实现了一个基于最大熵的中文指代消解系统。通过预处理获得相关信息,抽取出12特征,采用最大熵算法训练生成分类器。在ACE05 bnews中文测试语料上的指代消解实验结果表明,本系统是一个中文指代消解研究的较好平台。
This paper realized a Chinese coreference resolution system based on maximum entropy model. Using the information obtained from a pipeline of natural language processing(NLP) modules, this paper extracted 12 features from raw texts, and trained a maximum-entropy-based classifier. Then tested it on ACE05 bnews corpus. The results show that this system is a good system for Chinese coreference resolution research.