Deep Web入口探测与分类方法研究

传统的使用语料库对入口标签字符串进行匹配的方法受限于语料库的完整性和匹配算法的灵活性。为突破这种局限，引入了基于表单元件统计特征的Deep Web入口探测方法和使用文本分类方法对其进行分类的双层分类模型，并提出了两种特征权重计算方法用于特征选取。在TEL-8 Query Interfaces数据集上，测试结果体现了双层分类模型的优越性和特征向量维归约的必要性。

Research on detecting and classifying Deep Web interfaces

Traditional method using library to match those labels is limited to the integrity of the library and the scalability of the matching algorithm. In order to break through this limitation, this paper introduced a bilateral-layer model based on the statistic characteristics of the interfaces to detect Deep Web entries and text classification approach to classify them. Meanwhile, it provided and applied two methods of computing feature-weight to feature selection. The test results got from TEL-8 Query Interfaces showed the superiority of bilateral-layer classification model and the necessity of dimensionality reduction.