Modeling Morphological Subgeneralizations


Languages' lexica have internal structure: lexical items which share phonological features tend to pattern together with respect to exceptionful or lexically conditioned morphophonological processes (Albright and Hayes, 2003; Zuraw, 2000). We present results from a computational model of lexically conditioned morphology in which lexical items form 'bundles' based in part on phonological similarity. Each bundle is indexed to an 'operational constraint', a set of operations which realize a particular morpheme. These constraints compete with markedness and faithfulness constraints in a Maximum Entropy (MaxEnt) grammar (Goldwater and Johnson, 2003). We use the English past tense system as a test case for this model.


Paper in the Proceedings of Phonology-2013
Slides from Phonology-2013

Structure of the Model

    Structure of the model.