High Impact Factor : 4.396 icon | Submit Manuscript Online icon |

Tag Recommendation System for Photos on Social Websites Based on User Preferences and Geo-Location Preferences


Priyanka S K , RNSIT, Bangalore; Kusuma S, RNSIT, Bangalore


Geo-location preference, subspace leaning, unified space, user preference


Photo tagging is becoming more and more important now-a-days to organize and search large number of photos on social websites. To generate high quality social tags and automatic tag recommendation is the main research topic. In this paper main focus is on the personalized and geo-specific tag recommendation. Consider users and geo-locations have different preferred tags assigned to a photo, a new subspace learning method is proposed to individually discover both the preferences. The goal is learn unified space which is shared by visual domain and textual domain to make visual features and textual features comparable. Visual feature is considered to be lower level representation on semantics than textual feature. Additionally intermediate space is introduced for the visual space and expecting it to have consistent local structure with text space. Unified space is mapped from the textual space and the intermediate space respectively. When an untagged photo with its geo-location is given based on the nearest neighboring search user preferred and geo-location-specific tags are found in the corresponding unified space. Then combine these obtained tags and the visual appearance of the photo to find semantically and visually similar photos, among which the most frequently used tags are suggested to the user and user is allowed to select based on his preference.

Other Details

Paper ID: IJSRDV3I31099
Published in: Volume : 3, Issue : 3
Publication Date: 01/06/2015
Page(s): 1056-1059

Article Preview

Download Article