SAFE: A system for the extraction and retrieval of semantic audio descriptors

We present an overview of the Semantic Audio Feature Extraction (SAFE) Project, a novel data collection architecture for the extraction and retrieval of semantic descriptions of musical timbre, deployed within the digital audio workstation. By embedding the data capture system into the music production workflow, we are able to maximise the return of semantically annotated music production data, whilst mitigating against issues such as musical and environmental bias. Users of the plug-ins are able to submit semantic descriptions of their own music, whilst utilising the continually growing collaborative dataset of musical descriptors. In order to provide more contextually representative timbral transformations, the dataset is partitioned using metadata, captured within the application.