the problem of insufficient information utilization in current knowledge representation learning models, a knowledge representation learning model based on neighborhood semantics is proposed. This model focuses on generating hybrid neighborhood representations that blend entity neighborhood details. In addition, it combines entity semantic category information with mixed neighborhood as a unit to extend the entity's semantic category to neighborhood semantic categories. Comparative experiments show that our model outperforms other models in terms of metrics, such as MR, HITS@K in the link prediction, and Micro-ACC and Macro-ACC for triplet classification.