英汉《小王子》抽象语义图结构的对比分析

发布时间：2018-04-18 22:03

  本文选题：抽象语义表示 + 语义图　； 参考：《中文信息学报》2017年01期

【摘要】：AMR(抽象语义表示)是国际上一种新的句子语义表示方法,有着接近于中间语言的表示能力,其研发者已经建立了英文《小王子》等AMR语料库。AMR与以往的句法语义表示方法的最大不同在于两个方面,首先采用图结构来表示句子的语义;其次允许添加原句之外的概念节点来表示隐含的语义。该文针对汉语特点,在制定中文AMR标注规范的基础上,标注完成了中文版《小王子》的AMR语料库,标注一致性的Smatch值为0.83。统计结果显示,英汉双语含图结构句子具有很高的相关性,且含有图的句子比例高达40%左右,额外添加的概念节点则存在较大差异。最后讨论了AMR在汉语句子语义表示以及跨语言对比方面的优势。
[Abstract]:AMRs (abstract semantic representation) is a new method of sentence semantic representation in the world.The biggest difference between the AMR corpus. AMR and other syntactic and semantic representation methods has been established by the authors. Firstly, the graph structure is used to represent the semantics of sentences.Secondly, it is allowed to add concept nodes other than the original sentence to express the implied semantics.In this paper, according to the characteristics of Chinese, based on the specification of Chinese AMR annotation, the AMR corpus of the Chinese version of Little Prince is completed, and the Smatch value of consistency is 0.83.The statistical results show that there is a high correlation between English and Chinese sentences with picture structure, and the proportion of sentences with pictures is about 40%, and there is a great difference between the additional concept nodes.Finally, the advantages of AMR in Chinese sentence semantic representation and cross-language comparison are discussed.
【作者单位】： 南京师范大学文学院;南京师范大学计算机科学与技术学院;布兰迪斯大学计算机系;
【基金】：江苏高校哲学社会科学研究项目(2016SJB740004) 国家科技支撑计划课题(2014BAK04B02) 国家自然科学基金(61272221)
【分类号】：H315.9;TP391.1
，

本文编号：1770259