Doctoral Dissertation Research: Extending and testing theories of language production by investigating speaker choice in a classifier language

自然语言通常为说话者提供多种方式来传达相同的含义。同时，语言交际也受到环境和认知的制约。当有多种选择来表达或多或少相同的意思时，什么一般原则支配说话者的选择？推进我们对这个问题的理解可以潜在地增强一系列广泛的人类语言技术，例如提供更像人类的语言生成，更好地理解说话者的选择，更准确的机器翻译，更好的语言学习和教学资源，以及改善语言障碍治疗的见解。在这个更广泛的研究计划，该项目的重点是上下文的可预测性的语言内容的编码表现在分类语言的扬声器选择的影响。在英语中，数词直接修饰名词（例如，三张桌子）。在诸如汉语普通话之类的分类语言中，必须使用带有数字和名词的分类器（CL）（例如，三CL.平面表，三CL.通用表）。虽然不同的名词与不同的具体量词兼容，但有一个通用量词“个”（CL.general）可以与大多数名词一起使用。本研究的重点是使用通用分类器与具有相同名词的特定分类器之间的交替选择，其中选项几乎在语义上不变。使用更具体的分类器将减少名词的重复，但是如果访问通用分类器需要更少的努力，则从生产的角度来看，可以不使用更具体的分类器。该项目结合了语料库分析，心理语言学行为实验和计算建模，使用统计学，自然语言处理和实验心理学的技术，研究语言使用者如何分配资源，使他们能够产生和理解语言，该奖项反映了NSF的法定使命，并被认为值得通过使用基金会的学术价值和更广泛的影响评审标准。

Natural language often gives speakers multiple ways to convey the same meaning. Meanwhile, linguistic communication takes place in the face of environmental and cognitive constraints. When multiple options are available to express more or less the same meaning, what general principles govern speaker choice? Advancing our understanding of this question can potentially enhance a broad array of human language technologies, such as providing more human-like language generation with better understanding of speaker choice, more accurate machine translation, better resources for language learning and teaching, as well as insights to improve treatment for language disorders. Within this broader research program, this project focuses on the influence of contextual predictability on the encoding of linguistic content manifested by speaker choice in a classifier language. In English, a numeral modifies a noun directly (e.g., three tables). In classifier languages such as Mandarin Chinese, it is obligatory to use a classifier (CL) with the numeral and the noun (e.g., three CL.flat table, three CL.general table). While different nouns are compatible with different specific classifiers, there is a general classifier 'ge' (CL.general) that can be used with most nouns. This study focuses on the alternating options between using the general classifier versus a specific classifier with the same noun where the options are nearly semantically invariant. The use of a more specific classifier would reduce surprisal at the noun, but the use of that more specific classifier may be dispreferred from a production standpoint if accessing the general classifier requires less effort. This project combines corpus analyses, psycholinguistic behavioral experiments and computational modeling using techniques from statistics, natural language processing, and experimental psychology, examining how language users allocate resources to prepare them to produce and comprehend language, shedding lights on why language is structured in the way it is.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.