网页文本中部分关系知识的获取与验证

Acquisition and Verification of Mereological Knowledge from Web Page Texts

重点研究了如何从大规模W eb网页文本中获取一种特殊的知识——部分关系知识(M ereo log ical know ledge)。介绍了部分关系知识的6种不同形式,在此基础上自动和半自动地获取这些部分关系知识的文法模式,建立文法模式库;然后基于文法模式获取例句来抽取准概念并进行局部验证;利用基于图论的方法构造部分整体图,将所有准概念从全局的角度进行分析验证。实验结果证明了该方法的可行性。

Knowledge acquisition from text （KAT） is one of the important research problems in knowledge engineering. How to acquire a special kind of knowledge-mereological knowledge from largescale Web page text is the research emphasis in this paper. Firstly, six different types of mereological knowledge are introduced, on the basis of that, grammatical patterns could be acquired automatically or semi-automatically to construct the extraction pattern base. Then quasi-concepts would be extracted and verified locally with those result files generated by patterns. At last, mereological graph is defined to verify all the quasi-concepts globally based on graph theory, which has been proven to work well by experi- ments.