Personalized Geo-Specific Tag Recommendation for photos on social websites

Nowadays social tagging is an important in social websites to provide a good tagging for photos uploaded on the websites to access high quality social tags. Tag recommendation by automatically assigning related tags to photos to find out particular interesting area. In this paper we concentrate on the personalized recommendation work and try to choose user preferred geo-location-specific as well as relevant tags for photos on social website. For users and geo-locations we assume they have various preferred tags assigned to a photo and purpose, a subspace learning method to individually uncover the user preference and geo-location preference. The goal of our work is to combine a visual and textual space into a unified subspace. According to unified subspace, it is mapped from the intermediate subspace and textual subspace respectively. We create formula for above learning problems into united form and present the minimization with its convergence rule. For a given untagged photo with its geo-location to a user we used the nearest neighbor search in the relating unified space. The user preferred and geo-location specified tags. Experiments on big scaled data sets collected from flicker examine the effectiveness of the proposed system.