基于组块分析的路径自然语言语义角色标注方法

Semantic role labeling method for route natural language based on chunk parsing

为了进行路径自然语言研究,首先设计了路径自然语言收集方案,主要面向室内环境收集了一定的路径自然语言语料,通过手工标注建立了一个小型的语料库.其次,对路径自然语言的特征进行了一定的分析,在此基础上概括了几种常见的语义角色,主要包括关于landmark和方位转换的语义角色.路径自然语言本身具有高度规律性,语义角色相对集中,因此语义分析采用了基于组块分析的语义角色标注方法,将语义分析问题转化为序列信号的切分和分类问题.最后,利用支持向量机进行了相关的组块标注实验.实验结果表明提出的方法具有较大的潜力.

To study route natural language, first, a plan of collecting the corpus is made. Route nat ural language corpus about the indoor envirenment is collected and a manual corpus is built. Second ly, the features of the route natural language are studied and some common semantic roles about the landmark and spatial transformation are generalized. The rule of route natural language is so obvious that semantic roles are centralized. So the chunk parsing can be used to analyze the semantic of cor pus. Then, semantic analysis is changed into a sequence labeling problem. Finally, the support vec tor machine is used as a tool to tag the semantic role. The results of the experiment show the huge potential of the method.