Compound stress in English: Modelling the prosody of NN and NNN constructions

一般认为英语中的复合词在左边的成分上是重读的（例如bláckboard，wátchmaker），然而，在重音分配上有相当大的变化（例如silk tíe，麦迪逊大道，singer-sóngwriter），这在文献中是没有考虑到的。本研究将检验三个假设。语义假设认为，构成成分之间的语义关系决定了特定构式的重音模式。某些关系引发右手压力，其他人则引发左手压力。结构假说认为，形态结构，即真实的复合词，触发左重音，而句法短语触发右重音。词汇化是造成许多变异的原因。类比假说认为压力遵循类比模式。新的复合词被赋予重音，这是基于它们与现有复合词的相似性，即那些已经存在于说话者的心理词典中的复合词。该项目的目的是为英语复合词的可变重音行为提供一个解释。为了克服早期研究的缺点，三种不同类型的数据将被使用：大量的自然发生的数据从语音语料库，实验与母语人士，和计算建模。处理语音，形态，语义和句法的接口，该项目还将揭示新的光的问题，语言知识是如何在一个充分的语法和词汇模型表示。

lt is generally assumed that compounds in English are stressed an the left-hand member (e.g. bláckboard, wátchmaker), However, there is a considerable amount of variation in stress assignment (e.g. silk tíe, Madison Avenue, singer-sóngwriter) that is unaccounted for in the literature. Three hypotheses will be tested in this project. The semantic hypothesis states that it is the semantic relationship between the constituents that is responsible for the particular stress pattern of a given construction. Certain relationships trigger right-hand stress, others left-hand stress. The structural hypothesis states that morphological constructions, i.e. real compounds, trigger left-hand stress, while syntactic phrases trigger right-hand stress. Lexicalization is responsible for the many aberrant cases. The analogical hypothesis states that stress follows analogical patterns. New compounds are assigned stress an the basis of their similarity to existing compounds, i.e. those already present in the speaker's mental lexicon. The aim of the project is to provide an explanation for the variable stress behavior of English compounds. To overcome the shortcomings of earlier studies, three different types of data will be used: large amounts of naturally occurring data from speech corpora, ex­periments with native speakers, and computational modelling. Dealing with the interface of phonology, morphology, semantics and syntax, the project will also shed new light an the question of how linguistic knowledge is to be represented in a descriptively and explanatorily adequate model of grammar and lexicon.