I'm looking for an algorithm, that classifies a webpage. But not the content. I know the readability algorithm (by arc90) to extract a main text from an webpage. That for instance can decide, wether a webpage contains an article (is a news or blog article). Thus it can decide wether a webpage is an article page or not.
But I want to go further and classify, wether a non-article-webpage is a product page or a landingpage promoting a product or (!) wether a page has some purpose to motivate the visitor to take some action.
The third case has to be anything else. Whatever: It could be an overview site or else.
Does anyone know such a classifier or has any idea how to analyse the DOM-tree of a webpage in terms of similarity of striking container constellations?