Doctoral Dissertation Research: Morphological Dependencies

这个项目调查说一种语言的人是如何知道一个单词的各种形式的。几十年来，语言学家已经确定，说话者可以学习和概括语言中的统计趋势，从而创造出他们以前从未遇到过的新形式的单词。另一方面，在某些情况下，说话者必须记住不遵循共同模式的个别形式的行为。这个项目通过考察说话者在考虑如何制造另一种形式时是否考虑到一个单词的一种形式，来探索泛化和记忆之间的相互作用。它的目的是表明说话者可以形成这些类型的概括，从而提高我们对说话者如何学习和维持一个词可以以多种不同形式出现的复杂系统的理解。在这个项目中研究的四种语言中，每一种语言都涉及到一种情况，即一个单词的两种形式都表现出可预测性：如果一个单词在一种上下文中带有后缀a，那么它在另一种上下文中也可能带有后缀B。这些统计倾向将首先通过分析该语言的词典或文本语料库来建立。接下来，将与使用相关语言的人进行一系列实验。在每个实验中，说话者将在不同的语境中看到看似合理但未经证实的带有不同后缀的母语单词，并被要求提供额外的后缀形式。给出的形式会有所不同；他们的假设是，说话者也会类似地改变他们为新形式选择的后缀，大致模仿他们语言的统计趋势。为了解释这些相关性，该项目开发了一种理论方法，其中一个单词的形式与该单词的心理入口上的标记相关联，并且说话者学习概括哪些标记倾向于在单个入口中同时出现。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This project investigates how speakers of a language know the various forms a word can take. For many decades, linguists have established that speakers can learn and generalize over statistical trends in the language to create new forms of words that they have never previously encountered. On the other hand, in some cases speakers must memorize the behavior of individual forms that do not follow common patterns. This project explores the interplay between generalization and memorization by examining whether speakers take one form of a word into account when figuring out how to make another form. It aims to show that speakers can form these sorts of generalizations, thus improving our understanding of how speakers learn and maintain complicated systems in which a word can appear in many different forms.Each of the four languages studied in this project involves a case where two forms of a word show interpredictability: if a word takes suffix A in one context, it is also likely to take suffix B in a different context. These statistical tendencies will first be established by analyzing a dictionary or corpus of texts in the language. Next, a series of experiments will be run with speakers of the relevant language. In each experiment, speakers will be shown plausible but unattested words of their language with different suffixes in a number of contexts and asked to provide an additional suffixed form. The given forms will vary; the hypothesis is that speakers will similarly vary the suffixes they choose for the new forms, roughly mimicking the statistical tendencies of their language. To account for these correlations, this project develops a theoretical approach in which the forms of a word are associated with markers on that word's mental entry, and speakers learn generalizations over which sets of markers tend to cooccur within a single entry.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.