Feature Structure Formalisms: Mathematical And ComputationalFoundations For Linguistic Theory

特征结构形式主义：语言学理论的数学和计算基础

• 批准号: 8806913
• 负责人: Carl Pollard
• 金额: $ 8.68万
• 依托单位: Carnegie-Mellon University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1989
• 资助国家: 美国
• 起止时间: 1989-01-01 至 1990-07-20
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=8806913&HistoricalAwards=false
• 关键词: Feature; Structure; Formalisms; Mathematical; ComputationalFoundations

In the past several years, the mathematical objects known as feature structures have been employed by numerous research traditions in theoretical and computational linguistics (collectively referred to as "unification-based grammar") for modeling a wide range of phenomena in the domain of natural language. The principal investigator will explore the thesis that the formalism of feature structures, appropriately extended, is capable of serving as the foundation for mathematically precise and computationally tractable theories of natural language structure that have determinate empirical consequences. The researcher would then apply such extensions to the formalization of explicit "principle-based" linguistic theories and to the construction of practical NLP programs, focussing upon such centrally important linguistic phenomena as word order, binding, long-distance dependencies ("A-bar movement"), and discourse structure. Extensions to be investigated include, inter alia, the following: (1) different varieties of negative constraints (classical negation, intuitionistic negation, and apartness); (2) different varieties of type regiments (via the logic, via structural constraints on graphs, via sorting of algebraic models) and associated inheritance schemes (inclusion polymorphism); (3) realization of such notions as list value and set value (parametric polymorphism); (4) predicates other than equality (e.g. inequality, append, subset, union) and "building-in" of equational theories (generalized unification); (5) methods for introducing recursion; (6) infinitary methods (e.g. "functional uncertainty); and (7) dynamic interpretation.

在过去的几年里，数学对象被称为 特征结构已经被许多研究所采用 理论和计算语言学传统 （统称为“基于统一的语法”）， 模拟自然领域的各种现象， 语言 首席研究员将探讨的论文， 特征结构的形式主义，适当地扩展， 能够作为数学上精确的基础， 自然语言结构的计算易处理理论 具有确定的经验后果。 研究人员将 然后将这种扩展应用于显式 “原则基础”语言学理论与 实用的NLP程序，集中在这些核心重要的 语序、约束、长距离等语言现象 依赖关系（“A-杠运动”）和话语结构。 除其他外，将调查的扩展包括： 以下几点：（1）不同种类的负约束 （古典否定、直觉否定和分离）; (2)不同种类的类型团（通过逻辑，通过 图上的结构约束，通过代数模型的排序） 和相关的遗传方案（包含多态性）; (3)表值和集值等概念的实现 （参数多态性）;（4）相等以外的谓词（例如， 不等式、附加、子集、并）与方程“内建 理论（广义统一）;（5）引入方法 （6）无穷方法（如“函数不确定性”）; （7）动态解释。