VISCORS: A Visual-Content Recommender for the Mobile Web
IEEE Intelligent Systems
A regression framework for learning ranking functions using relative relevance judgments
SIGIR '07 Proceedings of the 30th annual international ACM SIGIR conference on Research and development in information retrieval
WI '07 Proceedings of the IEEE/WIC/ACM International Conference on Web Intelligence
Flickr tag recommendation based on collective knowledge
Proceedings of the 17th international conference on World Wide Web
Personalized, interactive tag recommendation for flickr
Proceedings of the 2008 ACM conference on Recommender systems
Ranking and classifying attractiveness of photos in folksonomies
Proceedings of the 18th international conference on World wide web
Stochastic gradient boosted distributed decision trees
Proceedings of the 18th ACM conference on Information and knowledge management
Leveraging user comments for aesthetic aware image search reranking
Proceedings of the 21st international conference on World Wide Web
Multimedia features for click prediction of new ads in display advertising
Proceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining
Image ranking based on user browsing behavior
SIGIR '12 Proceedings of the 35th international ACM SIGIR conference on Research and development in information retrieval
Rare is interesting: connecting spatio-temporal behavior patterns with subjective image appeal
Proceedings of the 2nd ACM international workshop on Geotagging and its applications in multimedia
Hi-index | 0.00 |
This paper focuses on the prediction of users' favourite photos in Flickr. We propose a multi-modal, machine learned approach that combines social, visual and textual signals into a single prediction system. Although each individual user has different motivations for calling a photo a favourite, we show that the textual, visual, and social modalities effectively capture the needs of most active Flickr users. We use gradient-boosted decision trees (GBDT) with a mod least squares loss function for the classification of a user's favourite photos, and evaluate the performance of our classifier with respect to the individual modalities and various combinations thereof. By using a combination of the social and visual modalities the GBDT creates a highly effective classifier. The addition of textual features allows us to significantly increase recall, with a slight trade off in precision.