This paper refers to the morphological analysis of words, as an important process in the domain
of natural language processing. We will present the classical solution, based on the use of inflected
paradigms and of an extended data base, containing all roots of the words, and then there are
emphasized some of the disadvantages of this method. Then we will present an original method, which dynamically generates the roots of words, using phonetic alternances in the context of flexionary rules.
There are also presented some optimisations of the morphologic analysis algorithm.