Automated information extraction from web APIs documentation
WISE'12 Proceedings of the 13th international conference on Web Information Systems Engineering
Hi-index | 0.00 |
The estimation of the block importance could be defined as a learning problem. First, a vision-based page segmentation algorithm is used to partition a Web page into semantic blocks. Then spatial features and content features are used to represent each block. Considering the difference of Web pages, an entropy-based method is adopted to analyze the individual contribution of each feature to the overall effectiveness in the given page. Thus, the entropy value of each feature is used to obtain feature's weight utilized in the further scoring algorithm. Experiments compare the influence both by adaptive weight and by constant weight. The result indicates that the BlockEvaluator algorithm could highly enhance the flexibility in the learning of block importance. The approach is tested with several important Web sites and achieves precise results, correctly extracting 96.2% of news in a set of 2430 pages distributed among 10 different sites.