A Study On Entity Property Grouping Methods For Browsing The Semantic Web

With the rapid growth of Semantic Web, tens of billions of RDF triples are col-lected in the Linked Data, describing various entities in the real world. Semantic Web browsers enable people to easily browse entities and exploit their values, encouraging people to contribute more high quality RDF data and promoting the development of Semantic Web.Entity property grouping is beneficial to entity browsing. Considering the de-scriptions of an entity may contain dozens or hundreds of properties, entity property grouping can enhance user experience and help users find information of interest more quickly. This thesis proposes two approaches to property grouping. Firstly, this thesis introduces a novel approach to property classification by allocating properties to ap-propriate classes. In the Linked Data, an entity usually has multiple types and these types can indicate various facets of the entity being considered. Properties in the entity descriptions are associated with these types. This thesis finds appropriate classes of properties according to property axioms and property usage and then allocates prop-erties to appropriate classes based on EBMC model. Secondly, this thesis introduces an approach to property grouping based on clustering algorithms. This thesis measures the relatedness of properties from five angles:lexical similarity of property names, se-mantic relatedness of property names, distributional relatedness of properties, overlap of property values and relatedness of property value types. Furthermore, an experiment is performed to compare these approaches, and the results show that the first approach is more beneficial in terms of entity browsing.