The Retrieval View, a Component of the Document Architecture

Abstract

The problem addressed in this paper is the efficient retrieval of hypermedia documents in large document collections. In contrast to traditional information retrieval, both the indexing and the retrieval process have to be extended to non-textual information such as graphics, images, and speech. In addition to content related properties of the documents, structural properties are used for indexing and retrieval. Such structural properties are usually described in the logical and layout view of the document architecture and consist mainly of the logical structure, format information, and intra- and inter-document links. As structural information establishes the coherence of a document, it is an important part of a document’s characterization and is, therefore, of high value when queries are evaluated. In order to facilitate access to the information necessary for the retrieval algorithms, we extended a traditional document architecture with a retrieval view. The content of this extended document architecture is created when multimedia collections are being put together, such as our test collection, consisting of patient data from an orthopedic clinic. Preliminary tests show that the quality of query results is significantly improved when information from the retrieval view is utilized by the retrieval algorithm.