A Quantitative Comparison of Semantic Web Page Segmentation Approaches

Abstract

We compare three known semantic web page segmentation algorithms, each serving as an example of a particular approach to the problem, and one self-developed algorithm, WebTerrain, that combines two of the approaches. We compare the performance of the four algorithms for a large benchmark of modern websites we have constructed, examining each algorithm for a total of eight configurations. We found that all algorithms performed better on random pages on average than on popular pages, and results are better when running the algorithms on the HTML obtained from the DOM rather than on the plain HTML. Overall there is much room for improvement as we find the best average F-score to be 0.49, indicating that for modern websites currently available algorithms are not yet of practical use.