Understanding typologies using Property Analysis: learnability, diachronic change, and formal structure

点击翻译按钮获取中文摘要

Languages change over time by accumulating small grammatical differences, and child language learners must be sensitive to these small differences. A robust theory of language change must then be able to articulate what are possible historical language changes and how these changes are transmitted intergenerationally. This project will address these issues by making concrete the notion of distance between languages and will determine how learners can utilize these small distinctions in learning. Dr. Merchant offers that there are parallels in the way that grammars change over time and learners navigate these changes to the way that RNA (ribonucleic acid) sequences change over time and the corresponding phenotypes produced by these RNA sequences. Genetic notions of distance between phenotypes are directly applicable to notions of distances between languages, and these distances are directly relevant for explaining language change. In addition to providing a better understanding of how languages change and how learners acquire these changes, the project will create a large repository of analyzed languages allowing researchers to build on the insights gained in this project. Furthermore, the project will provide valuable STEM training for undergraduate students from underrepresented groups who will help to develop the online database. The technical core of the project involves developing and extending formal theories of language, language learning, and language change by incorporating ideas from genetics, topology, cognitive science, and linguistics. In particular, the project will accomplish these tasks by developing and expanding three interrelated components: (1) software tools will be developed that will allow linguists to calculate and compare formal theories of grammars, determine distances and adjacencies between languages, including distances arising from the genetic structure of grammars and those arising from the properties of the grammars (in the sense of Alber & Prince 2017), and evaluate the outcomes of learning algorithms that rely on different notions of grammatical distance. (2) An online repository of diachronically relevant analyzed linguistic systems will be created. (3) The learning algorithms using ideas of grammatical nearness will be applied to the analyzed systems of the repository to evaluate diachronically realistic and informative learning algorithms. This work will produce new models of language learning, diachronic change, and will contribute a significant body of publicly available data.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.