During application development under reuse, much effort is spent in re
trieving suitable components. This paper presents an approach to retri
eval based on software descriptions of components, and on a mechanism
for computing the functional similarity of clusters of reuse candidate
s. Descriptions are given in a semi-formal fashion and using fuzzy wei
ghts for keywords. Algorithms for similarity computation are presented
, The experimental tool for this approach manages imprecise queries po
sed by developers against a software repository; results about reuse o
f object-oriented code based on the use of the tool are presented, giv
ing measures of recall and precision of the proposed approach.