基于规则的中文框架元素自动标注方法研究

摘要

计算机自动语义分析是当前制约自然语言信息检索、信息抽取与机器翻译等应用技术发展的一个瓶颈问题.语义角色标注是语义分析的一种主要实现方式,而目前语义角色的自动标注主要采用基于统计的方法,由于训练数据的规模有限、语义角色类型多,面临严重的数据稀疏问题,处理结果一致性差.本文则采取基于规则的方法,选择汉语框架语义知识库(CFN)所提供的框架和框架元素作为语义标注体系,利用CFN的语义标注句子库,根据短语类型、句法功能以及短语内部构成和外部语境等其他句法语义特征分析框架元素的实现规律,构建语义标注规则,经测试取得了令人满意的结果,为解决语义分析问题探索了一条可行的路线.%Automatic semantic parsing is a bottleneck for the progress of such natural language applications as information searching, information retrieval and machine translation and so on. Semantic roles labeling is one of the main implementations for semantic parsing. However, the present researches on semantic roles labeling are mostly based on statistical methods, facing a serious problem of data sparse with poor consistency, because of the limit of training data' s size and the multiplicity of semantic roles. So this paper chooses frames and frame elements from the Chinese FrameNet (CFN) as the tag system for semantic annotation, analyses frame elements' realization patterns involving phrase types, grammatical functions and other grammatical and semantic properties such as phrase structure and context and so on, generalizes semantic annotation rules and tests them. The results are satisfatctory and thus a feasible method for semantic analysis is gotten.