Effective searching of RDF knowledge bases

RDF data has become a vital source of information for many applications. In this thesis, we present a set of models and algorithms to effectively search large RDF knowledge bases. These knowledge bases contain a large set of subjectpredicate-object (SPO) triples where subjects and objects are entities and predicates express relationships between them. Searching such knowledge bases can be done using the W3C-endorsed SPARQL language or by similarly designed triple-pattern search. However, the exact-match semantics of triple-pattern search might fall short of satisfying the users needs by returning too many or too few results. Thus, IR-style searching and ranking techniques are crucial. This thesis develops models and algorithms to enhance triple-pattern search. We propose a keyword extension to triple-pattern search that allows users to augment triple-pattern queries with keyword conditions. To improve the recall of triple-pattern search, we present a framework to automatically reformulate triple-pattern queries in such a way that the intention of the original user query is preserved while returning a sufficient number of ranked results. For efficient query processing, we present a set of top-k query processing algorithms and for ease of use, we develop methods for plain keyword search over RDF knowledge bases. Finally, we propose a set of techniques to diversify query results and we present several methods to allow users to interactively explore RDF knowledge bases to find additional contextual information about their query results.