Distributed Morphology as a regular relation
Marina Ermolaeva · ScholarWorks@UMassAmherst (University of Massachusetts Amherst) · 2018
This research reorganizes the Distributed Morphology (DM) framework to work over strings. Typically, DM operates on binary trees, with the syntax-morphology interface implicitly treated as a tree-transducer. We contend that using (binary) trees is overpowered, predicting patterns unattested in natural language. Assuming the standard Y-model, DM operating on trees presumes that the flattening of the derivation for PF takes place post-morphology. We however flatten the structure above the morphological module, between the syntax and morphology. Restricting the morphological component to working on strings, we correctly predict that morphology can be modeled with regular string languages.