Modelling Derivational Morphology: A Case of Prefix Stacking in Russian.

In order to automatically analyse Russian texts, one needs to model complex verb formation, as it is a productive mechanism and dictionary data is not sufficient. In this paper I discuss two implementations that aim to produce all and only the existing complex verbs built from the available morpheme inventory for the same fragment of Russian grammar. The first implementation is based on the syntactic theory approach to prefix combinatorics by Tatevosov (2009) and the other one uses the combination of basic syntactic restrictions and frame semantics to construct all possible combinations. I show that a combination of basic syntactic and semantic restrictions provides better results than a set of elaborated syntactic restrictions, especially for the complex verbs that are not normally tested by introspection.