Modeling Phonetic Structure Using Articulatory Dynamics

研究人员将继续开发语音结构的计算模型，并利用它来帮助理解各种语言中出现的某些发音同化的地方。该模型基于一种特殊的动态系统方法，称为任务动力学，该方法根据语音产生过程中发音机构运动的协调模式抽象地描述语音手势。包括给定话语的手势的相对定时和参数化由由模型的语言手势组件生成的手势分数来指定。为了将该模型应用于同化，研究人员计划从说英语、俄语、阿拉伯语和西班牙语的人那里收集关于手势结构(根据该模型的定义)作为各种条件的函数而表现出的变化的数据，例如语速和韵律结构。他们将对观察到的变化模式进行建模。由于该模型在手势规范和声学输出之间提供了明确的、原则性的联系，因此它可以用来测试变异模式如何引起同化感知。

The investigators will continue the development of a computational model of phonetic structure, and employ it to help understand certain kinds of place of articulation assimilation that occur in various languages. The model is based on a particular dynamical systems approach called task dynamics, which abstractly characterizes speech gestures in terms of coordinated patterns of articulator movements during speech production. The relative timing and parameterization of the gestures comprising a given utterance are specified by a gestural score that is generated by the model's linguistic gestural component. To apply the model to assimilation, the investigators plan to collect data on the variation exhibited by the gestural structures (as defined by the model) as a function of variety of conditions, such as speaking rate and prosodic structure, from speakers of English, Russian, Arabic, and Spanish. They will model the observed patterns of variation. Because the model provides an explicit, principled connection between gestural specification and an acoustic output, it can be used to test how patterns of variation give rise to the percept of assimilation.