Relevant information is required in all kinds of information-based systems. It is not a trivial task to locate the right piece of information since it may be difficult (or impossible) specify query requests precisely and completely. Thus, a flexible retrieval algorithm is required, allowing for an imprecise query specification. The document presents an approach to judging relevance of retrieved information based on a novel approach to similarity assessment. Contrary to other systems, we define relevance measures (context in similarity) at query time. This is necessary if since without a context in similarity one cannot guarantee that similar items will also be relevant.