Knowledge graphs (KG) have evolved in the World Wide Web as a standard for representing facts. KG are based on entities, representing persons, places, an abstract things, such as events. KG also contain facts describing and connecting entities using predicates. Ontologies provide entailment mechanisms such as the ability to group entities into a class, create sameAs links between entities, equivalence relationships between classes, and denote predicates as subProperties of a general property.
These mechanisms have allowed creating one of the most substantial achievements of modern information management, namely, the Linked Open Data (LOD) cloud, network of KG from a diverse set of domains containing over 193 billion facts.
Some KG are built by method of semi-automated extraction performed on unstructured texts. The process of extracting entities and predicates from unstructured text to extend existing knowledge graphs is the focus of our work.
The need to resolve extracted entities against those which exist in the KG leaves many entities and predicates behind. Knowledge Vault (KV) is widely considered state-of-the-art in this realm, but is trained under a Local closed world assumption, where training and testing statements are ignored if the statement does not appear in Freebase (its evolving knowledge-graph). Thus, KV is trained in the absence of true new information.
Fader, Anthony, Stephen Soderland, and Oren Etzioni. "Identifying relations for open information extraction." Proceedings of the conference on empirical methods in natural language processing. Association for Computational Linguistics, 2011.
Niu, Feng, et al. "Elementary: Large-scale knowledge-base construction via machine learning and statistical inference." International Journal on Semantic Web and Information Systems (IJSWIS) 8.3 (2012): 42-73.
Sagi, Tomer, and Avigdor Gal. "Schema matching prediction with applications to data source discovery and dynamic ensembling." The VLDB Journal—The International Journal on Very Large Data Bases 22.5 (2013): 689-710.
Nickel, Maximilian, et al. "A review of relational machine learning for knowledge graphs." Proceedings of the IEEE 104.1 (2016): 11-33.