摘要： 为有效解决网页信息抽取所需知识的获取问题，提出了一种基于多层模式的网页信息抽取方法（简称ＨＰＩＥ方法）。将网页信息抽取知识分为若干层，由抽象到具体逐层描述信息识别模式知识。ＨＰＩＥ方法能够利用各抽取对象之间存在的相互联系，以及抽取过程与结果所形成的新学习样本，不断完善多层模式的知识内容，并帮助最终从多个信息内容类似但其描述格式各异的ＨＴＭＬ网页中，抽取出所需的多记录信息内容。有关多个（美国大学教员）论文目录网页的抽取实验结果表明，ＨＰＩＥ方法具有较强的网页信息自适应抽取能力。

关键词: 半结构化数据, 信息抽取, 模式识别

Abstract: To tackle the problem of knowledge acquisition in the HTML page information extraction, a new approach based on hierarchical pattern for HTML page information extraction is put forward. The knowledge used in this method (called HPIE) is composed of a few kinds of pattern descriptions, from abstract to concrete for information recognition patterns. Besides the HPIE combines the knowledge about the different objects' relation with information extraction process and its results to help modify its knowledge of information extraction, in order to extract the desired multiple record information from different HTML pages with similar contents successfully. The information extraction experiment made on the American university faculty's paper page shows that the HPIE has a power and adaptive ability to extract the desired information from HTML page having different display format.

中图分类号: 

• TP393.09