Matching Ads in a Collaborative Advertising System

Classical contextual advertising systems suggest suitable ads to a given webpage, without relying on further information - i.e. just analyzing its content. Although we agree that the target webpage is important for selecting ads, in this paper we concentrate on the importance of taking into account also information extracted from the webpages that link the target webpage (inlinks). According to this insight, contextual advertising can be viewed as a collaborative filtering process, in which selecting a suitable ad corresponds to estimate to which extent the ad matches the characteristics of the "current user" (the webpage), together with the characteristics of similar users (the inlinks). We claim that, in so doing, the envisioned collaborative approach is able to improve classical contextual advertising. Experiments have been performed comparing a collaborative system implemented in accordance with the proposed approach against (i) a classical content-based system and (ii) a system that relies only on the content of similar pages (disregarding the target webpage). Experimental results confirm the validity of the approach