Study on the labeling of discourse structure and its electronic representation

目前的研究旨在研究如何创建、存档和交换记录的对话数据，重点关注标记或“标签”。今年，考虑了口语对话语料库的需求，并在对基本标签集的考虑基础上创建了一个原型样本语料库。口语对话语料库由3或4个部分组成。第一个组件是一组声音文件，其中包含两个不同扬声器发出的单独声音。每个文件代表一个对话，使研究人员能够对话语进行时间戳。第二个组件是一个电影文件，但在当前的技术状态下，它可以是可选的。第三部分是对语音事件的一组注释。注释包括正字法注释、语言（句法、语义）注释和语篇注释。从口语对话语料库交流的角度来看，目前的研究已经成功地表明，抄写本质上是一种标注，而不是被标注的对象。利用128个日语地图任务对话，提出并验证了一个TEI标签集。为创建和交换准备了指南。

The present research aims to examine the methods with which to create, archive and exchange the recorded dialogue data, focusing on the markup or "tags." This year, the requirements for the spoken dialogue corpus was considered and a prototypical sample corpus was created based on the consideration on the basic tag set.The spoken dialogue corpus consists of 3 or 4 components. The first component is a set of sound files, which contains the separate voices uttered by the two different speakers. Each file represents one dialogue, enabling researchers to timestamp utterances. The second component is a movie file, but at the current state of the technology, it can be optional. The third component is a set of annotations to the phonetic events. Annotations includes the orthographical transcription, linguistic (syntactic, semantic) annotations and discourse annotations. The current research has been successful in showing that the transcription is essentially a kind of annotation, not that which is annotated, from the perspective of the exchange of spoken dialogue corpora.A tag set was proposed and verified in term of TEI,using the 128 maptask dialogues in Japanese. A guideline is prepared for the creation and exchange.