ROTATCY: LINK PREDICTION MODEL BASED ON ROTATION EMBEDDING IN CYLINDRICAL COORDINATE SYSTEM

Link prediction has always been an important research topic in the field of knowledge graph completion. Nowadays, there are still many link prediction models that ignore the semantic relationship between entities in the real world. In order to effectively use the semantic information between entities, a link prediction model based on rotation embedding in cylindrical coordinate system-RotatCY is proposed. The entities and relationships were embedded into the cylindrical coordinate system, and the entities were divided according to the semantic level by the combination of radial distance, height and angle, so as to achieve the modeling of the semantic level between entities. Experiments were conducted on five standard data sets. The experimental results show that the performance of the RotatCY model is improved in four evaluation indicators. Compared with RotatH, RotatCY has the largest improvement in Hits @ 1 of FB15k dataset, with an increase of 0. 143.