基于规则的非典型有标被动句的语义角色自动标注研究  

Rule-Based Automatic Labeling of Semantic Roles of Atypical Marked Passive Sentences

作　　者：汪梦翔[1] WANG Mengxiang

机构地区：[1]北京联合大学师范学院,北京100010

出　　处：《语言文字应用》2022年第2期122-132,共11页Applied Linguistics

基　　金：北京市社科基金2020年度青年项目“基于人工智能的汉语动名搭配内在语义组配机制研究”(20YYC021)资助。

摘　　要：采用被动标记特征辅助语义角色自动标注是语义关系标注的一种重要思路,但被动标记处理还面临诸多难题,最突出的就是非典型被动标记难以准确识别。本文以介词“让”为例,从语言理论和语言工程角度探讨非典型被动标记的自动识别及提取方法,结合有标被动句句法语义分布特点,总结有标被动句的句法构成规律,进而生成有标被动句的语义角色标注规则。这些规则能在自建被动句测试集的语义角色自动标注过程中发挥重要作用。实验结果表明被动标记的准确识别有助于语义角色的准确标注,且本文提出的语义角色标注规则在有标被动句的语义角色自动标注过程中是有效的。Using passive marker features to assist automatic labeling of semantic roles is an important approach of semantic relationship labeling.But passive marker processing still faces many problems,the most prominent of which is that atypical passive markers are difficult to identify accurately.Taking the preposition“Rang”as an example,this paper discusses automatic recognition and extraction methods of atypical passive markers from the perspective of language theory and language engineering.Combined with the syntactic and semantic characteristics of marked passive sentences,this paper summarizes the syntactic rules of marked passive sentences,and then generates the rules of semantic role labeling for marked passive sentences.These rules play an important role in the automatic labeling of semantic roles in the self-built passive sentence test set.The experimental results show that accurate recognition of passive markers is conducive to the accurate tagging of semantic roles.The rules of semantic role tagging proposed in this paper are effective in automatic tagging of semantic roles of marked passive sentences.

关 键 词：被动句 中文信息处理 语义角色标注 词性标注 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]