Sensitivity to Meaningful Morphological Information Acquired through Reading Experience

英语和其他语言中的大多数单词都是由称为语素的较小意义单位组合而成的(例如，lean+ly，un+lean)。理解一种语言的词法是如何工作的是至关重要的，因为它允许我们进行概括；例如，我们可以理解‘Mislean’，因为我们知道[mis-]和[lean]作为有意义的元素发挥作用。研究表明，语素知识为词汇增长提供了重要的启发，而且这种知识有助于成人的快速阅读理解。这个项目的目的是发现我们如何获得词缀语素的抽象知识(例如-ify，-ly)。这些单元通常不是单独出现的，所以它们的功能必须通过对整个单词的经验来推断(例如，Purify、Falsify、Simplify)。最近的理论表明，我们学习这些单位是因为它们提供了关于词义的强大信息(例如，-ify的意思是‘制造[词干]’)。然而，这些理论在很大程度上是通过使用简单的微型语言进行的实验室实验发展起来的。语素在真实语言中传达意义的方式要复杂得多：我们不知道是什么驱使我们在野外学习，也不知道孩子们正在学习的是什么。这个项目开发了两个研究流来量化儿童的语言经验如何塑造他们的语素知识。我们的第一个研究流提供了一种前所未有的尝试，以量化儿童文学中词缀信息的性质。我们将建立一个大规模的儿童语料库，包括适合7至16岁的书籍，并将开发理论驱动的度量标准，以捕捉词缀信息的性质。这些指标的发展将利用新的计算技术，使我们能够捕捉到语素如何以比以前更丰富、更细微的方式对意义做出贡献。我们将计算整个语料库的这些指标，但也会计算适合不同年龄段的材料的文本。后一种分析很重要，因为在这个年龄段的书籍中使用的词汇可能会有很大的变化(特别是对较长的、形态复杂的单词)，这将影响儿童对个别词缀的学习。然后，我们的第二个研究流探讨了在第一个研究流中发现的词缀规则如何影响儿童的语素知识。我们将进行一项大规模的行为研究，测量7到16岁儿童的词缀知识。这项工作是最先进的，因为它让我们超越了相对较小的儿童群体是否表现出形态影响的问题。相反，我们的预测是，语素知识应该是(A)单个儿童的语言经验的函数；以及(B)特定词缀对词义贡献的统计规律性。基于项目的措施将从语料库分析中得出；基于参与者的措施将从各种语言和扫盲措施中得出。这项工作的直接结果将是一种新的理论，即语言经验如何支持语素知识的获得。这一新理论将有助于我们更广泛地理解语言经验如何塑造语言知识，并将激发人们对为什么阅读可能是语言经验的一个特别重要的来源的新思考。该项目还将产生一个新的儿童文本语料库和由该语料库建立的高维语义表示。这些产品将尽可能以用户友好的界面提供，以方便今后的研究。

The majority of words in English and in other languages are built by combining smaller units of meaning called morphemes (e.g. clean+ly, un+clean). Understanding how a language's morphology works is vital because it allows us to generalise; for example, we can understand 'misclean' because we know that [mis-] and [clean] function as meaningful elements. Research suggests that morpheme knowledge provides an important heuristic for vocabulary growth, and that this knowledge facilitates rapid reading comprehension in adults. The aim of this project is to discover how we acquire abstract knowledge of affix morphemes (e.g. -ify, -ly). These units typically do not occur in isolation, so their functions must be inferred through experience with whole words (e.g. purify, falsify, simplify). Recent theories suggest that we learn these units because they provide powerful information about word meaning (e.g. -ify means 'to make [stem]'). However, these theories have been developed largely through laboratory experiments using simple miniature languages. The way that morphemes communicate meaning in real language is far more complex: we do not know what drives learning 'in the wild' or what it is that children are learning. This project develops two research streams to quantify how children's language experience shapes their morpheme knowledge. Our first research stream offers an unprecedented attempt to quantify the nature of affix information in children's literature. We will build a large-scale children's text corpus including books suitable for ages 7 to 16, and will develop theoretically-driven metrics that capture the nature of affix information. The development of these metrics will capitalise on new computational techniques that permit us to capture how morphemes contribute to meaning in a richer and more nuanced way than has previously been possible. We will calculate these metrics across the whole corpus but also as text accumulates across material suitable for different age bands. This latter analysis is important because the vocabulary used in books across this age range is likely to change considerably (particularly for longer, morphologically-complex words), and this will influence what children can learn about individual affixes. Our second research stream then probes how the affix regularities uncovered in the first research stream influence children's morphemic knowledge. We will conduct a large-scale behavioural study that measures affix knowledge in children between the ages of 7 and 16. This work advances the state-of-the-art because it allows us to move beyond questions of whether or not relatively small groups of children show morphological effects. Instead, our prediction is that morpheme knowledge should be a function of (a) an individual child's linguistic experience; and (b) the statistical regularity with which specific affixes contribute to word meaning. Item-based measures will be derived from the corpus analysis; participant-based measures will be derived from a variety of measures of language and literacy. The immediate outcome of this work will be a new theory of how language experience underpins the acquisition of morphemic knowledge. This new theory will contribute more broadly to our understanding of how language experience shapes language knowledge, and it will stimulate new thinking about why reading might be a particularly important source of language experience. This project will also produce a new children's text corpus and high-dimensional semantic representations built from that corpus. These products will be made available in user-friendly interfaces to the fullest extent possible to facilitate future research.