This paper treats the development of taxonomies that can be used for musical audio mining. Addressed are: the need for a user dependent taxonomy, and the description of categories and concepts that emerge from a statistical analysis of 1148 vocal queries of 72 different persons.