Complex words in context

辨别性词汇模型(DLm，Baayen et al.，2019；Chuang&Baayen，2021)实现了心理词汇的计算理论。这一理论是针对单独考虑的单词而发展起来的，没有考虑任何上下文。然而，如何说话，以及它们的意思，取决于它们所在的语境。例如，英语Cut可以表示用电锯、刀子或剪刀进行的动作，荷兰语和普通话使用三个不同的动词来表示这些动作。英语中的Cut显示了广泛的其他含义，包括派生词(Cutter，一种船)、复合词(地沟虫，一种蛾子幼虫)、词汇化表达(Cut Cross)和习语(用来逃课)。Elman(2009)和Jackendoff&Audring(2020)指出，我们的词汇知识不是由简单和复杂的单词组成的，而是由成千上万的多个单词组成的。此外，在一种语言中用一个形态复杂的词来表达的东西，可能需要其他语言中的短语。本项目提案的核心是这样一个假设，即小话语的意义可以被表示为高维词汇/句法/语用空间中的点。这样一个融合了分布语义、形态和简单句法的空间，不仅需要开发算法来概念化屈折特征，如数字或时态，而且还需要开发算法来概念化句法角色，如施事和受事，以及语用功能，如敬语(如在韩语和日语中发现的)。至关重要的是，算法的建立必须以这样的方式进行，即在保持词汇相似性的同时，正确区分具有不同句法或语用功能的实体。将DLM从孤立的词扩展到上下文中的词不仅对基于用法的词法的DLM形式化表示形式和词汇意义的方式提出了挑战，而且还解决了语言学理论中关于形式和意义之间关系的基本问题。作为该项目的核心成果，增强模型将被应用于以词为基础的DLM目前无法处理的一系列现象，如联系、外部变音、词法范式中的边缘现象、复合口译、格屈折的解释、固定短语和习语、词汇意义的上下文相关性和敬语的形态。重要的是，由于DLM将分布语义学融入到形态理论中，因此有可能详细研究意义上的细微差异如何调节言语产生和听觉理解。

The Discriminative Lexicon Model (DLM, Baayen et al., 2019; Chuang & Baayen, 2021) implements a computational theory of the mental lexicon. This theory has been developed for words considered by themselves, without taking any context into account. However, how words are spoken, and what they mean, depends on the context in which they are used. For instance, English cut can denote actions car- ried out with chainsaws, knives, or scissors, actions for which Dutch and Mandarin use three different verbs. English cut displays a wide range of other meanings, across derivations (cutter, a type of ship), compounds (cutworm, a moth larva), lexicalized expressions (cut across), and idioms (to cut classes). Elman (2009) and Jackendoff & Audring (2020) have pointed out that our lexical knowledge does not consist of just simple and complex words, but of tens of thousands of multi-word expressions. Furthermore, what in one language is expressed with a single morphologically complex word, may require a phrase in other languages. Central to the present project proposal is the hypothesis that the meanings of small utterances can be represented as points in a high-dimensional lexical/syntactic/pragmatic space. Such a space, which integrates distributional semantics with morphology and simple syntax, requires the development of algorithms for the conceptualization not only of inflectional features such as number or tense, but also of syntactic roles such as agent and patient, and pragmatic functions such as honorifics (as found in Korean and Japanese). Crucially, the algorithms have to be set up in such a way that entities with different syntactic or pragmatic functions are properly distinguished, while maintaining lexical similarities. Extending the DLM from isolated words to words in context not only poses challenges for the way in which form and lexical meaning are to be represented within the DLM formalization of usage-based morphology, but also addresses fundamental questions in linguistic theory about the relation between form and meaning. The enhanced model that is the central deliverable of this project will be applied to a wide range of phenomena that are currently out of reach of the word-based DLM, such as liaison, external sandhi, periphrasis in morphological paradigms, compound interpretation, the interpretation of case inflection, fixed phrases and idioms, the context-dependence of lexical meaning and the morphology of honorifics. Importantly, because the DLM incorporates distributional semantics into morphological theory, it becomes possible to study in detail how subtle differences in meaning modulate speech production and auditory comprehension.