Aggregate Queries on Knowledge Graphs: Fast Approximation with Semantic-aware Sampling A summary of the IEEE ICDE 2022 research paper by Yuxiang Wang, Arijit Khan, Xiaoliang Xu, Jiahui Jin, Qifan Hong, and Tao Fu . Background: Knowledge graphs (KGs), such as DBpedia [1], YAGO [2], Freebase [3], and NELL [4], manage large-scale and real-world facts as big graphs in a schema-flexible manner. The same kind of information can be represented as diverse substructures [5], [6]. This schema-flexible nature should be carefully considered in the study of KG querying. Consider the factoid query [7]: e.g., “Find all cars produced in Germany” (Q117 from QALD-4 benchmark [8]). Given the KG in Figure 1(a), we expect answers as all entities having type Automobile that satisfy the semantic relation product to the specific entity Germany, e.g., Audi TT (u10), BMW 320 (u6), etc. Notice that these correct answers are linked with Germany in structurally different ways in Figure 1(a), for instance, u