期刊论文

中文

微计算机信息

1008-0570

2007

23

30

200-202

10.3969/j.issn.1008-0570.2007.30.077

本课题得到国家重点实验室开放研究基金（SKLSE04-018） 湖北省科技公关项目（2005AAl01C43）

本校为第一机构

复句关系词的自动标注是自然语言理解领域的基础性研究课题，是层次关系标注和机器翻译等问题的研究基础。本文采用概率统计方法，建立相应的渡越矩阵研究复句关系词的自动标注方法，并进行了有效标注。标注后。对复句关系词中容易产生歧义的“结果”和“如”进行了封闭性测试和开放性测试，其准确率分别达到98．32％和96．41％，85％和83％。

Automatically tagging relatives in complete sentences is a basal researehful problem in the nature language procession field. It is the base of tagging level and machine translation etc. The du-yue matrix is established to research the method of automatical- ly tagging relatives in complete sentences effectively in this paper. The relatives “jieguo” and “ru” easily produce problems in tag- ging relatives. They are chosen in our tagging experiments in which the method gets precisions of 98...