Abstract: Bag of words method used in Deep Web sources classification shows many limitations. This paper proposes a novel Deep Web sources enhancing classification model based on world knowledge. It sets up the feature mappings by topic analysis of external knowledge, constructs an auxiliary classifier based on domain concepts, and enriches feature set of Deep Web forms. Experiment is performed based on Wikipedia encyclopedia, and experimental results verify this method is effective and scalable.

Key words: Deep Web, data sources classification, topic analysis, feature mapping, world knowledge

摘要： 针对传统词袋方法在深网(Deep Web)数据源分类应用中的局限性，提出一种基于世界知识的Deep Web数据源增强分类模型，通过对外部知识库的主题分析，建立特征映射，构造基于领域概念的辅助分类器，丰富Deep Web查询表单的特征集合。基于Wikipedia百科知识库对真实Web数据进行分类。实验结果证明该模型有效。

关键词: 深网, 数据源分类, 主题分析, 特征映射, 世界知识

CLC Number: 

• TP393