User - friendly Information Retrieval in Data

The paper describes two methods for realizing a user-friendly access to distributed information ressources. The rst method (Method I) is based on a form-driven dialogue, which is used in the project named "MEDOC". It aims at an experienced user who is familiar with attribute value structures of the data base schemes of typical information retrieval systems (IRS) and who knows the de nition of boolean operators. The second method (Method II) applied in the system LINAS is from the very beginning oriented towards natural language communication between end-user and IRS. Both methods can be used in an interface between the user and an information brokering system helping him/her to nd an appropriate information provider for his/her demands in networked information systems. Method I gives the user a certain guidance in formulating his/her queries but has a restricted expressive power. It almost never supports the user in automatically nding more complicated descriptional elements, as for instance classi cators of a standardized classi cational system. Method II, on the other hand, is devoted to the "naive user" having no experience with information retrieval techniques. Allowing for an unrestricted natural language input, it is distinguished by a greater expressive power and gives valuable support in automatically nding descriptors and classi cational categories used in the description of documents. In comparison with Method I, there is less guidance in formulating the users demands.