XML Structured and Unstructured Query Processing

xml has become a standard for representing, storing and exchanging data, so effectively and efficiently retrieving data from xml data sources becomes increasingly important. In this thesis, we mainly study two types of xml queries: xml structured query and xml keyword search. For xml structured query, we focus on twig pattern matching, which lies in the center of most xml query languages (e.g., XPath, XQuery). Existing twig pattern matching algorithms can be classified into two-phase algorithms and one-phase algorithms. We first propose two novel one-phase holistic twig matching algorithms, TwigMix and TwigFast, which combine the efficient selection of useful elements (introduced in TwigStack) with the simple lists for storing final solutions (introduced in TwigList). TwigMix simply introduces the element selection function getNext of TwigStack into TwigList to avoid manipulation of useless elements in the stack and lists. TwigFast further improves this by introducing some pointers in the lists to completely avoid the use of stacks. On the other hand, previous twig pattern matching algorithms may incur other redundant computation, so we propose two approaches, namely re-test checking and forward-to-end, which can reduce the redundant computation and can be easily applied to both holistic one-phase and two-phase algorithms. Improving the effectiveness of xml keyword search remains an open problem. In this thesis, we first present XKMis, which divides the nodes into meaningful and self-containing information segments, called minimal information segments (MISs), and return MIS-subtrees which consist of MISs that are logically connected by the keywords. The MIS-subtrees are closer to what the user wants. XReal [1] utilizes the statistics of underlying data to resolve keyword ambiguity problems. However, we found their proposed formula for inferring the search-for node type suffers from inconsistency and abnormality problems. Therefore, we propose a dynamic reduction factor scheme as well as a novel algorithm DynamicInfer to resolve these two problems. Then, we resolve the ambiguities of keywords by exploiting users’ typing habit in constructing keyword queries. We propose an approach which infers and ranks a set of likely search intentions. In a search intention, each keyword has a specific meaning. The result subtrees of the inferred likely search intentions are returned to users in clusters, which can significantly save users’ browsing time. Finally we explore the application of query suggestion in xml keyword search and propose a novel interactive xml query system XQSuggest, which mainly targets non-professional users who roughly know the contents of the database. In summary, this thesis presents several novel algorithms to improve the efficiency of twig pattern matching. It also presents several approaches to resolve the ambiguity of keywords and improve the effectiveness of xml keyword search. Statement of Originality This work has not previously been submitted for a degree or diploma in any university. To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where due reference is made in the thesis itself.