Optimizing the finite-state description of Estonian morphology

The research on modeling the Estonian morphology by finite state devices has been influenced mostly by (Koskenniemi, 1983), (Lauri Karttunen and Zaenen, 1992) and (Beesley and Karttunen, 2000). We have used lexical transducer combined with twolevel rules as a general model for describing Estonian morphology. As a novel approach we can emphasize the application of the rules to the both sides of the lexical transducer – both to the lexical representation and to the lemma. In the paper the criteria of optimality of the finite-state description of a natural language morphology and the means of fulfilling these criteria are discussed on the example of Estonian – a language with very rich and complex morphology. Other builders of finite-state morphological transducers may profit from the ideas proposed.