汉语块扩句的自动分析

Automatic Analysis of Chunk-extension Sentences

块扩句是一类概念预期知识十分明确的句子。基于概念层次网络理论介绍了块扩句对应的块扩句类,总结了能够激活块扩句类的典型块扩动词。根据块扩动词的概念知识得到句子的特征语义块及块扩句类,依据块扩句类的知识对句子进行检验后可给出句子句类的分析结果。在已有的句类分析系统的基础上对真实语料中的块扩句进行了自动分析,实验表明正确率达到了71.29%,错误主要来自特征语义块动词辨识、动词多句类代码等。正确分析块扩句将有助于解决汉语句子的多动词处理难点。

The chunk-extension sentence is a kind of sentence, which is defined with specific conceptual knowledge in the theory of hierarchical network of concepts. This paper introduced some semantic categories of chunk-extension sentences and typical head verbs in them. Through these typical verbs, the eigen semantic chunk and semantic category of a chunk-extension sentence can be activated. After testing with the conceptual knowledge of chunk-extension semantic categories, the semantic structure of a sentence can be revealed. An experiment to automatically analyze some chunk-extension sentences from corpus was done based on a system which is designed for analyzing semantic categories of sentences. The precision of experimental result is 71.28%. The errors are mainly caused by recognizing head verbs and multi-semantic-category of a verb. To analyze chunk-extension sentences well will be help to tackle the puzzle of multiverb in processing Chinese sentences.