Generated in the context of Web2.0, social media is a new way of multimedia information exchange and sharing. Its rapid development has attracted attentions from more users than ever before all over the world. The concept of social media includes text information like blog, social networking sites, microblog and wikis; multimedia information like image, video and music sharing websites; and services on mobile device like location-based social network, social game, etc. With the extensive infiltration of various social medias, massive rich media information has been generated. How to effectively and efficiently conduct data mining for accurate search and recommendation, becomes the key problem to the development of modern Internet. Social media is very different from traditional multimedia on the web. Traditional web multimedia search and recommendation technologies focus on directly analyzing multimedia content. However, social media has local symbiotic and collective characteristics. There exist complicated information associations between users and multimedia content. A variety of media collaborations and user interactions, in turn, provide broader semantic for the content. Collective search and recommendation, which exploits the interrelation between users and medias to improve social media understanding, and then accurately and efficiently push desired information to users, serves as one of the most popular research fields in social media. Generally, it includes the following three key issues: · Multimedia content analysis in social media. Multimedia information is the core component of social media, and multimedia content analysis serves as the basis of collective search and recommendation. Traditional image and video analysis methods utilize low-level feature descriptors like color, edge, texture, which has inevitable “semantic gap” with the high-level semantics of human understanding. In the context of social media, interactions and interrelations between users and multimedia content provide a feasible solution to this problem. It is of great significance to leverage the context, annotation, reviews metadata and cross-media associations for better multimedia content analysis. · User understanding in social media. To truly bridge the gap between data and user requirements in collective search and recommendation, besides multimedia content analysis, another key issue is user interest and intent understanding. Learning user interests and prefer...
修改评论