The thesis presents a semi-supervised morphology learner developed by extending Paramor (Monson, 2009), an unsupervised system, to accept easy to obtain manually provided data in the form of inflections with marked morpheme boundary. In addition, a hierarchical clustering framework allowing combination of multiple sources of information was developed as a part of the thesis. The approach was tested on Czech, Slovene, German and Catalan and has shown increased F-measure in comparison with the Paramor baseline.

This dissertation is an analysis of two types of nominal compounds in Turkish, primary compounds and synthetic compounds within the framework of Distributed Morphology. A nominal primary compound is formed by two nouns, and its meaning is largely determined by world knowledge. A synthetic compound, on the other hand, is formed by a noun and a derverbal noun, such that the former is a true argument of the latter. The meaning of such compounds is always compositional. In many languages, the structural difference between these two types of compounds is not immediately observable. However, in Turkish, a primary compound would be obligatorily marked with the compound marker, -(s)I(n) while a synthetic compound would never be marked as such. In this dissertation, I claim that primary compounds in Turkish are underlyingly possessive phrases, a claim that has been previously made by several others. My analysis differs from those previous analyses in that it maintains that -(s)I(n) figures in a morphological component that follows syntax but precedes PF. Such a post-syntactic analysis has a number of advantages as it can account for a wide range of descriptive observations about the behavior of -(s)I(n). I claim that -(s)I(n) and an agreement marker never form a sequence at any stage in the grammar. I test this claim in an experiment conducted with Turkish-speaking individuals with aphasia, and show that only a vanishingly rare number of -(s)I(n)-agreement sequences are attested in aphasic speech. My analysis of synthetic compounds in Turkish is based on three types of nominalizers and the types of categories they can select. I show that only event-denoting nominals can form true synthetic compounds. I also show that nominals that are derived directly from roots can never form true synthetic compounds, which casts doubts on roots as projecting categories. I also consider a third group of seemingly synthetic compounds, which have an overt complex verbal stem, and yet, fail to derive true synthetic compounds. Following Marantz (2013), I claim that such pseudo-synthetic compounds, in fact, have semantically null verbalizing morphemes, and therefore, the root and the nominalizing head are semantically adjacent at LF.

No description available.