Mechanisms of predictability in reference production: Effects of semantic role, utterance planning, and audience design

促进人类交流的核心认知能力之一是利用各种类型的指称表达（名称，描述和代词）来识别句子中充当谓词论元的对象，实体和概念的能力。交际的成功很大程度上取决于说话者如何在所有可能的指称表达中选择一个特定的谓词论元。人们擅长追踪世界上事件的可预测性。这种可预测性对语言的使用有很大的影响。例如，当信息是可预测的时，说话者倾向于使用更短和更简单的语言形式，例如代词（他/她）而不是更长的表达（Sue，女人），以及更多的声音减少发音而不是更长的发音。本研究探讨了一个领域，这种效果是有争议的：语义角色的可预测性。 如果一个说话者报告了一个事件，比如“Ann passed Sue the ball”，有些事件相对更可能是延续，例如“.苏投进了一个球“这源于Sue的语义角色：她是球的接收者，因此可能会被期望用球做些什么。但研究人员不同意这种可预测性对代词产生是否有影响。尽管有大量的证据表明，声学减少是受可预测性的影响，但还没有人研究过声学减少是如何受到语义角色可预测性的影响的。在这个由美国国家科学基金会资助的项目中，詹妮弗·阿诺德博士和她的学生旨在解决语义角色可预测性对参考形式的影响。 他们使用新的实验方法，旨在模仿自然语言生产的特点，和语料库分析来衡量语义角色的可预测性，从真实的语言使用。他们采取全面的方法，检查代词的使用和声学减少。语义角色的可预测性将在其他众所周知的效应（如句法可预测性）的背景下进行研究。 眼动追踪和时间分析将为参考产生的认知机制提供证据。这项研究将揭示人们如何跟踪世界上的信息并将其纳入日常行为（如语言）的基本问题。 它将提供一个培训研究生和本科生的框架，并将提供必要的信息，以发展一个参考资料生产的机械模型

Among the core cognitive abilities that facilitates human communication is the ability to exploit various types of referring expressions--names, descriptions, and pronouns--to identify the objects, entities, and concepts that serve as predicate arguments in sentences. Communicative success depends heavily on how speakers choose among the full range of potential referring expressions to single out a specific predicate argument. People are good at tracking the predictability of events in the world. Such predictability has strong effects on language use. For example, speakers tend to use shorter and simpler linguistic forms when information is predictable, such as pronouns (he/she) instead of longer expressions (Sue, the woman), and more acoustically reduced pronunciations instead of longer/emphatic ones. This study examines one domain where this effect is debated: semantic role predictability. If a speaker reports an event like "Ann passed Sue the ball," some events are relatively more likely continuations, e.g. "... and Sue shot a basket." This stems from the semantic role of Sue: she was the recipient of the ball, and therefore might be expected to do something with it. Yet researchers disagree whether this type of predictability matters for pronoun production. And despite extensive evidence that acoustic reduction is influenced by predictability in general, no one has looked at how acoustic reduction is affected by semantic role predictability.In this NSF-funded project, Dr. Jennifer Arnold and her students aim to solve the puzzle of semantic role predictability effects on reference form. They use new experimental methods that aim to mimic the characteristics of natural language production, and corpus analysis to measure semantic role predictability from real language use. They take a comprehensive approach, examining both pronoun use and acoustic reduction. Semantic-role predictability will be examined against the backdrop of other well-known effects, like syntactic predictability. Eye-tracking and timing analyses will provide evidence about the cognitive mechanisms of reference production.This study will shed light on fundamental questions about how people track information in the world and incorporate it into daily behaviors like language. It will provide a framework for training both graduate and undergraduate students, and will provide information necessary for developing a mechanistic model of reference production