适应多领域多来源文本汉语依存句法数据标注规范.PDF

第 ** 卷 第 * 期 中文信息学报 Vol. **，No. * 201* 年 * 月 JOURNAL OF CHINESE INFORMATION PROCESSING ***. ，201* 文章编号：1003-0077 （2017）00-0000-00 适应多领域多来源文本的汉语依存句法数据标注规范 郭丽娟 李正华 彭雪 张民 （苏州大学 计算机科学与技术学院，江苏 苏州 215006） 摘要：近十年来，依存句法分析由于具有表示形式简单、灵活，分析效率高等特点，得到了学术界广泛关注。为 了支持汉语依存句法分析研究，国内同行分别标注了几个汉语依存句法树库。然而，目前还没有一个公开、完整、系 统的汉语依存句法数据标注规范，并且已有的树库标注工作对网络文本中的特殊语言现象考虑较少。为此，我们充分 参考了已有的数据标注工作，同时结合实际标注中遇到的问题，制定了一个新的适应多领域多来源文本的汉语依存句 法数据标注规范。我们制定规范的目标是准确刻画各种语言现象的句法结构，同时保证标注一致性。利用此规范，我 们已经标注了约2万句汉语依存句法树库。 关键词：依存句法；标注规范 中图分类号：TP391 文献标识码：A DataAnnotationGuidelineofChineseDependency SyntaxforMulti-domain andMulti-sourceTexts GUO Lijuan,LI Zhenghua,PENG Xue and ZHANG Min (School of Computer Science and Technology, Soochow University, Suzhou, Jiangsu, 215006, China) Abstract: In the past decade, dependency parsing has attracted a lot of attention in the research community, due to its simplicity and flexibility in syntax representation schema, and its high analysis efficiency. In order to support Chinese dependency parsing research, researchers in China have annotated several Chinese dependency treebanks.However, there still lacks a public, integrated and systematic annotation guideline for Chinese dependency treebanking. Moreover, the existing works on Chinese dependency treebanking do not give much consideration to special linguistic phenomena in web texts. Therefore, after making full reference to previous works and carefully studying the problems encountered in real annotation practice, we propose a new annotation guideline for Chinese dependency treebanking, which is suitable for multi-domain and multi-source texts. Our annotation guideline aims to accurately depict the syntactic structures of