基于SVM的中文名词短语指代消解研究

指代消解是自然语言处理领域中要研究的关键问题之一。在自然语言中,为了使语言简明,减少冗余,往往对同一意思的单词、句子或某一事件用不同的单词来代替。相对于人而言,计算机理解这些指代现象就比较困难,因此近年来关于指代消解的研究越来越多。由于中文指代消解研究起步较晚,因此关于中文名词短语指代消解的研究还比较少,大多研究是关于英文指代消解的。给出了一个基于SVM的中文名词短语指代消解平台并详细介绍了整个实现过程,语料库采用OntoNotes 3.0的中文新闻语料。利用3种评测算法对系统性能进行了评测,结果表明本系统是一个比较好的中文指代消解平台。

Research of Chinese Noun Phrase Anaphora Resolution : A SVM-based Approach

Coreference resolution is an important subtask in natural language processing systems. In natural language, to make the natural language clear and explicit illusions, it is common that two or more words, sentences or events which have the same meaning are replaced by different words. In compare to people, it is difficult to understand these phenomenon by using the computer, so more and more researchers focus on noun phrases coreference resolution. A great deal of research has been done on this task in English. In Chinese, because the research about coreference resolution starts late,much less work is done in this area. We presented a Chinese noun phrase coreference resolution system based on a SVM approach and gave the details of the platform in the paper. We adopted three tools to evaluate the performance of the platform. Experiments on the Chinese portion of OntoNotes 3. 0 show that the platform achieves a good performance.