Identifying a matching component is a recurring problem in software engineering, specifically in software reuse. Properly generalized, it can be seen as an information retrieval problem. In the context of defining the architecture of a comprehensive software archive, we are designing a two-level retrieval structure. In this paper we report on the first level, a quick search facility based on analyzing texts written in natural language. Based on textual and structural properties of the documents contained in the repository, the universe is reduced to a moderately sized set of candidates to be further analyzed by more focussed mechanisms.