Extracting web content for personalized presentation

Printing web pages is usually a thankless task as the result is often a document with many badly-used pages and poor layout. Besides the actual content, superfluous web elements like menus and links are often present and in a printed version they are commonly perceived as an annoyance. Therefore, a solution for obtaining cleaner versions for printing is to detect parts of the page that the reader wants to consume, eliminating unnecessary elements and filtering the "true" content of the web page. In addition, the same solution may be used online to present cleaner versions of web pages, discarding any elements that the user wishes to avoid. In this paper we present a novel approach to implement such filtering. The method is interactive at first: The user samples items that are to be preserved on the page and thereafter everything that is not similar to the samples is removed from the page. This is achieved by comparing the path of all elements on the DOM representation of the page with the path of the elements sampled by the user and preserving only elements that have a path "similar" to the sample. The introduction of a similarity measure adds an important degree of adaptability to the needs of different users and applications. This approach is quite general and may be applied to any XML tree that has labeled nodes. We use HTML as a case study and present a Google Chrome extension that implements the approach as well as a user study comparing our results with commercial results.