Constructing cognitive and communicative models for dialog based on utterance-unit annotations

(1)通过语料库分析和听力实验，确定了与话语结束增量预测相关的句法和韵律因素。(2)通过对话语开头和其他地方的要素信息状态（new / given）的注释，阐明了它们与话语韵律和说话者认知负荷的关系。(3)研究了具有某些结构形式的问题对其他参与者可能的回答的相互作用约束。(4)研究了不同类型的响应令牌如何根据顺序上下文使用，以及它们在交互中的作用。(5)为了实现这些结果，基于新开发的模式，生成了丰富的对话语料库注释，并将其实现为关系数据库。

(1) Syntactic and prosodic factors relevant to the incremental prediction of utterance ending have been identified through corpus analyses and listening experiments. (2) Based on annotations of the information status (new vs. given) of elements in utterance beginning, and other places, their correlates with utterance prosody and speaker's cognitive load have been elucidated. (3) Questions with some forms of construction were investigated with respect to their interactional constraints on possible responses from other participants. (4) Various types of response tokens were examined as to how they are used depending on sequential contexts and to what effects they have in interaction. (5) To achieve these results, rich annotations of dialog corpora were produced and implemented as relational databases based on a newly-developed schema.