论文部分内容阅读
回指消解一直以来是自然语言处理发展的主要障碍之一。本文旨在探讨运用向心理论消解汉语回指。向心理论是一种语篇理解的计算模型,它可以考察注意状态、指称形式和推理过程的控制之间的相互关系,因其简洁、易处理,一直作为回指消解的主要算法之一。本文首先结合汉语的特点,提出了适合处理汉语语篇的向心理论模型—“汉语向心模型”。考察整个汉语回指系统,发现向心模型只适合消解汉语中的零形回指和第三人称代词回指。通过逼近法,推导出了汉语回指消解算法。最后,构拟了基于“汉语向心模型”的全自动汉语回指消解系统流程。
Anaphora resolution has long been one of the major obstacles to the development of natural language processing. This article aims to explore the use of centripetal theory to digest Chinese anaphora. Centripetal theory is a computational model of discourse comprehension, which can examine the relationship between attention state, reference form and the control of inference process. Because of its simplicity and ease of handling, centripetal theory has been used as one of the main algorithms for anaphora resolution. Based on the characteristics of Chinese, this paper proposes a centripetal theory model suitable for Chinese discourse - “Chinese centripetal model ”. Examining the entire Chinese anaphora system, we find that the centripetal model is only suitable for resolving the zero anaphora and the third-person anaphora in Chinese. By the approach of approximation, the Chinese anaphora resolution algorithm is deduced. Finally, a fully automatic Chinese anaphora resolution system based on “Chinese centripetal model ” is constructed.