基于平行论元的隐式篇章关系推理研究

发布时间：2018-03-17 01:26

本文选题：隐式篇章关系　切入点：平行论元推理　出处：《计算机应用与软件》2012年09期 　论文类型：期刊论文

【摘要】：基于Penn Discourse TreeBank(简称PDTB)语料中的隐式篇章关系类型,提出一种无指导的识别方法。主要依据显式与隐式平行论元对之间的映射关系实现显式到隐式关系的推理,即利用显式论元对的篇章关系,推理与之平行的隐式论元对的篇章关系。推理过程主要包括:基于大规模语料库以及基于搜索引擎的平行语料挖掘,平行语料中显式连接词映射到篇章关系过程的消歧。与传统基于监督学习的分类方法相比,基于统计的无指导方法在隐式篇章关系推理的性能上获得显著提升,识别精确率提高了近15.6%。此外,在分析比较各研究方法的同时,指出目前隐式篇章关系推理研究所面临的主要困难与挑战。
[Abstract]:Based on the type of implicit text relation in Penn Discourse Tree Bank corpus, an undirected recognition method is proposed, which is based on the mapping relationship between explicit and implicit parallel argument pairs to realize the reasoning from explicit to implicit. That is to say, using the text relation of explicit argument pair and the text relation of implicit argument pair which is parallel to the explicit argument pair, the reasoning process mainly includes: based on large-scale corpus and parallel corpus mining based on search engine. In parallel corpus, explicit conjunction mapping to disambiguation in the process of text relationship is performed. Compared with the traditional classification method based on supervised learning, the performance of implicit text relation reasoning is improved significantly by the undirected statistical method. The accuracy rate of recognition is increased by 15.60.In addition, the main difficulties and challenges in the research of implicit textual relationship reasoning are pointed out while analyzing and comparing various research methods.
【作者单位】：苏州大学计算机与科学技术学院;
【基金】：国家自然科学基金项目(60970056,60970057,61003152) 江苏省苏州市自然科学基金项目(SYG201030)
【分类号】：TP391.1

【相似文献】