Exploiting Context-Dependent Quality Metadata for Linked Data Source Selection

The traditional Web is evolving into the Web of Data which consists of huge collections of structured data over poorly controlled distributed data sources. Live queries are needed to get current information out of this global data space. In live query processing, source selection deserves attention since it allows us to identify the sources which might likely contain the relevant data. The thesis proposes a source selection technique in the context of live query processing on Linked Open Data, which takes into account the context of the request and the quality of data contained in the sources to enhance the relevance (since the context enables a better interpretation of the request) and the quality of the answers (which will be obtained by processing the request on the selected sources). Specifically, the thesis proposes an extension of the QTree indexing structure that had been proposed as a data summary to support source selection based on source content, to take into account quality and contextual information. With reference to a specific case study, the thesis also contributes an approach, relying on the Luzzu framework, to assess the quality of a source with respect to for a given context (according to different quality dimensions). An experimental evaluation of the proposed techniques is also provided