The Development of Spoken Word Recognition

口语识别的发展

• 批准号: 7154137
• 负责人: JAMES L MORGAN
• 金额: $ 32.29万
• 依托单位: BROWN UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 1996
• 资助国家: 美国
• 起止时间: 1996-02-01 至 2009-11-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/7154137
• 关键词: Development; Spoken; Word; Recognition

DESCRIPTION (provided by applicant): Spoken word recognition is at the root of language comprehension and acquisition. For adults, recognition of known words (numbering in the tens of thousands) is fast, accurate, robust, and effortless. Yet the perceptual computations required are highly complex, as anyone listening to speakers conversing in an unfamiliar language begins to discern: what one hears is a babble lacking readily identifiable words. Nevertheless, within the first year of life, infants must develop the skills needed for spoken word recognition. Such skills are prerequisite to further acquisition of language, for if a language learner cannot break input utterances into their constituent words, it will be impossible to learn how these words fit together or what individual words mean. Studies proposed in this application address acquisition of the three chief components of spoken word recognition: segmentation of continuous speech into word-like chunks, representation of speech in a compact fashion that maximizes efficiency of processing, and identification of tokens as exemplars of particular lexical types. Issues considered include: How does developing lexical knowledge affect segmentation? How are knowledge-driven and signal-driven segmentation related? What are the constraints on properties of speech that infants initially entertain as contributing to lexical representations? The range of such properties narrows with development: what are the statistical properties of input that underlie this change and guide the emergence of native-language phonological categories? Adult word recognition is affected by number and frequency of competitors. How are competitive neighborhoods structured in the developing lexicon? How do processing deadlines and perceptual characteristics of speech tokens interact in word identification? The proposed research will employ a variety of empirical methods to provide converging measures of the development of spoken word recognition in infants from 6 to 18 months: headturn preference, conditioned head-turning, habituation, and intermodal preferential looking. Results from experimental studies with infants will be integrated with computational modeling to assess the feasibility of continuity theories of word recognition development, according to which fundamental processing architecture is shared by infants and adults.

描述（由申请人提供）：口语单词识别是语言理解和习得的基础。对于成年人来说，识别已知的单词（数以万计）是快速，准确，健壮和毫不费力的。然而，所需的感知计算是非常复杂的，因为任何人听到说话者用一种不熟悉的语言交谈时都会开始辨别：你听到的是一种缺乏容易识别的单词的胡言乱语。然而，在生命的第一年，婴儿必须发展口语识别所需的技能。这些技能是进一步习得语言的先决条件，因为如果语言学习者不能将输入的话语分解成其组成词，就不可能学习这些词如何组合在一起或单个词的含义。 本申请中提出的研究解决了口语识别的三个主要组成部分的获取：将连续语音分割成类似单词的块、以紧凑的方式表示语音以最大限度地提高处理效率，以及将标记识别为特定词汇类型的样本。考虑的问题包括：如何发展词汇知识影响分割？知识驱动和信号驱动的细分是如何关联的？婴儿最初认为有助于词汇表征的言语属性的限制是什么？这些属性的范围随着发展而缩小：是什么统计输入的属性，这种变化的基础，并指导母语语音类别的出现？成年人的单词识别受竞争者的数量和频率的影响。在不断发展的词典中，竞争性社区是如何构建的？在单词识别中，语音标记的加工期限和感知特征是如何相互作用的？ 拟议的研究将采用各种实证方法，以提供收敛的措施，在6至18个月的婴儿口语单词识别的发展：头转向偏好，条件性头转向，习惯化，和联运优先看。婴儿实验研究的结果将与计算建模相结合，以评估单词识别发展的连续性理论的可行性，根据该理论，婴儿和成人共享基本的处理架构。