Speaker: Patrick Niedzielski (MIT)
Title: Distributed Morphology is more powerful than Finite-State Morphology, but not much more
Time: Wednesday, November 3rd, 5pm – 6:30pm
Abstract: Computational research in morphology has focused on finite-state technologies as grammatical descriptions (Beesley & Karttunen 2003), placing the computational power of the morphology on par with that of the phonology. At the same time, the rise of Syntax-All-The-Way-Down theories of morphology such as Distributed Morphology seem to complicate this picture, by making the syntactic component handle the word-formation aspects of the morphology. As the syntax is mildly-context sensitive (Shieber 1985), this would seem to place the computational power of the morphology on par with the more powerful syntax rather than the weaker phonology. To investigate this, we introduce a simple extension of Minimalist Grammars (Stabler 1997, 2001), called Minimalist Grammars with Complex Words (MGCWs), which enable us to reason formally about the set of words that head movement, Distributed Morphology’s primary word-formation operation, can build. We find that the generative capacity of head movement is precisely the class of linear context-free languages, which is strictly larger than the set of languages produced by finite-state morphology, but strictly smaller than the class of mildly context-sensitive languages the syntax is otherwise able to produce. This suggests even if morphological structure is built by the syntax, its power is still quite weak.