Combining Free Collaborative and Controlled Annotation in Social Media

Social media are usually made of flows of contents (e.g. texts, images, products), which are annotated by content producers and/or consumers. The resulting annotations form a sort of index that is often used for content access. However, free collaborative annotation is known to produce low-quality indexes. In this paper, we present an original approach to overcome this weakness by combining free annotation based on tags with categorization relying on a controlled vocabulary. We propose tools that support the creation and maintenance over time of controlled indexes, so that these operations are manageable for human annotators. Our work is based on a corpus of blogs, in which posts are associated with both tags and categories, but we assume that the proposed approach can be adapted to other types of social media.