Recomendação de tags para mídia social colaborativa: da generalização à personalização

Abstract

Social media systems such as Flickr, Youtube and Picasa have become very popular as they provide a collaborative environment to share photos and videos supporting tags, ratings and comments. This kind of interaction includes a lot of content provided by users, which may bring meaningful information to recommendation systems. The aggregation of tags is also a way to cluster items and provide tag-based search content. We investigate how to support tag recommendation by ranking the co-occurrence, popularity and relevance of commonly-used tags in similar items and by similar users. We developed a tag recommendation system to recommend of possibly relevant tags. We use Flickr s dataset to analyze our algorithm s behavior and present the results provide by the experiment. A new model using personalized recommendation was developed using the experiment results and the behavior of each user.