Speech Reduction across Languages and Dialects

人类每天在随意交谈中发出的语音是令人难以置信的多种多样的，声音和整个音节要么改变，要么缺失。美国英语的听者在听到语境中的这种“简化语音”时并没有什么不寻常的地方；然而，说第二语言的人，甚至来自其他英语国家的听者，往往会发现美式英语的简化语音很难理解。目前的研究集中在说话者和听话者如何跨语言和方言使用减少的、自发的言语，以及这种言语如何阻碍甚至促进不同背景的说话者之间的交流。该项目将对说荷兰语、西班牙语、日语和英语三种方言的人进行测试，以确定1)缩写在多大程度上是语言特有的，是语法的一部分，而不是随机或物理决定的可变性；2)母语的发音模式是否在第二语言自发言语水平上影响语音变异性；3)方言对简化语音的理解有多大影响；以及4)熟练程度、年限、民族/民族认同感等对缩写的产生和理解有多大影响。最重要的理论问题是，什么是所学语法的一部分，什么是低水平可变性。此外，该项目将提供有关言语感知的范例模型、说话者第一语言和第二语言之间的相互影响以及发音规划等理论问题的数据。通过全球化、移民和电信，现代世界中的人类经常跨越语言背景进行互动。以英语为母语的人与非英语母语的人互动，不同英语方言的人相互互动。目前的项目解决了人类如何在交流环境中处理会话语言的可变性问题。通过该项目获得的关于自然、简化语音的详细知识将影响语音技术以及人类如何通过语音与计算机交互，从而有利于语音合成和自发语音的自动识别。由于该项目包括对实验参与者的英语水平和语言背景进行广泛的调查，它还将揭示哪些因素会使非母语听众更容易或更难理解用第二语言进行的会话演讲。该项目组成了一个协同的国际协作小组来回答这些问题。

The speech humans produce every day in casual conversation is incredibly varied, with sounds and whole syllables changed or missing. American English listeners notice nothing unusual when hearing such "reduced speech" in context; however, second-language speakers and even listeners from other English-speaking countries often find American English reduced speech difficult to understand. The current research centers on how speakers and listeners use reduced, spontaneous speech across languages and dialects, and on how such speech may hinder or even facilitate communication among speakers of different backgrounds. The project will test speakers of Dutch, Spanish, Japanese, and three dialects of English to determine 1) to what extent reduction is language-specific and part of the grammar rather than random or physically-determined variability, 2) whether the sound patterns of the native language influence phonetic variability at the level of spontaneous speech in the second language, 3) how strongly dialect affects understanding of reduced speech, and 4) how degree of proficiency, years of experience, strength of ethnic/national identity, etc. affect production and understanding of reduction. The overarching theoretical question is, what is part of the learned grammar and what is low-level variability. Furthermore, the project will provide data on theoretical questions about exemplar models of speech perception, mutual effects between speakers' first and second languages, and articulatory planning. Through globalization, immigration, and telecommunications, humans in the modern world often interact across language backgrounds. Native English speakers interact with non-native speakers and speakers of different English dialects interact with each other. The current project addresses how humans handle the variability of conversational speech in communicative situations. Detailed knowledge of natural, reduced speech, gained through this project, will impact speech technology and how humans interact with computers by voice, benefiting speech synthesis and automatic recognition of spontaneous speech. Because the project includes extensive investigation of English proficiency and language background for the experiment participants, it will also shed light on what factors make it easier or harder for non-native listeners to understand conversational speech in their second language. This project forms a synergistic international collaborative group to answer these questions.