CONNECTIONIST MODELS IN COGNITIVE NEUROLINGUISTICS

认知神经语言学中的联结模型

• 批准号: 2124301
• 负责人: Steven L Small
• 金额: $ 8.84万
• 依托单位: UNIVERSITY OF PITTSBURGH AT PITTSBURGH
• 依托单位国家: 美国
• 财政年份: 1991
• 资助国家: 美国
• 起止时间: 1991-07-01 至 1996-06-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/2124301
• 关键词: brain disorders; computer simulation; data collection; dementia; human subject; language disorders; model design /development; nervous system disorder; neural information processing; neuroanatomy; patient /disease registry; semantics; speech; speech therapy; statistics /biometry; verbal behavior

The individual projects outlined in this proposal are aimed at the description of an overall processing architecture of the human neurolinguistic system. As research progresses toward this goal, theoretical notions motivated by empirical studies will be tested with connectionist network models, and concomitantly, predictions from network models will be tested with empirical studies. The desired result is the development of a general theory of cognitive neurolinguistic processing that is both testable and falsifiable. The proposed research and training uses research methods from experimental psychology, speech and language pathology, neuroanatomy and neuropathology, and computer modelling to construct useful cognitive theories of language disorders. Specifically, the project described here proposes to (a) organize a language disorders registry and database and implement it at Presbyterian-University Hospital (PUH); (b) design and execute a study of semantic priming in patients with language disorders; (c) extend an existing prototype connectionist simulator (DYSNET) to permit network lesions of both diffuse and focal types, in order to simulate neurological damage; and (d) perform computer modelling experiments leading to predictions for the design of empirical studies and suggestions for language therapy. Initial modelling experiments will be based on semantic priming data from a study of language disorders in diffuse brain disease. It will first account for the data from normal subjects (independent of age) that supportive context makes sentence completion tasks easier (i.e., faster); subsequently, it will be systematically disrupted (lesioned) to account for the results in the patients with language disorders, and lastly, the model will be examined for approaches to repairing lesioned networks, with attempts to correlate computational strategies and clinical language therapy. In the final two years of the project, the laboratory framework for building computer models should be sufficiently operative to initiate experiments in individualized patient models, with full-scale attempts to use the results of such tests in specific individually tailored language rehabilitation efforts.

本提案中概述的各个项目旨在 对人类的总体处理架构的描述 神经语言系统 随着研究朝着这一目标的进展， 由实证研究激发的理论概念将被测试， 连接主义网络模型，以及随之而来的网络预测， 将通过实证研究对模型进行检验。期望的结果是 发展认知神经语言学处理的一般理论 既可检验又可证伪的理论 拟议的研究和培训使用的研究方法， 实验心理学、言语和语言病理学、神经解剖学和 神经病理学和计算机建模来构建有用的认知 语言障碍理论 具体来说，这里描述的项目 建议（a）组织一个语言障碍登记处和数据库， 在长老会大学医院实施;（B）设计和 对语言障碍患者进行语义启动研究; (c)扩展现有的原型连接模拟器（DYSNET）， 允许弥漫性和局灶性两种类型的网络病变，以便 模拟神经损伤;以及（d）进行计算机建模 为实证研究设计提供预测的实验 以及语言治疗的建议 最初的建模实验将基于语义启动数据 一项关于弥漫性脑疾病中语言障碍的研究。 它将 首先解释来自正常受试者（与年龄无关）的数据， 支持性上下文使句子完成任务更容易（即，faster）; 随后，它将被系统地破坏（损坏）， 语言障碍患者的结果，最后， 模型将被检查的方法来修复受损的网络， 试图将计算策略和临床语言联系起来 疗法 在项目的最后两年， 建立计算机模型应该足够有效， 在个体化患者模型中进行实验，全面尝试 使用这种测试的结果，在特定的个别定制的语言 康复工作。