Research on Discourse Understanding based on Discourse Structure

本研究旨在发展一种语篇理解理论，尤其是对话中信息如何流动的理论，并建立一个计算机与人类对话的计算模型。本文以分段语篇表征理论为理论框架，从语言学的理论研究视角和计算语言学的应用视角。由于研究的主要对象是日语语篇，我们不得不将语篇关系扩展到包括日语语篇结构的各个方面，并发现分析和解释语篇中描述的信息之间的关系是有效的，特别是短小的信息，如推文。

This research was intended to develop a theory of discourse understanding, especially how the information is flowed in dialogue, and to build a computational model of dialogue between computer and human. Adopting SDRT(Segmented Discourse Representation Theory) as its theoretical framework, we took the theoretical research perspective from linguistics and the application perspective from computational linguistics. As the main target was Japanese discourse, we had to extend the discourse relations to include various aspects of Japanese discourse structure, and found that it is effective to analyze and explain the relationship among the information described in the discourse, especially short one such as tweets.