Recommendations and object discovery in graph databases using path semantic analysis

Abstract

This paper presents a novel approach to recommendation systems based on graph databases (e.g. LinkedData). Graph databases contain large amounts of heterogeneous and interlinked data from many sources, hence different algorithms to analyze these data are necessary. Moreover, it must be said that these properties of collected data make it impossible to store them in a relational data model. As for now, there are few methods of intelligent data exploration from graph databases. In this paper, we propose new methods of discovering, searching and recommending objects in a graph database. The proposed methods can be applied to many domains or semantics, e.g. books, movies, etc. Paths in the graph (sequences of graph edges) have weights assigned, on the base of a training set according to their semantic importance. Recommended objects (vertices of the graph) may be found by analysing existing paths and where they lead to. Additionally, the method is able to determine if a found object belongs to a given problem from the point of view of a chosen semantics, hence only objects intuitively linked one to each other, are recommended. The proposed method is verified in two case studies, using an open graph database. © 2014 Springer International Publishing.