The amount and complexity of information that is electronically available results in non-efficient information retrieval. To tackle this problem we propose to use a document knowledge base that contains semantic and structural information concerning the retrievable documents that is extracted from the actual documents. We discuss how such a document knowledge base can be implemented based on description logics technology. We have previously used this technology successfully in a document management setting. We show that using such a document knowledge base results in support for advanced query functionality where structural, non-structural and semantic information can be taken into account in an integrated way. The use of the document knowledge base supports retrieval and maintenance of information as well.