Coordinating Information in Sentence Processing

协调句子处理中的信息

基本信息

批准号：
8512748
负责人：
Michael Kehas Tanenhaus
金额：
$ 29.08万
依托单位：
UNIVERSITY OF ROCHESTER
依托单位国家：
美国
项目类别：
财政年份：
1991
资助国家：
美国
起止时间：
1991-06-01 至 2015-06-30
项目状态：
已结题

项目摘要

DESCRIPTION (provided by applicant): The overarching goal of this project is to address both the methodological and theoretical challenges in real- time language processing in order to understand how listeners coordinate linguistic and non-linguistic information as they construct interpretations. Research from this project and research from other labs has strongly established that in addition to form-based probabilistic constraints, listeners make rapid use of information provided by visual context and action-relevant affordances of objects and task goals. It is less clear whether speakers and addressees take into account their interlocutor's knowledge and commitments during the earliest moments of language processing. Modeling the knowledge states of other individuals is presumably complex and resource-intensive. Inferring and tracking the knowledge states and intentions of interlocutors is often conceived of as a task separated from core real-time language processing. In contrast, we propose that targeted and probabilistic tracking of some aspects of interlocutors' knowledge states and intentions may be an essential part of language processing itself; interlocutors, in the linguistic choices they make, may be implicitly telling other individuals about elements of their own knowledge states and intentions that are relevant to the task of carrying on the current linguistic interaction. These choices of form by the speaker, and their interpretation by the addressee, involve integrating the content of the sentence with considerations of who knows (and does not know) what, a matter that relates more directly to the conduct of the discourse than to its descriptive content. We will examine processing of linguistic devices whose function is to shape the presentation of descriptive content rather than to form a description, including sentence-type, prosody, and forms of acknowledgment. The proposed research builds on our work with definite reference and our recent success in using eye movements to study real-time language processing in collaborative tasks. We address two central questions: (1) how and when interlocutors take into account each other's likely knowledge, both shared and privileged and (2) whether, and if so, when, interlocutors signal and monitor each other's intentions in conversation. The proposed work should make important empirical and theoretical contributions; it will helping resolve existing controversies and it will break new ground. We further develop the relatively novel eye-tracking targeted language games methodology and explore optimal cue-integration models and new methods for statistical modeling of eye-tracking data. Our results and methods should continue to: (a) advance our understanding of normal language comprehension; (b) inform investigations of language processing in children and special populations, including those with impairments that arise from brain injury, and they are beginning to influence development and evaluation of dialog systems with health-related applications.

描述（由申请人提供）：该项目的总体目标是解决实时语言处理中的方法论和理论挑战，以了解听众在构建解释时如何协调语言和非语言信息。该项目的研究和其他实验室的研究已经强烈确定，除了基于形式的概率约束外，听众还快速使用视觉上下文提供的信息，而与对象和任务目标相关的可容纳行动。尚不清楚说话者和收件人是否会考虑到对话者在语言处理的最早时刻的知识和承诺。对其他人的知识状态进行建模可能是复杂且资源密集的。推断和跟踪对话者的知识状态和意图通常被认为是与核心实时语言处理分开的任务。相比之下，我们建议对对话者知识状态和意图的某些方面进行针对性和概率跟踪可能是语言处理本身的重要组成部分。在他们做出的语言选择中，对话者可能会隐含地向其他人介绍其自身知识状态的要素和与执行当前语言互动的任务相关的意图。演讲者的这些形式选择及其对收件人的解释涉及将句子的内容与谁知道（并且不知道）的考虑，这与话语的行为更直接有关，而不是与其描述性内容有关。我们将研究其功能的语言设备的处理，其功能是塑造描述性内容的呈现，而不是形成描述，包括句子类型，韵律和确认形式。拟议的研究以我们的工作为基础，并以明确的参考以及我们最近使用眼动作来研究协作任务中的实时语言处理的成功。我们解决了两个中心问题：（1）对话者如何以及何时考虑彼此的可能知识，包括共享和特权，以及（2），如果是，如果是，何时，对话者在对话中发出信号并监视对话的意图。拟议的工作应做出重要的经验和理论贡献；它将有助于解决现有的争议，并将打破新的基础。我们进一步开发了相对新颖的目标语言游戏方法，并探索最佳的提示整合模型和新方法，用于眼睛跟踪数据的统计建模。我们的结果和方法应继续：（a）提高我们对正常语言理解的理解；（b）告知对儿童和特殊人群中语言处理的调查，包括脑损伤引起的损害的人群，并且他们开始影响与健康相关应用的对话系统的发展和评估。