Predicting Grammatical Categories

We investigate to what extent the grammatical class(es) of a word can be predicted exclusively on the basis of phonological and prosodic information. We report on several experiments with an artificial learning system which has to assign English nouns and verbs to their appropriate grammatical class using various types of phonological and prosodic information. We replicate the claim articulated by Kelly (1996) that specific phonological cues are particularly good predictors of grammatical class. Our results indicate that these cues are partially predictive for the problem at hand. We show experimentally that ‘raw’ segmental and suprasegmental information (word stress) convey most information about the ‘phonology syntax’ interface. * Preparation of this paper was supported by a VNC project of NFWO NWO (contract number G.2201.96), and by a GOA grant.