K-Nearest-Neighbors (KNN) search is a fundamental problem in many application domains such as database and data mining, information retrieval, machine learning, pattern recognition and plagiarism detection. Locality sensitive hash (LSH) is so far the most practical approximate KNN search algorithm for high dimensional data. Algorithms such as Multi-Probe LSH and LSH-Forest improve upon the basic LSH algorithm by varying hash bucket size dynamically at query time, so these two algorithms can answer different KNN queries adaptively. However, these two algorithms need a data access post-processing step after candidates' collection in order to get the final answer to the KNN query. In this thesis, Multi-Probe LSH with data access post-processing (Multi-Probe LSH with DAPP) algorithm and LSH-Forest with data access post-processing (LSH-Forest with DAPP) algorithm are improved by replacing the costly data access post-processing (DAPP) step with a much faster histogram-based post-processing (HBPP). Two HBPP algorithms: LSH-Forest with HBPP and Multi- Probe LSH with HBPP are presented in this thesis, both of them achieve the three goals for KNN search in large scale high dimensional data set: high search quality, high time efficiency, high space efficiency. None of the previous KNN algorithms can achieve all three goals. More specifically, it is shown that HBPP algorithms can always achieve high search quality (as good as LSH-Forest with DAPP and Multi-Probe LSH with DAPP) with much less time cost (one to several orders of magnitude speedup) and same memory usage. It is also shown that with almost same time cost and memory usage, HBPP algorithms can always achieve better search quality than LSH-Forest with random pick (LSH-Forest with RP) and Multi-Probe LSH with random pick (Multi-Probe LSH with RP). Moreover, to achieve a very high search quality, Multi-Probe with HBPP is always a better choice than LSH-Forest with HBPP, regardless of the distribution, size and dimension number of the data set.

The Resource Description Framework (RDF) is a specification that aims to support the conceptual modeling of metadata or information about resources in the form of a directed graph composed of triples of knowledge (facts). RDF also provides mechanisms to encode meta-information (such as source, trust, and certainty) about facts already existing in a knowledge base through a process called reification. In this thesis, an extension to the current RDF specification is proposed in order to enhance RDF triples with an application specific weight (cost). Unlike reification, this extension treats these additional weights as first class knowledge attributes in the RDF model, which can be leveraged by the underlying query engine. Additionally, current RDF query languages, such as SPARQL, have a limited expressive power which limits the capabilities of applications that use them. Plus, even in the presence of language extensions, current RDF stores could not provide methods and tools to process extended queries in an efficient and effective way. To overcome these limitations, a set of novel primitives for the SPARQL language is proposed to express Top-k queries using traditional query patterns as well as novel predicates inspired by those from the XPath language. Plus, an extended query processor engine is developed to support efficient ranked path search, join, and indexing. In addition, several query optimization strategies are proposed, which employ heuristics, advanced indexing tools, and two graph metrics: proximity and sub-result inter-arrival time. These strategies aim to find join orders that reduce the total query execution time while avoiding worst-case pattern combinations. Finally, extensive experimental evaluation shows that using these two metrics in query optimization has a significant impact on the performance and efficiency of Top-k queries. Further experiments also show that proximity and inter-arrival have an even greater, although sometimes undesirable, impact when combined through aggregation functions. Based on these results, a hybrid algorithm is proposed which acknowledges that proximity is more important than inter-arrival time, due to its more complete nature, and performs a fine-grained combination of both metrics by analyzing the differences between their individual scores and performing the aggregation only if these differences are negligible.