We exploit the combinatorial structure of conceptual graphs in order to obtain faster execution time when computing projection, which is a core generalisation-specialisation relation over conceptual graphs. We show how the problem of finding this relation can be translated into the Maximum Clique problem. Consequently, approximation techniques developed for the Maximum Clique problem can be used to compute projection in conceptual graphs. We show that there are “simple queries” which can be answered quickly, thus providing efficient reasoning support in a knowledge management environment based on conceptual graphs.