CCCMR 2015 - International Workshop on Content-, Context- and Crowd-based Multimedia Recommendation

Recommender systems are becoming increasingly important due to the overload of information brought by today’s Internet. With the rapid growth of digital devices and media sharing platforms, the amount of multimedia information has explosively increased in recent years. Therefore, multimedia recommender systems are crucial for proactively helping users in their information seeking process within the huge volume of Internet multimedia. In traditional recommender systems, user-item collaborative information (such as Collaborative Filtering and Matrix Factorization based methods) and multimedia contents (such as Content-based Filtering methods) are two major sources to predict the user-item interaction behaviors. However, with the growth of data scale, traditional recommender systems seriously suffer from the sparsity problem of user-item interactions. In addition, it has been demonstrated by the industry (e.g. Netflix, YouTube etc.) that the collaborative information is quite limited in capturing the user behavior patterns and intentions in information seeking.
With the fast development of social networks and online media sharing platforms, user profiles, tagged multimedia contents and the interaction behaviors between users and multimedia contents are digitally recorded in an unprecedented level. This provide us a precious opportunity to deeply investigate the user-information interaction mechanism, and design more accurate recommender systems. Take video recommendation for example. From the video aspect, we have not only video contents, but also user-labeled tags, geographic information, and user comments etc. From the user (crowd) aspect, we also have user profiles, user relationships and user interactions, etc. From the user-information interaction aspect, we can reconstruct fine-granular user behavior logs and also the temporal and spatial context information for these interaction behaviors. The information above gives us a good opportunity to better understand the user-content interaction behavior patterns. Further, it can be regarded as a supplement of collaborative information to solve the sparsity problem. Therefore, developing computational methods to jointly utilize the information in content, context and crowd dimensions for multimedia recommendation is of paramount importance to improve the performance of recommender systems, and provide us new advances including new approaches and directions on multimedia recommendation.
The objective of this special issue is therefore to provide a forum for researchers in multimedia recommendation and multimedia retrieval to review pressing needs, discuss challenging research issues, and showcase the state-of-the-art research and multimedia recommender systems in the modern Internet environment.
Topics of Interests
The topics of interests of this special issue include, but not limited to, the followings:
Data representation for multimedia recommendation
Knowledge discovery for multimedia recommendation
Social-sensed multimedia recommendation
Multimedia-focused recommendation models and applications
User profiling and modeling for multimedia recommendation
Multimedia recommendation for new users and/or contents
Cross-platform multimedia recommendation
Collective user behavior modeling for multimedia recommendation
Crowd sourcing for multimedia recommendation
Personalized multimedia recommendation for large scale data
Context-aware multimedia recommendation
HCI and presence for multimedia recommendation
Mobile Multimedia recommender systems