english summary

GlobalMusic2one is developing a new generation of hybrid search engines including new methods of music information retrieval (MIR) and Web 2.0 technologies. It aims at reaching a better quality in the automated recommendation and online marketing of global music collections. Music recordings will be automatically analysed by a self-learning software, which incorporates rhythm, melody and other characteristics. This allows an efficient and exact filing of new content into existing collections. The user may create new categories to allow the system to flexibly adapt to new musical forms of expression and regional contexts. These categories can, for example, be regional sub-genres which are defined through exemplary songs or song snippets.

A prototypical work flow consisting of semantic indexing, user interaction, model adaptation as well as search and recommendation functions will be implemented. It will be based on the research of new core technologies and evaluated with global music content by user groups.

Target:

The main idea of the project is to use folksonomies as training basis for dynamical adaptation and to refine the content-based models for music indexing. Universal models and core technologies of content-based music retrieval will provide a flexible and modular tool box. The self- learning MIR framework will then be continuously expanded with precise content-based descriptors. These are defined by specifications and exemplary training data from user groups. This framework will provide main test and research functionalities for future global artist marketing portals. Combining the knowledge of music content of artists and labels from around the world with scientifical concepts and hybrid metadata will enhance the prospects of global marketing.