This dissertation addresses several problems in the context of publishing and consuming Linked Data. It describes these problems from the perspectives of three stakeholders: the Linked Data provider, developer and scientist. The Linked Data provider is faced with impractical data re-use and costly Linked Data hosting solutions. Developers face difficulties in finding, navigating and using Linked Datasets. Scientists lack the resources and methods to evaluate their work on Linked Data at large.
This dissertation presents a number of novel approaches that address these issues, such as:
- The LOD Laundromat: a centralized service that re-publishes cleaned, queryable and structurally annotated Linked Datasets. In 2015 the Laundromat was awarded first prize in the Dutch national Linked Open Data competition, and third prize in the European equivalent.
- SampLD: A relevance-based sampling algorithm that enables publishers to decrease Linked Data hosting costs;
- YASGUI: A feature-rich query editor for accessing SPARQL endpoints;
- LOD Lab: An evaluation paradigm that enables scientists to increase the breadth and scale of their Linked Data evaluations.
This work provides a unique overview of problems related to publishing and consuming Linked Data. The novel approaches presented here improve the state-of-the-art for Linked Data publishers, developers and scientists, and are a step towards a web of Linked Data that is more accessible and technically scalable.