基于规则的把字句语义角色标注Rule-basedSemanticRoleLabeling.PDF

基于规则的把字句语义角色标注 何保荣，邱立坤，徐德宽 （鲁东大学文学院，山东 烟台 264025） 摘要：把字句是现代汉语中一种重要的特殊句式，本文尝试用基于知识库的规则方法对把字句进行语义角 色自动标注。首先，我们从 《人民日报》语义角色标注语料库中收集把字句例句，形成一个覆盖范围较广 的把字句例句库。之后，对例句库中把字句的句法和语义构成规律进行手工标注，标注内容包括谓语动词 的配价类型、把字句谓语结构类型、把字句句模类型等。在上述标注的基础上，对把字句的句模构成规律 进行分析，总结出若干条语义角色标注规则。最后，在测试数据上对前述规则进行验证，语义角色标注的 最终正确率为98.61%，这一结果说明本文所提出的规则在把字句语义角色标注上是有效的。 关键词：把字句；语义角色标注；句模 中图分类号：TP391 文献标识码：A Rule-based Semantic Role Labeling of Ba-sentences HE Baorong, QIU Likun , XU Dekuan (School of Chinese Language and Literature, Ludong University , Yantai, Shandong 264025, China) Abstract: Ba-sentence is one of the most important Chinese special sentence patterns. This paper proposed a rule-based method to address the task of automatic semantic role labeling, especially focusing on ba-sentences. Firstly, we collected a set of ba-sentences from our annotated semantic corpus, including texts from People s Daily’ , and thus formed a sample gallery of ba-sentences. Then, we tagged the valence type of each predicate, the syntactic structure type and semantic structure type of each ba-sentence manually. Based on this annotated corpus, we analyzed the rules of semantic formation, and summed up several rules of semantic role labeling. Finally, we evaluated these rules in a test set. The overall precision is 98.61%, which shows that the proposed method is effective in labeling the semantic roles of ba-sentences. Key words: ba-sentence; semantic role labeling; semantic sentence pattern 1 引言 语义角色标注是一种浅层语义标注，其主要内容是识别谓词的论元，并为每个论元标注 [1] 一个语义角色 。现有研究一般将语义角色标注视为分类问题或者序列标注问题，使用最大 [2] [3] 熵模型 、条件随机场模型 等予以实现。在训练数据较为充足的情况下，已取得较高精度。 但现有自动标注方法主要使用句法信息和词汇信息，较少考虑谓词的格框架以及语义角色与 句式 （句子的句法结构类型）之间的配合关系。 语言学中与语义角色标注相关的研究包括句模研究