The typology of subordinate clauses: A case study

这个项目调查了可以从带有从句的句子中得出的推论，比如“我忘了锁门”或“我忘了锁门”。人们观察到，语言使用特定的动词形式来编码特定的推理。例如，上面的形式“to lock”意味着门没有锁，而使用形式“locking”则意味着相反。这种形式与意义/推理之间的映射已经在英语以及少数其他密切相关的语言中进行了研究，这些研究导致了关于人类语言如何将形式与意义联系起来的有趣假设。然而，目前的情况是不完整的，因为不清楚这些假设是否涵盖了它们赖以发展的有限集合之外的语言。本项目旨在通过研究一种未被充分研究和资源不足的语言的主题来解决这一差距，这种语言的从句与文献完备的语言中的从句在重要方面存在差异。该项目有两个主要组件：一个大规模数据收集组件和一个计算建模组件。数据收集组件使用一种新颖的分阶段启发方法，将标准的启发方法与基于语料库的方法和基于网络的实验交织在一起。计算建模组件开发了一个计算模型，用于查找跨多种不相关语言的谓词的底层语义类。这项研究不仅有助于揭示人类语言如何将形式与意义联系起来的更全面的图景；它还有助于开发对自然语言理解的应用技术至关重要的资源。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 inferences that can be drawn from sentences with subordinate clauses, such as 'I forgot to lock the door' or 'I forgot locking the door'. It has been observed that languages use specific verb forms to encode specific inferences. For instance, the form 'to lock' above implies that the door wasn’t locked, while using the form 'locking' implies the opposite. Such mappings between form and meaning/inference have been investigated in English as well as a small number of other closely related languages, and these investigations have led to interesting hypotheses about how human language connects form with meaning in general. The current picture is incomplete, however, since it is not clear whether these hypotheses cover languages outside the circumscribed collection on which they were developed. This project aims to address this gap by investigating the topic in an understudied and under-resourced language whose subordinate clauses differ in important ways from those found in well-documented languages.The project has two main components: a large-scale data collection component and a computational modeling component. The data collection component uses a novel staged elicitation methodology that interleaves standard elicitation methodologies with both corpus-based methods and web-based experiments. The computational modeling component develops a computational model for finding underlying semantic classes of predicates across multiple unrelated languages. This research not only helps to reveal a fuller picture of how human language connects form with meaning in general; it also serves to develop resources crucial for developing applied technologies for natural language understanding.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.