汉语篇章结构分析的资源建设与计算模型研究

项目名称： 汉语篇章结构分析的资源建设与计算模型研究

项目编号： No.61273320

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 周国栋

作者单位： 苏州大学

项目金额： 81万元

中文摘要： 篇章结构分析针对篇章的连贯性,是篇章级语义分析的基础。目前自然语言处理研究的重点聚焦在词法和句法层面，对篇章内在规律的研究相对较少，缺乏对篇章进行有效分析的理论和计算方法，从而严重制约了基于篇章的相关应用。本项目将从理论体系完善、语料库构建和计算模型研究等方面系统开展汉语篇章结构分析的研究：1）从研究汉语篇章结构组织的规律和特点入手，建立一套适用于汉语篇章结构表示与分析的理论体系。拟采用一种连接依存树的形式表示汉语的篇章结构，并基于连接词构建汉语的篇章关系体系。2）提出一套汉语篇章结构标注规范，并结合自举学习和主动学习方法半自动构建一定规模的高质量汉语篇章结构标注资源。3）以子句为基本单位，自底向上，借鉴依存句法分析算法，分析篇章中的层次结构及各组成成分之间的结构关系。有指导性和无指导性方法双管齐下，重点解决隐式（连接词缺省）篇章结构关系的识别。

中文关键词： 篇章级语义分析；篇章级结构分析；篇章连接依存树；汉语篇章结构表示机制；汉语篇章结构分析器

英文摘要： Discourse structure analysis deals with discourse coherence and is fundamental to discourse-level semantic analysis. Currently, most of research in NLP focuses on morphological and syntactic levels and there is much less research on the inherent regulations in discourse. This results in the lack of theoretical and computational methodologes towards effective discourse analysis and serevely restricts its wide applications. This project addresses Chinese discourse structure analysis from following aspects: 1)Studying the general regulations and special characteristics in the structure organization of Chinese discourses and building a theretical framework suitable to the representation and analysis of Chinese discourse structures. Specially, a connective-driven depepdency tree structure is proposed to represent Chinese discourse structures and a Chinese discourse relationship framework will be constructed on connectives. 2)Proposing a annotation scheme for Chinese discourse structures and semi-automatically building a certain-scale high-quality annotated Chinese discourse structure corpus by integrating bootstrapping and co-training. 3)Proposing a bottom-up approach to parse the hierarchical structure in a Chinese discourse and the structure relationship between its compomnents. Specially, both supervised and u

英文关键词： discourse-level semantic analysis；discourse-level structural analysis；discourse-level connective-driven dependency tree；Chinese discourse structure representation scheme；Chinese discourse structure analyzer