Automatic tag generation for photos using contextual information and description logics

We present the design of a system for automatic photo annotation. The system provides a set of tags by combining the context in which the photo was captured with prior knowledge about popular annotation concepts. We exploit semantic web technologies, such as RDF and OWL, to store structured metadata and for representing domain knowledge. The inferencing capabilities of Description Logics allow new knowledge to be deduced resulting in more useful and meaningful tag suggestions. The system is evaluated quantitatively using randomly chosen photos which were manually annotated. Experiments have demonstrated good performance which strengthens our belief that combining several information sources and using inference to make educated guesses has great potential for automatic tag generation.