Linear Optimality Theory as a Model of Gradience in Grammar
Frank Keller · 2006
Abstract This chapter outlines the linear optimality theory (LOT), a variant of optimality theory primarily suggested by Keller to model gradient linguistic data. LOT is a framework intended to account for gradient judgment data; gradience in processing data and in corpus data has different properties from gradience in judgment data, and it is unlikely that the two types of gradience can be accounted for in a single, unified framework. The chapter provides a summary of the empirical properties of gradient judgments that motivate the design of LOT. It defines the components of an LOT grammar, and introduces the LOT notions of constraint competition and optimality. It gives a comparison with other variants of OT, particularly with standard OT and with harmonic grammar. This chapter contains a survey of more recent developments, such as probabilistic OT and variants of OT on maximum entropy models.