基于混合统计模型的中文基本名词短语识别

摘要

采用一种基于混合统计模型的方法来实现中文基本名词短语识别.首先简要分析目前的研究现状,明确中文Base NP识别的任务,然后采用以基于转换的标注和条件随机域模型为底层,支持向量机模型为高层的混合统计模型来进行中文BaseNP的识别.在ACE2005中文语料上的实验表明,F值比使用单一模型提高了1.37%,达到了88.67%,能提高中文基本名词短语的识别性能.%This paper proposes a mixed statistical model based method for identifying Chinese base noun phrase (NP). After the brief overview of the current study, we confirmed the mission of Chinese base NP identification, and then adopted mixed statistical model, which consists a base tier of conversion-based tagging and conditional random field model and a senior tier of SVM model, to conduct the identification of Chinese base NP. Experiment on ACE 2005 Chinese corpus shows that the F-measure of the mixed model achieves 88.67% with the improvement of 1.37%. It is capable to ameliorate the identification performance on Chinese base NP.