A Novel Ranking Framework for Linked Data from RDB

Web is evolving from a global information space of linked document to linked data, of which a critical source is from the structured data stored in RDB. More and more enterprises publish their RDB data into linked data and provide a simple browsing end point on it. However, this kind of browsers is usually lack of data organization and ranking, resulting in limited usability. In this paper, we want to investigate the problem of ranking linked data exposed from RDB. The relationships in this kind of data are usually deduced from mapping rules based on table schemas, except for the general feature that relationships are attached with different types. The feature determines that the ranking criteria in linked data sourcing from RDB will be different from other linked data sources, and also from document Web. We propose a novel ranking framework for solving the problem. The core idea is to group relationships by their types, then firstly rank the types, secondly rank the instances attaching to each type. More detail, we propose respective ranking criteria for each step by considering the mapping rules and heterogeneous graph structure of data Web, etc.