Phillip WU
It must be a very exciting speech, since the speaker gave a lot of audio demonstration to the audience. And the music examples are what I need, because I can hardly study the RmusicS without listening to it.
The speaker told the story about the history of development of computer-composition. Some important conclusions in each period have been mentioned. Researchers tried to find the musical RvocabularyS and RgrammarS, that is, the rules to extract sounds from chaos and organize them into music. But it is a great challenge since what counts in music is not the physical relation between sounds, but the relation we perceive when we listen to them. And we didn't learn more about ourselves.
So what we could do is to tell the computer what not to do, instead of what to do. Just like we are searching a right path in a map which contains millions and millions of paths. We are far from doing the right thing, but we never do something wrong by taking away the wrong ways. And that's why, I think, the professor keeps telling me the research work in this music cognition project is knowledge-based.
Shivani YARDI
The paper by Jean Claude Risset discusses the history of progress in the field of computational algorithms in the understanding of music. About 50 years ago, people experimented by randomly selecting fragments of a musical piece and building a sensible whole from those blocks based on intuition. later on, the focus was on the wise selection of various music parameters such as pitch, tone, etc. Based upon observations, a number of rules were established which laid constraints on the final structure. These rules, when quantified led to the idea of developing 'compositional algorithms' . Research is being carried out in the area of AI where one can train a machine to ape the thought process of a musician so that it can accompany him/her in real time. This involves the study of behaviour of a composer and the ability to extract the nearly exact attributes of the composer's style.
Posted 22 Jan 2003