Sonic Annotator Web Application (SAWA) is a simple framework for writing Web-based applications involving audio analysis. Currently there are two applications: SAWA-Feature Extractor for analysing uploaded audio files using Vamp plugins, and SAWA-Recommender which searches our content based audio similarity database and generates playlists based on one or more uploaded seed tracks.
Our current database contains information of about 150000 audio tracks collected by SoundBite. SAWA is built on Sonic Annotator, a command line Vamp plugin host, various Semantic Web tools developed in the OMRAS2 project (for instance Mopy) and Web Ontologies such as the Music Ontology and the Audio Features Ontology. SAWA uses RDF and other Semantic Web technologies to exchange information between its components. This is summarised by the diagram (please ignore the reference to possible future work such as code sharing via this system which might be implemented using VamPy).
about this releaseThis release focuses on the usability through an enhanced interface for configuration and transform selection. Several VAMP transforms (VAMP plugin output configurations) can be created and selected for each query. All selected transforms will be executed using all uploaded (and selected) audio files. The results can be viewed in place, downloaded individually, or downloaded in batch as a zip compressed file. The next release will focus on machine access and provide a SPARQL interface (currently in testing stage).
browser compatibilityThis is a public development version (beta) of SAWA. As such, not all features work flawlessly in all configurations. In particular multi-platform cross-browser compatibility is not yet achieved. However, the application is expected to work well using Firefox 3.6 on all platforms, we also tested Chrome 6 (Windows, OS/X), Safari 3.1.1 (OS/X), Internet Explorer 8 (Windows, Note: Limited support only. Older versions of IE will not work properly. IE6 is no longer supported.) Opera 10 (OS/X, Note: Some display flaws may be experienced).
SAWA has been made possible through the work of many past and present members of the Centre for Digital Music (C4DM) who contributed with their research. Special thanks to Chris Cannam, Mark Levy and Chris Sutton for the underlying Vamp plugin system and Vamp API, Chris Landone and Chris Cannam for coding many of the feature extractor plugins, Yves Raimond, Chris Sutton, Chris Cannam, David Pastor Escuredo and Kurt Jacobson for their work on ontologies used by this system, Mark Levy and Chris Sutton for the original SoundBite software and data collection system, Dan Tidhar and Sefki Kolozali for their work on the SoundBite Dataset, and Kurt Jacobson for his help in server configuration and maintenance. SAWA is developed and maintained by George Fazekas, supported by the School of Electronic Engineering and Computer Science, Queen Mary University of London, the EPSRC-funded ICT project OMRAS-2 (EP/E017614/1), and through the NEMA project, the Andrew W. Mellon Foundation.