基于词库与词法的分词不一致研究  

On the Disagreement of Participles Based on Lexicon and Morphology

作　　者：董宇[1] 陈小荷[2] 

机构地区：[1]金陵科技学院龙蟠学院,江苏南京211169 [2]南京师范大学文学院,江苏南京210097

出　　处：《浙江教育学院学报》2008年第3期96-102,共7页Journal of ZHEJIANG Education Institute

摘　　要：分词不一致问题一直严重影响带标注语料库的标注质量,利用词库与词法知识,可以合理地阐释分词不一致产生的根源,并结合建立的规则库、组合型歧义库、固定词表和特殊单字词表,可以解决汉语分词语料库中相同结构类型的分词不一致问题。计算机利用这些知识较好地识别出了"大+动词(单字)""动补结构"和"颜色词+物体名",召回率在96%以上,精确率在95%以上,并能根据用户的要求统一处理成"分"或"合"的形式。The problem of participles disagreement has been affecting the marking quality of labelled corpus. Utilizing the knowledge of morphology and lexicon, the source of participles disagreement can be explained rationally and with the set-up of regular lexicon and equivocal words, list of fixed words and list of individual words, the disagreement of participles, which are of the same structural type in the lexicon of Chinese participles, can be solved. By using the knowledge, computers can recognize ＂大 ＋ individual words＂, ＂verb ＋ complement structure＂ and ＂colour words ＋ names of things＂, and the calling-in rate is above 96 % and the precision rate is above 95 %. In addition, computers can also make them into forms of ＂separation＂ or ＂combination＂ according to consumers＇ demands.

关 键 词：分词不一致 词库 词法 自动分词 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]