Morphological complexity and unsupervised learning: validating Russian inflectional classes using high frequency data

Abstract

This paper addresses the question of whether it is possible to use machinelearning techniques on linguistic data to validate linguistic theory. We determine how readily inflectional classes recognized by linguists can beinferred by an unsupervised learning method when it is presented with the paradigms of a small number (80) of high frequency Russian noun lexemes.We interpret this as a measure of the validity of the linguistic theory.Inflectional classes are of particular interest, because they constitute a kindof autonomous morphological complexity which has no direct relationshipto other levels of linguistic description, and hence there is no other objectiveway of assessing a theoretical characterisation of them. Using the samemethod, we also examine the status of principal parts and defaults ininflectional classes, and the relationship between inflectional classes andstress in Russian nominal morphology. Our experiments suggest that this isan effective and interesting technique for shedding additional light ontheoretical claims.