Proceedings of the Annual Meeting of the Cognitive Science Society

Many computational models of morphology represent complex words by n-grams to account for lexical processing and acquisition. However, while n-gram models are simple and efficient, they are not without problems. From a cognitive perspective, it is unclear how n-gram words are represented in the mental lexicon and how these representations affect language use and acquisition. From a computational perspective, these models are problematic because n-gram representations are often ambiguous and redundant: they make very limited use of distributional information and neglect the role of efficiency and sequential processing in language use and acquisition. In this paper, we present a new computational approach to morphology that is cognitively more plausible than standard n-gram models. By analyzing data from the nominal number system in German, we show that a task-specific algorithm of linear processing guided by the principles of efficiency and reliability outperforms state-of-the-art n-gram models and also makes predictions about lexical processing that are consistent with the judgments of German native speakers in a psycholinguistic experiment.