In this paper we discuss some topics related to knowledge representation, and the mechanisms that can be used to bring that knowledge to the user, mostly through the Web. This topic became relevant, when we faced the problem of publishing large amounts of data from several different sources, in the context of project GEIRA, an EC supported project under the INTERREG II program, that is developing a service to publish multimedia information about science and technology, cultural resources and environmental protection in the North of Portugal and Galiza (North-West Spain). The problem is not one of making a visible presence for each institution on the Web, but of taking advantage of the complete information delivered by them all. For the participating museums, we would like user to visit each museum but, more than that, to see related collections or objects, available in other museums. The diversity of institutions participating in the project GEIRA is so rich, that the user can see other information related to some objects, like the biography of the author, a list of publications citing the object, and even see (through a GIS plug-in) the archaeological site where an object was found. The major discussion in this paper is of two different models for data representation: the relational database and annotated documents. As we will show the former has advantages when manipulating structured data and has several sophisticated and affordable tools available. The later model is suited for less structured information, or at least, before a formal structure is identified. To be less structured does not mean that we cannot use useful tools, like thesaurus and encyclopedia views. As an open question, for future work, is the possibility of using some hybrid model, with the advantages of both data models.