Presentation by Przemyslaw Kazienko at the 1st World Summit of the Open Knowledge Society, Athens, 24-26 September 2008. Track: Social & Humanistic Computing for the Knowledge Society: Emerging Technologies and Systems for the Society and Humanity.
Two users participate in common activity related to the certain object with the same/ different role a: e.g. two users comment on the same image. The weights of relations depend on intensity, frequency and quantity. Distinction of different layers of the social network: contact lists, tags, groups, favourites, opinions, multi relational social network. Some layers have rather social (contacts, opinion-author, author-opinion); others have more semantic relations (tags, opinion-opinion relation).
The goal of the system is to recommend people to people. First relations are extracted – building the different layers of the social network (distinction between direct relations (contacts) and object-based relations (tags, opinions, favourites, groups). Based on the layers we create weights for the importance of each layer (consisting personal weight = the user’s individual weight of the user for each layer; and a system weight = aggregation over all users). Afterwards, a social filtering is applied: that is rejection based on the user’s contact lists; rejection of users blocked by the user, damp already viewed users. Rotation mechanism for more random results. Finally, the recommendation is presented to the user. Users are then asked to rate the recommendations.