In everyday life one has to take a variety of decisions. So there is a need for recommendation which information may be relevant and which is rather unimportant to support decision making. Frequently we find recommendation systems to assist clients in online-shops and other internet environments. The objective of these systems is the implementation of user-friendly interfaces with a high degree of personalization and efficient decision support. This paper evaluates several machine-learning techniques for recommendation systems which are suitable to find appropriate decision-relevant text documents like product descriptions or test reports. We propose recommendation systems which are based on an active data warehouse where we link adaptive user profiles and textual product descriptions.