Designing Experiments to Compare Web Page Segmenters

Web pages are mainly designed to be visualized, and users with no visual limitations can easily recognize the role of the parts and functional elements. On the other hand, computer programs can hardly identify the semantic of blocks and, therefore, tasks such as indexing the pages have, necessarily, limitations. Tools that can automatically identify the role can be useful to accomplish such a task. Web Page Segmenters are tools that, using visual and other clues in the page, can extract the functions of blocks. We suggest that a study to compare segmenters could be of interest to the community. In this paper, we highlight the research questions that the study should answer, suggest a possible methodology, and indicate the steps to create it.