SUBSUMPTION DEGREES BETWEEN ENTITY TYPES AND NAMES FOR APPROXIMATE KNOWLEDGE RETRIEVAL

Abstract

The Web has become a huge and indispensable source of information to be used and shared globally, where knowledge is commonly represented and stored in RDF, or alternatively, in conceptual graphs. Managing and searching for web information have gone beyond the relational database model, as the data are semi-structured and inexact answers are often the case. Usually, approximate searching results are due to mismatching between entity types and names in a query and an answer. Firstly, this research work focuses on partial subsumption of a query graph to an answer graph, which is an unsymmetric measure in contrast to similarity. Secondly, it proposes a population-based method for defining subsumption degrees between entity types, one to another, and a class-sensitive soft TF-IDF method for entity names. Lastly, on the one hand, for a user-friendly interface and easily readable query expressions, conceptual graphs are employed at the front-end. On the other hand, in order to take the advantage of the existing platform of SeRQL, an exact RDF query language, the query modification tactic is used to retrieve the knowledge graphs that are close to a query graph, before the subsumption degrees of the query graph to those answer graphs are calculated.