REPRESENTATION AND PROCESSING OF WORDS AND SENTENCES

单词和句子的表示和处理

• 批准号: 6302618
• 负责人: MARK S SEIDENBERG
• 金额: $ 12.42万
• 依托单位: CARNEGIE-MELLON UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2000
• 资助国家: 美国
• 起止时间: 2000-03-15 至 2001-02-28
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/6302618
• 关键词: behavior prediction; behavioral /social science research tag; clinical research; cognition; comprehension; computational neuroscience; computer simulation; human subject; language; model design /development; neural information processing; neuropsychology; performance; psychomotor reaction time; reading; semantics; sensory discrimination; speech; syntax; vocabulary

This project will examine basic phenomena concerning word and sentence processing using a set of theoretical principles derived from the connectionist/parallel distributed processing framework. The proposed research is a continuation of the previous project "Normal and Disordered Word Identification and Naming," which resulted in the development of a general theory of word recognition in reading based on these principles, the implementation of computational models that instantiate parts of the theory, the assessment of such models against behavioral data, and the acquisition of new data testing predictions derived from this an competing theories. The specific aims of the project are: 1. To develop current computational modeling architechtures that will address a) division of labor between components of the word recognition system in determining the meanings of words b) phonological representation and the reading of multisyllabic words, the assignment of syllabic stress; and c) morphological structure and the reading go of multimorphemic words; focusing on variations in the degree of morphological transparency. 2. To extend the framework to questions concerning the processing of words in sentences. Specific foci will include a)the representation of verbs and their argument structures; b) the use of multiple, interacting, probabilistic constraints in resolving local syntactic ambiguities; and c)the effect on sentence processing of structural constraints owing to the architecture of the language comprehension system and the sequential nature of language comprehension. 3. To develop a next-generation architecture that will be used to address issues related to the temporal nature of language, focusing on a) how underlying representations for temporally structured knowledge develop under the constraints imposed by the serial nature of comprehension and production; and b) how the degree of shared structure among constituents (e.g.; phonemes, morphemes, words) within the representations provides insight into graded similarity effects in human language performance. 4. In addition to this theoretical and computational work, we will also acquire critical behavioral data in each area, focusing on individual differences in division of labor; peoples~ performance in assigning syllabic stress to novel words; and effects of degree of morphological transparency in cross-modal priming.

该项目将研究有关词和词的基本现象 使用源自以下的一组理论原理进行句子处理 连接主义/并行分布式处理框架。 这 拟议的研究是之前项目“正常 以及无序的单词识别和命名”，这导致 阅读中单词识别一般理论的发展 基于这些原理，计算的实现 实例化部分理论的模型，对此类的评估 针对行为数据的模型以及新数据的获取 测试由此得出的预测和相互竞争的理论。 这 该项目的具体目标是： 1. 开发当前的计算 建模架构将解决 a) 之间的分工 单词识别系统的组成部分，用于确定 单词的含义 b) 语音表征和阅读 多音节词、音节重音的分配；和c) 形态结构和多形态的阅读 字;关注形态程度的变化 透明度。 2. 将框架扩展到以下问题： 句子中单词的处理。 具体重点包括a) 动词及其论证结构的表示； b) 使用 解决局部问题时存在多个相互作用的概率约束 句法歧义； c）对句子处理的影响 由于语言架构的结构限制 理解系统和语言的顺序性 理解。 3. 开发下一代架构 将用于解决与时间性质有关的问题 语言，重点关注 a) 的底层表示如何 时间结构化知识在约束下发展 由理解和生产的连续性所强加；和 b) 各组成部分之间的共享结构程度如何（例如； 表示中的音素、语素、单词）提供 深入了解人类语言表现的分级相似性效应。 4. 除了理论和计算工作外，我们还将 还获取每个领域的关键行为数据，重点关注 分工的个体差异；人们~的表现 为新词分配音节重音；以及程度的影响 跨模式启动的形态透明度。