Content oriented retrieval on document centric XML

XML is the perfect format for storing (mostly) textual documents in a digital library; its flexibility enables users to store both highly structured data (like database records) and free text in the same document. The data-centric parts can be searched using query languages like XPath and XQuery, where exact conditions on the structure can be imposed. For digital libraries, however, it is important to be able to search the free-text parts effectively. Standard information retrieval systems would return complete documents as retrieval results, which is useful for short documents such as web pages. If the document collection consists of books, however, this result granularity is too coarse. Users should be able to find the information that helps them solve their problem without having to wade through much information that is not relevant for their problem. To this end, content-oriented XML retrieval can help: In content-oriented XML retrieval, documents are not considered atomic entities as they are in traditional text-based information retrieval. A retrieval result can not only contain complete documents, but also parts of documents such as chapters or paragraphs. This thesis investigates several major aspects of content-oriented retrieval on XML documents: • The adaptation of standard information retrieval techniques to XML. Although the adaptation is mostly straightforward, but several peculiarities of XML have to be taken into account. • A space-efficient implementation of this base retrieval engine using customized index structures. Although the base retrieval engine uses the same concept of similarity as standard information retrieval systems, it is possible to take the XML structure into account when indexing to save space and time. • A novel method for improving retrieval quality by making use of the document structure. In particular, section titles are exploited for finding highly relevant sections in the documents. • A detailed evaluation of the retrieval quality for the base retrieval system and the proposed method. This evaluation is based on my own implementation of the retrieval system and a standard benchmark. • Preliminary ideas for searching data-centric parts of the documents. The only major part missing for a usable retrieval system is the user interface.