Utilizing Social Breadcrumbs for User Profiling in Personalized Applications

The rapid growth of World Wide Web and huge amount of information on it gave birth to “information overload”; an effective way to deal with this problem is through personalized applications such as personalized Web search engines and recommendation systems. An important aspect of any personalization effort however is a model that can serve as an accurate representation for the user. This model plays a central role as it serves as the engine based on which personalized content will be served to a specific user. Given the widespread activities of users on social networking platforms, these platforms can provide useful indications relating to a user’s interests and preferences. Traditional sources for user modelling are controversial in nature in line with the loss of privacy associated with them. More recent sources such as bookmarks and social tags have proven to be less effective; we therefore propose microblog acitivity of users as an alternate source to the aim of creating effective user profiles. Moreover, microblog activity of those to whom a user is connected is utilized in our model. Experimental evaluations in two application scenarios namely personalized Web search and scientific articles’ recommendation demonstrates the superiority of our model.