The paper is concerned with the problem of representing knowledge for expert-consultation systems in the area of information retrieval. The objective in such a system is to assist users, who have an information need, to select the right vocabulary terms for a database search. The knowledge base of the system originates mainly from the knowledge provided in the thesaurus, and it is represented as a semantic network where nodes are terms/concepts and links are various types of relationships between them. This knowledge base enables a program to conduct a search, to evaluate findings and to interact with the user - similar to work patterns of a (human) information consultant. The paper concentrates on the characteristics of the systems which utilize the knowledge are also discussed.