Abstract:

We have developed an audio information retrieval system called, Audio Metaphor, that utilizes large online collaborative databases for real time soundscape composition. Audio Metaphor has been used in a contemporary dance piece IN[A]MOMENT. The audience interacts with the system by sending the system Tweets. At the heart of the Audio Metaphor is a sub-query generation algorithm. This algorithm which we name SLiCE, for string list chopping experiments, accepts a natural language phrase that is parsed into a text feature string list, which is chopped into sub-query search to find a combination of queries with non-empty mutually exclusive results. Employing SLiCE, Audio Metaphor processes natural language phrases to find a combination of audio file results that represents the input phrase. In parallel, the parsed input phrase is used to search for related Tweets which are similarly put through SLiCE. Search results are then sent to a human composer to combine and process into a soundscape.