Intelligent retrieval of hypermedia documents requires sophisticated document representations and querying facilities that allow for content-based and fact-based querying as well as considering the structure of documents. This paper describes POOL, a Probabilistic Object-Oriented four-valued Logic, which allows a uniform view on hypermedia documents for the purpose of their retrieval: documents, images, authors, dates, etc. are treated as objects and POOL models the content of objects, the facts about objects, and the structure of objects to provide for a relevance-based ranking of hypermedia documents.