A Computational Model of Language Generation Applied to Japanese Wh-questions

A Computational Model of Language Generation Applied to Japanese Wh-questions⋆ Jason Ginsburg Center for Language Research, University of Aizu, Tsuruga, Ikki-machi, Aizu-Wakamatsu 965-8580, Japan jginsbur@gmail.com Abstract. This paper discusses a computational model of language generation, based on work in Phase Theory, that attempts to shed light on how the human mind generates sentences. This model presents explicit algorithms that a) determine selection and merger of Lexical Items, b) determine the labels of Merged elements, c) account for movement of Lexical Items within a derivation, and d) account for when chunks of a sentence are sent to Spell-Out. We demonstrate how this model accounts for generation of a wh-question in Japanese. Keywords: Phase Theory, computational modeling, wh-questions, Japanese 1 Introduction Examination of a theory via a computational model can be an enlightening exercise. A computer requires precision. Thus, ambiguious and/or conflicting rules can become immediately apparent, since these cannot be processed by a computer. While we acknowledge that the ability to model a theory on a computer does not necessarily correlate with the correctness of the theory, awareness of problems for a computational implementation of a theory and development of solutions to these problems can lead to a more refined and accurate theory. To this end, we developed a computational model of language generation in which sentences are derived from an underlying set of Lexical Items (LIs), via a process of selection and Merge, following a version of Phase Theory (Chomsky, 1999, 2000, 2004, 2006), the most recent implementation of work in the Minimalist Program (Chomsky, 1995b). In Phase Theory, a sentence is constructed via a bottom-up process whereby LIs are selected from a numeration, which consists of subnumerations, and Merged together. A derivation is bro- ken up into phases - v* (transitive v) and C (also possibly D) are phase heads. A derivation is subject to the Phase Impenetrability Condition (PIC), which determines when chunks of a deriva- tion are sent to Spell-Out (or to Transfer). Crucially, any elements contained within a chunk of a derivation that has been sent to Spell-Out are not accessible to higher operations. The PIC has been given (at least) two formulations (cf. Grewendorf and Kremers (2009)). In one version (Chomsky 1999, 2000), when v* is reached in (1), the complement of v*, which is VP, is sent to Spell-Out, and when C is reached, C’s TP complement is sent to Spell-Out. In another version of the PIC (Chomsky 2004), in (1), when C is reached, VP, which is the complement of v*, is sent to Spell-Out. (1) [CP C [TP T [vP∗ v* [VP . . . ]] Within Phase Theory, a head with an uninterpretable feature functions as a probe that Agrees with a goal (if present) that has a matching interpretable feature. Crucially, the probe and goal must be accessible to each other - a probe cannot Agree with a goal that has been sent to Spell-Out. ⋆ I thank Sandiway Fong for helpful discussion of various issues presented in this paper. …

Content truncated to 500 words