CUbRIK is a 36 month Collaborative Project, partially funded by the European Commission's 7th Framework, that aims to introduce real innovative patterns inside the Multimedia search domain, proposing the paradigm of human-enhanced time-aware multimedia search, driven by openness at all levels.
Multimedia search engines today are ‘black-box’ systems, powerful engines but complex inside. As a consequence, besides being complex in design and modelling, results from search-driven applications, especially in multimedia, are often far from being relevant.
The multimedia search domain will be scientifically and technically advanced by interdisciplinary sectors such as architectures for content-intensive and human-intensive scalable processes, multimedia and multimodal search techniques, social and human computation.
A unified approach will include both machine tasks and a variety of human, individual, social and crowd-sourced tasks, orchestrated in CUbRIK by programmable pipelines for all the three coordinated processes typical of multimedia search: content, query and relevance feedback processing.
CUbRIK open composability will also encompass entity-based time and space-aware search functions, based on highly accurate repositories of spatio-temporal entities (locations, events, trends) correlated with rich semantic associations. Workflows will be open to incorporate human and social processing by coupling machine-based computation abilities with human abilities so as to augment the precision of results. The semantic gap will be reduced about content and relevance of search results: i.e. by disambiguating low-level features (colour, texture, shape,..) and high-level features (meaning, purpose,..) of depicted objects or scenes; by considering human judgements, direct opinions and the opinions of people socially sharing same common values. Innovators will be offered an open stage to experiment the CUbRIK ‘white-box’ ability.