与词法、句法分析相比，篇章分析研究相对滞后，特别是，汉语篇章分析的研究处于起步阶段，由于尚未形成成熟的理论体系，资源极为匮乏，因此相关计算模型的研究受到了严重的制约。篇章逻辑语义结构和话题结构从不同视角描述了篇章的连贯性，本项目将从汉语篇章逻辑语义结构和话题结构出发，基于已有的篇章分析和话题结构理论，重点研究汉语篇章的连贯性。主要的研究内容包括：1）提出并实现融合多种信息的隐式篇章逻辑关系识别方法，构建高性能的端到端的篇章逻辑语义分析平台，并基于该平台进行基于ILP和结构化感知器的全局优化研究；2）提出并实现基于话题-评述关系理论的微观话题结构分析方法，并在此基础上进行基于话题链的宏观话题识别研究；3）从连贯性视角分析汉语篇章逻辑语义结构和话题结构的关联性，并基于此提出并实现基于句法和谓词论元结构的汉语篇章逻辑语义结构和话题结果的联合学习方法。

The research on discourse analysis lags behind morphological and syntactic analysis. In particular, the research on Chinese discourse analysis just started. The lack of effective theoretical methodologies and corpora severely restricts the research on the computational model of Chinese discourse analysis. Logical semantic structure and topic structure describes discourse coherence from different perspective. This project mainly focuses on Chinese discourse coherence based on Chinese logical semantic structure and topic structure analysis. The project addresses Chinese discourse coherence from following aspects: 1) Propose an algorithm for implicit discourse relation identification combining multi-level knowledge, and implement an effective end-to-end Chinese logical semantic structure parser. Based on this platform, research on global optimization via ILP and structured perception framework. 2) Design and implement a high-performance algorithm for micro-topic structure analysis based on the theory of topic-comments relationship. Then research on macro-topic identification based on the micro-topic chains. 3) Analyze the relationship between logical semantic structure and topic structure from discourse coherence perspective. Propose and implement the joint learning of logical semantic structure and topic structure analysis based on syntactic and predict-argument structure.