Semantic Sonification

The system reads a text and returns a piece of sound about it. It works from a library of sounds, each recorded or synthesised, and each described in words. A linguistic pipeline I programmed in Java reads those descriptions alongside the input text and measures how close their meanings lie, so that a passage about the sea calls up the sounds whose descriptions are near it. From those matches the system builds a piece with a structure: movements that succeed one another, material that returns, so the result holds as a composition rather than a sequence. It writes the score as a SuperCollider program that assembles the piece event by event, naming each passage after the fragment of text it stands for.