An automatic approach to extracting review link from Chinese news pages

Review links are widely used in some special kinds of web pages, especially news pages. They are very useful pieces of information in many applications, such as hot topic discovery and public opinion monitoring. Unfortunately, extracting review links manually from news pages is time-consuming and error-prone. Though lots of works on web data extraction have been developed, we argue that this is still not a trivial problem due to the diversity on both DOM tree structure and visual presentation. In this paper, a novel approach is proposed for automatically extracting the review links from web pages. This approach consists of two steps: first segment each news page into a set of blocks, and then identify the block(s) that contain the review link using a machine learning technique. Experimental results over a large number of Chinese news pages indicate that this approach is highly accurate.