篇章衔接性分析是篇章理解的基础，而指代和省略是衔接机制中的核心问题，同时也是难点所在。受限于语料资源，目前国内外对于汉语篇章衔接性中指代和省略的研究进展缓慢。本项目将结合汉语自身的特点和规律，针对汉语篇章衔接性中的指代和省略问题展开研究。首先建立较大规模的汉语篇章衔接性标注语料库；在此基础上，通过先行事件候选的触发词识别、动态事件指代关系树的获取以及适用于事件指代消解的基于实体指代链的上下文相似度计算方法的研究，提出并实现综合使用多层面信息的事件指代消解方法；对于汉语中的省略，将从句法、语义和语用三个视角识别省略及对其进行消解；最后再利用联合学习机制将指代和省略消歧融合，建立统一的篇章消歧平台。本项目开展的研究工作对于推进汉语篇章衔接性分析的研究，推动计算语言学研究和中文信息技术的发展具有重要的理论意义和应用价值。

Discourse cohesion analysis plays a critical role in discourse understanding. This project addresses two essential issues in discourse cohesion mechanism, anaphora and ellipsis, where there is only a little research due to the lack of annotated corpora. In particular, this project focuses on anaphora and ellipsis resolution in Chinese. First, a large scale corpus will be annotated with various kinds of anaphora and ellipsis information. Second, an event anaphora resolution framework is proposed to combine multi-level information in effectively identifying probable antecedent candidates of an event anaphor, constructing adaptable dynamic event anaphora tree spans for event anaphora resolution, and presenting context similarity algorithms based on entity coreference chains to compute the context similarity between the event anaphor and the antecedent candidate more effectively. Third, various solutions are explored to address ellipsis from the perspectives of syntax, semantics and pragmatics. Finally, a unified platform is proposed for both anaphora and ellipsis resolution via joint learning model. .In summary, this project will promote the research on Chinese discourse cohesion analysis, and are thus valuable for the development of computational linguistics and Chinese information processing.