A Block Segmentation Based Approach for Web Information Extraction

This paper addresses the issue of web information extraction to support automatic teacher information management. We propose an effective approach based on block segmentation. First, the teacher introduction web pages are divided into independent blocks, where html tags and punctuation marks are used as segmentation criterion. Then CRF model is employed to label the text. We apply this approach on a teacher web page dataset collected from heterogeneous sources. Experimental results indicate that for basic info and contact info extraction our approach achieves an accurate result just using word level features. As extending value features related to education to block level, the performance of our system on the complex educational information extraction task is dramatically improved.