The trouble with searching through multimedia
Mind-boggling quantities of multimedia content are generated every day. For example, some 60 hours of video are uploaded every minute on YouTube alone. However, the contemporary many-to-many paradigm of information flow presents a number of significant challenges, particularly when it comes to searching for and finding relevant information. This is especially tricky for multimedia content, since searches are mostly conducted using text, not images or sounds. With EU backing, the 'Multimedia information extraction from social networks' (MIESON) project aimed to bolster the effectiveness of tools designed to access, discover or create multimedia content in real-world applications. The project focused on leveraging cutting-edge technology to create a novel user-centric approach to information retrieval. Project members developed methods for learning to develop statistical models from the combination of content and context information. This has resulted in three new methods, including an approach that leverages unstructured contextual information from user-generated content. To date, two scientific articles and a conference paper on the project have been published, and MIESON has attended several talks and various scientific research events. Once incorporated into applications, the findings of the project should help make the searching of multimedia content that bit more intuitive.
