How the shapes of words adapt to communicative pressures in spoken language

语言是一个复杂的系统，在我们生活的几乎所有方面都扮演着关键的角色。考虑到它的多种作用——沟通的工具、抽象思维的载体、社会身份的标志——它的演变受到多种不同因素的影响也就不足为奇了。我研究的长期目标是了解口语的声音系统是如何受到口语交流功能的限制和塑造的。拟议的研究计划聚焦于一个具体问题：单词的形状如何适应高效交流的压力。该领域最近的研究表明，与不可预测的单词（如recover）相比，易于预测的单词（如go）也更短，而且产生起来更省力。使用和形状之间的这种联系符合人们在信息论优化下的期望。然而，一个关键的问题仍然没有答案：单词的形状和用法是如何演变成这种方式的？我自己的工作已经开始解决这个问题，但是到目前为止，已经确定的用法和形状之间共同进化的梯度模式不太可能是口语词汇中单词长度和信息内容之间强烈关联的来源。目前的提案提出了三个短期目标来解决这个问题。首先是使用语料库语音方法评估130多年来语音样本中单词形状的变化，作为强大的形状-用法关联的潜在来源。其次，我将使用自然语言处理的方法评估大型历史文本集合（例如b谷歌Ngram语料库）中单词用法的变化，作为形状-用法关联的潜在来源。第三个目标是使用迭代的人工语言学习实验来复制实验室中形状-用法关联的演变，并加强目标1和目标2中观察研究的结果。这项工作将在语音模式起源实验室（OoPS-Lab）进行，借助最先进的语音科学计算方法基础设施。研究团队将包括2名研究生HQP和6名本科生HQP，他们将接受计算和语音方法方面的培训，并接受专业指导。OoPS-Lab致力于创造一个多元化的研究环境，通过仔细考虑和以科学为基础的公平、多样性和包容性方法，支持成员的学术成长，无论他们的背景如何。通过揭示口语中交际效率优化的进化途径，这项工作将对认知科学和语言学产生广泛的影响：它将帮助我们理解人类语言结构的交际和认知来源。

Language is a complex system that plays a pivotal role in nearly all aspects of our lives. Given its many roles - a tool for communication, a vehicle for abstract thought, a marker of social identity - it is no surprise that its evolution is shaped by a multitude of diverse factors. The long-term goal of my research is to understand how the sound systems of spoken languages are constrained and sculpted by their function in spoken communication. The proposed program of research zooms in on a specific question: how word shapes adapt to pressures towards efficient communication. Recent work in the field has shown that words that tend to be predictable (e.g. go) are also shorter and produced with less effort than words that tend to be unpredictable (e.g. recover). This association between usage and shape is in line with what one would expect under information-theoretic optimisation. A key question, however, remains unanswered: how do word shapes and usage evolve to become aligned this way? My own work has started to address this question, but the gradient patterns of coevolution between usage and shape identified so far are unlikely to be the source of the strong association between word length and information content in the lexicons of spoken languages. The current proposal lays out three short-term objectives to address this issue. The first is to assess changes in word shape in speech samples covering over 130 years as a potential source of robust shape-usage associations using corpus phonetic methods. Second, I will assess changes in word usage in large collections of historical texts (e.g. Google Ngram corpus) as a potential source of shape-usage associations using methods from natural language processing. The third objective is to use iterated artificial language learning experiments to replicate the evolution of shape-usage associations in the lab and strengthen the results of the observational studies in Objectives 1 and 2. The work will be conducted at the Origins of Patterns in Speech Lab (OoPS-Lab) with the help of state-of-the-art infrastructure for computational approaches to speech science. The research team will involve two graduate HQP as well as 6 undergraduate HQP, who will receive training in computational and phonetic methods as well as professional mentorship. The OoPS-Lab is committed to creating a diverse research environment that supports members in their academic growth regardless of their background through a carefully considered and science-based approach to equity, diversity and inclusion. By uncovering the evolutionary pathways through which optimisation for communicative efficiency takes place in spoken languages, this work will have broad implications for cognitive science and linguistics: it will help us understand the communicative and cognitive sources of the structure of human languages.