This invention relates particularly to a method and apparatus in which portions of a speech stream (1) can be categorized with or without a visual representation, by user command and/or by automatic recognition of speech qualities and (2) can then be selectively retrieved from a storage.