Interactive processing in speech perception

语音感知中的交互处理

• 批准号: 6792968
• 负责人: Daniel Mirman
• 金额: $ 4.27万
• 依托单位: CARNEGIE-MELLON UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2004
• 资助国家: 美国
• 起止时间: 2004-01-01 至 2005-12-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/6792968
• 关键词: adult human (21+); attention; behavioral /social science research tag; clinical research; computer simulation; human subject; language; learning; neural information processing; phonology; predoctoral investigator; speech; speech recognition

DESCRIPTION (provided by applicant): One of the central dichotomies in theories of cognitive processing is between autonomous theories based on strictly feed forward processing and interactive theories that posit bi-directional flow of information. Speech perception is a particularly interesting battleground for this debate because there is now a vast store of evidence demonstrating lexical level effects on phoneme identification. Some researchers have argued that interactive processing in speech perception is not supported by existing data and that an interactive model is lot consistent with the data. There are three challenges that have not been addressed by the interactive view: lack of lexical inhibition, effects of shifting attention, and lexically mediated phonetic category tuning. The proposed research addresses the lack of lexical inhibition using simulations to show that an interactive model is consistent with the existing data and by experimentally testing predictions from further simulations. A model extension is proposed that adds a mechanism for shifting attention within an interactive framework and proposed experiments test predictions from this mechanism. Finally, proposed simulations test a learning mechanism designed to account for lexically mediated tuning of phonetic categories.

描述（由申请人提供）：认知处理理论中的一种核心二分法是基于严格喂养前进处理的自主理论和互动理论，这些理论认为信息的双向流动。语音感知是这场辩论的特别有趣的战场，因为现在有大量证据证明了对音素识别的词汇水平影响。一些研究人员认为，现有数据不支持语音感知中的交互式处理，并且交互式模型与数据一致。交互式观点尚未解决三个挑战：缺乏词汇抑制，转移注意力的影响以及词汇介导的语音类别调整。拟议的研究解决了使用模拟缺乏词汇抑制，以表明交互模型与现有数据一致，并通过实验测试进一步模拟的预测。提出了一个模型扩展，该模型扩展是在交互式框架中转移注意力的一种机制，并提出了从该机制测试预测的实验。最后，提出的模拟测试了一种学习机制，旨在说明语音类别的词汇调整。