【作者】 邱立坤； 张晓巧； 毛宁；

【Author】 Qiu Likun~(1,2),Zhang Xiaoqiao~1,Mao Ling~1 (1 Institute of Artificial Intelligence,Beijing City University,Beijing 100083; 2.Department of Chinese Language and Literature,Peking University,Beijing 100871)

【机构】 北京城市学院人工智能研究所； 北京大学中文系；

【摘要】 本文介绍了构建现代汉语复合词内部结构词典的方法和流程。在构建该词典的过程中,我们使用人工分析与自动分析相结合的方法。在自动分析方面,我们共使用了三种方法,包括双向平行类推、成对替换类推两种自动类推方法,以及基于形式与意义同构思想的推导方法。Hownet 54000双字词中,这三种方法可以给出全部或部分内部结构的词语有4万余条。在人工标注和自动标注的基础上,我们通过两种方法来辅助校对。第一种方法是采用启发式规则从人工标注结果中自动发现标注不一致的现象,并从中挑出可能错误的标注结果,收到了良好的效果。第二种方法是将人工标注结果和自动标注结果进行比较,将不一致的提交给校对人员。更多还原

【Abstract】 This paper presents a novel approach of building a contemporary Chinese dictionary on internal structure of compounds.Manual tagging and three automatic tagging methods,including bi-direcuon parallel analogy,paired parallel analogy and inference based on the consistency between form and meaning,are used.More than 40,000 words of all the 54000 bi-syllabic words in Hownet are fully or half semantically tagged by those automatic methods.After manual and automatic tagging,two methods are used for checking.First,several heuristic rules are used on manual tagging results to mine abnormal tagging.Second,the manual and automatic tagging results are comparing together and those inconsistent tagging results are considered as abnormal tagging.All those abnormal tagging results are reserved for further checking.更多还原