The volume and variety of information available from network information services have been increasing steadily. A lot of network information retrieval (NIR) systems have been developed. The user satisfaction with a network information service depends not only on information provided by the service, but also on the quality of an NIR system used to retrieve the information.
My purpose in this dissertation is to analyze the current state of NIR system design, to identify the most serious problems with current approaches and systems, and to propose new approaches and elements of a new NIR system architecture that will solve or alleviate selected problems in the NIR field. Particularly, I shall present design proposals providing the following benefits for NIR system users (when compared to current approaches and systems): a more extensible and universal framework for the description of the document content and behaviour, increased consistency and possibility of automation of the presentation of documents of various types, more accurate conveying of document semantics from the author to the user, improved document search results, and an easier and more reliable implementation of new NIR protocols.