自然言語計算における重力原理 : 疑問詞移動・優位性効果再考

概要

論文の詳細を見る
This paper basically argues for Pesetsky's (2004) hypothesis that Case is an uT (uninterpretable Tense feature) of DP (determiner phrase), call it uT analysis. The uT analysis treats the traditional distinction of structural Case such as [NOM] (nominative Case) and [ACC] (accusative Case) as an epiphenomenon, and instead treats Case as a single structural feature of uninterpretable Tense. By so doing, T, which bears the interpretable Tense feature (iT), checks off uT of DP, for the T feature of uT is redundant, according to the Economy Principle. At the same time, [iT] must not be deleted, because it is interpretable, according to the Conservation Law.There are complications that need to be solved in the uT analysis. Consider how the uT analysis derives a sentence such as `Who read the book?' Suppose the derivation reached the point:C [TPwho read the book].For the attractor C, there are two closest elements that can check off [uT] of C;the DP `who,' bearing [uT], and TP, bearing [iT]. The attractor C has three choices, i. e. , (a) C attracts `who,' (b) C attracts TP, or (c) C attracts both `who' and TP. If (a) is chosen, `who' is remerged in [Spec, CP], yielding the acceptable output `who read the book?' If (b) is chosen, C attracts TP, and in this case, C attracts T, yielding the unacceptable output `Did who read the book?' If (c) is chosen, C attracts `who' and T, yielding the unacceptable sentence `Who did read the book?' The right choice is (a). There are two questions that arise at this point: (1) Why is it that C ends up attracting the head T when C has a choice to attract TP? (2) If 'who' and TP are both equidistant from C and qualified as elements capable of checking off [uT] of C, why is it that C attracts `who,' and not TP? For question (1), the answer is simple:T is lighter than TP. Mass energy (M) of TP is larger because TP has more merge structures. The number of merge structures of the head T is 0. If one defines M as the number of merge structures, M of T is 0, like the photon and graviton (not discovered). Since M equals energy, C is forced to attract the lightest T, not heavy TP, given the Principle of Least Energy (Effort), a species of the Economy Principle. This is the last resort left for C. For the question (2), the answer is again simple: `who' is lighter than TP. M of non-branching DP is 0. M of TP is 3 (at least), counting a merge structure as M=1, given <TP, vP, VP>. The attractor C is forced to attract `who (M=0),' given the Principle of Least Energy.The Gravitation Law proposed in this paper explains gravitational mass (M) that is computed in the natural language system in the human brain. Each structure has M. The language computation obeys the Gravitation Law, which states the following:Attract the lightest. Attracting the lightest element saves energy, and the computation load remains the smallest. Given the least energy and the fastest speed of mother language acquisition by human infants, it is natural that Mother Nature created the information-processing system that obeys the Gravitation Law, a kind of Economy Principle.
桃山学院大学の論文
2007-06-08

自然言語計算における重力原理 : 疑問詞移動・優位性効果再考

スポンサーリンク

概要

著者

関連論文

スポンサーリンク