Modelling and Factorizing Large-Scale Knowledge Graph (DBPedia) for Fine-Grained Entity Type Inference

Abstract

Recent years have witnessed a rapid growth of knowledge graphs (KGs) such as Freebase, DBpedia, or YAGO. These KGs store billions of facts about real-world entities (e.g. people, places, and things) in the form of triples. KGs are playing an increasingly important role in enhancing the intelligence of Web and enterprise search and in supporting information integration and retrieval. Linked Open Data (LOD) cloud interlinks KGs and other data sources using the W3C Resource Description Framework (RDF) and makes accessible on web querying. DBpedia, a large-scale KG extracted from Wikipedia has become one of the central interlinking hubs in the LOD cloud. Despite these impressive advances, there are still major limitations regarding coverage with missing information, such as type, properties, and relations. Defining fine-grained types of entities in KG allows Web search queries with a well-defined result sets. Our aim is to automatically identify entities to be semantically interpretable by having fine-grained types in DBpedia. This paper embeddings entire DBpedia, and applies a new approach based on a tensor model for fine-grained entity type inference. We demonstrate the benefits of our task in the context of fine-grained entity type inference applying on DBpedia, and by producing a large number of resources in different fine-grained entity types for connecting them to DBpedia type classes.