Farspredict: A benchmark dataset for link prediction

Knowledge graphs have received much attention in recent years due to their applications that offer significant economic benefits. A Knowledge graph contains the knowledge obtained from the sources, including texts and tables. It has many applications in natural language processing and has been investigated as a potential reasoning source for explainable artificial intelligence. Although the impact of creating knowledge graphs in non-English languages has been explored recently, little attention has been paid to preparing a suitable knowledge graph for use in the link prediction field. At the same time, one of the main reasons that significant progress has yet to be made in Persian reasoning, recommendation systems, and other similar fields is the need for a proper knowledge graph in these languages. Although some attempts have been made to construct a Persian knowledge graph, the most successful is the Farsbase project. By applying Farsbase for link prediction through KGE models, we realized it is too weak to be used for link prediction. In approach to state-of-the-art link prediction methods, we come to the KGE methods. These methods were introduced with TransE, which falls into translational distance models.