YAGO2s: Modular High-Quality Information Extraction

Abstract

Abstract: In this paper, we present YAGO2s, the new edition of the YAGO ontol-ogy [SKW07, HSBW12]. The software architecture has been refactored from scratch, yielding a design that modularizes both code and data. This modularization enables us to add in new data sources more easily, while still maintaining the high accuracy and coherence of the ontology. Thus, we believe that YAGO2s occupies a sweetspot between a centralized design and a completely distributed design. In this demo, we present an application of this design to the task of planning a flight. Our proposed system finds flights between all airports close to the departure city to all airports close to the destination city. 1 Knowledge Base Construction In recent years, many projects have successfully created large-scale knowledge bases (KBs) in an automated fashion. The KBs contain millions of entities (such as rivers, universities, people, and movies), and millions of facts about them (such as who acted in which movie, which river is located in which country, etc.). There are several strategies to build such KBs. One strategy is to accumulate and reconcile as much data as possi-