依存消解、一致性计算与浮现和语段计算理论

Emergentist and Phase-Theoretic Approaches to Dependency Resolution & Agreement Computations

本文首先用O’G rady的浮现计算理论分析论元依存、wh-依存、所指依存等的消解,结合UG一致性数特征计算方法和Chom sky语段理论(PT)探讨一致性计算。阐明浮现计算理论的本质是效率驱动的线性计算系统,结合数据分析论证依存消解要遵循计算有效性原则。跨语言系统研究发现,浮现计算理论以自左向右(L→R)线性组装、第一时间消解依存实现降低工作记忆处理负荷的效率驱动为指导的计算原理与Sh iobara语段音系推导中自L→R线性结构组装的多重拼读(MSO)和递增推导提高计算效率的方法一致,而两者与Chom sky的语段理论(PT)以最小搜索(MS)降低计算系统CHL负荷(复杂度)、提高计算有效性的总的计算设计原理也是一致的。本文进一步分析论证得出:语段不可渗透条件(PIC)等局部性计算制约遵循限制探针搜索空间的最小搜索(MS)基本原则;优选计算的本质是效率驱动的最小化计算(MC)。

This paper first analyzes the resolutions of argument dependency,wh-dependency and referential dependency under O＇Grady＇s（2005,2008） emergentist theory（ET）,and then explores the agree computation based on UG number agreement computation,parallel assembly computational system and particularly Chomsky＇s（2001,2004,2005,2007,2008） Phase Theory（PT）.Based on data analysis,we illustrate that the nature of the linear computational system of emergentist theory is efficiency-driven,and dependency resolution must follow computational efficiency with significant implications for language acquisition.And we illustrate the phonological view of phases（Shiobara 2009） which requires left-to-right（L→R）structure-building in the computational component.In L→R derivation,linearization and prosodic phrasing originate from the Multiple Spell-Out（MSO） of the prosodically determined phase.And by systematic cross-linguistic comparative researches,we find that ET＇s efficiency-driven L→R linear processor with computational maxims realized by resolving dependency at first time with the aim of reducing working-memory processing burden are in accordance with PT＇s optimal computational maxims of raising computational efficiency and reducing computational complexity realized by minimal search（MS）.Additionally the incremental model in agreement computation also follows the L→R linear order.We further conclude that Phase Impenetrability Condition（PIC） and other locality constraints can be reduced to the minimal search（MS）,the fundamental principle in PT,and the nature of optimal computation is efficiency-driven minimal computation（MC）.