SOURCE OF CROSS-LANGUAGE VARIATIONS IN SYLLABLE SYSTEMS

音节系统中跨语言变异的来源

• 批准号: 6070528
• 负责人: Melissa A Redford
• 金额: $ 3.09万
• 依托单位: UNIVERSITY OF TEXAS AT AUSTIN
• 依托单位国家: 美国
• 财政年份: 2000
• 资助国家: 美国
• 起止时间: 2000-03-01 至
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/6070528
• 关键词: language; mathematical model; multilingualism; phonology; speech recognition

It has been argued that the three most prominent cross-language syllable system patterns can be explained in terms of a few specific production and perception constraints (Redford, 1999), which are sufficient to explain the emergence if a week-organized, language-like syllable system from a randomly-organized system (Redford, Chen, Mikkulainen, 1998; 1999). Since the constraints on syllable patterns are universal, the question that arises from this research is how or why cross-language variation occurs. Two distinct possibilities are explored in the proposed research: (1) Variation occurs naturally as a result of the optimized process. If the production and perception constraints are fixed relative to one another, several good solutions (local minima) to the problem of constraint satisfaction may exist; (2) language communities may weight the same universal set of constrains differently because the constraints are flexible and often in conflict with one another. In this view, different languages represent different total solutions (global optima) to the problem of constrain satisfaction. Since we cannot exert direct control over language evolution, these possibilities are best examined via computational simulations. Thus a significant part of the work proposed is the development of a realistic, phonetically- and functionally-motivated model with which we will be able to stimulate the emergence of syllable systems. At first, genetic algorithms will be used to evolve vocabularies in which syllables and phonomes will co-evolve in response to functional constraints on the system. Later, a second level of evolution will be introduced to allow the model to match its output with actual language data and to adjust the weightings of the constraints associated with the output. The results of simulations conduct with these broad and theoretically-motivated model will help us understand how language variation occurs. The model will also give us insight into the evolution of the complex sound patterns of language and into the constraints that shape syllable systems. In future work, the model will be used to simulate the acquisition of speech in different language environments.

有人认为，可以用一些特定的生产和感知约束来解释三个最突出的跨语言音节模式（Redford，1999），这足以解释出现的出现，如果从随机的随机系统中使用了一周的，类似于语言的语言的音节系统（雷德福德，Chen，Chen，Mikkkulainen，1998; 1998; 1999）。由于对音节模式的限制是通用的，因此这项研究引起的问题是如何或为什么发生跨语言变化。在拟议的研究中探索了两种不同的可能性：（1）由于优化过程而自然发生变化。如果生产和感知约束是彼此固定的，则可能存在一些良好的解决方案（本地最小值）来解决约束满意度问题。 （2）语言社区可能会加重相同的通用限制集，因为约束是灵活的，并且经常相互冲突。在此观点中，不同的语言代表了不同的总解决方案（全局最佳），以解决限制满意度的问题。由于我们无法直接控制语言演变，因此最好通过计算模拟来检查这些可能性。因此，提出的工作的重要部分是开发了一个现实的，语音和功能动机的模型，我们将能够刺激音节系统的出现。首先，遗传算法将用于进化词汇，其中音节和音量将根据系统上的功能约束而共同进化。后来，将引入第二级的进化，以允许该模型与实际语言数据匹配其输出，并调整与输出相关的约束的权重。 通过这些广泛和理论上动机的模型进行模拟的结果将有助于我们了解语言变化的发生方式。该模型还将让我们深入了解语言复杂声音模式的演变，并使我们介绍了塑造音节系统的约束。在将来的工作中，该模型将用于模拟在不同语言环境中的语音获取。