利用标点符号自动识别分句

On Making Use of Punctuation Marks to Identify Clauses Automatically

为了建设一个面向中文信息处理的大规模复句"精加工"语料库,首先要完成自动分词和词性标注工作,在此基础上进行分句层次和关系的自动划分和标注。由于标点符号是最直观而明晰的断句标记,我们编制的初始程序让计算机一"碰到"标点符号就把它前后的语言片段都判定为分句。这就为今后基于句法和语义知识的非分句识别奠定了良好的基础。

In order to build an ＂advanced＂ compound-sentence corpus for Chinese Information Process,automatic word segmentation and POS tagging work should be completed first of all.Then on this basis,automatic classification and labeling of levels and relationship between clauses should be conducted.As punctuation marks are the most intuitive and clear marks,we programmed the computer to regard the language fragments between punctuation as clauses.Doing so much is risking,because it will ＂victimize＂ a lot of non-clause language fragments which are subject to be identified through other means.