基于渡越矩阵的复句关系词自动标注初探

Preliminary Investigation on Tagging Relatives in complex sentences based on the du-yue matrix

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 98.32% and 96.41% in closed test, 85 and 83% in open test.