人工智慧與作曲的研究項目（音樂與科技的研究與實踐中心 2022年1月參觀）AI in Musical Composition Research Project: visit to PRiSM research centre (Jan.’22)

藉由南華大學研發處的資助，本人已開始進行新的專題研究：“人工智慧作為作曲的輔助工具以及靈感的來源：探索機器學習的風格轉移在音樂創作中的運用之潛力”。

Supported by funding from the Nanhua University Research and Development Office, I’ve begun a new research project: “AI as a compositional aid and a source of inspiration: exploring the potential application of machine learning style transfer to musical creativity”.

我2022年1月應研究中心總監Emily Howard的邀請曾參觀英國皇家北英格蘭音樂學院的音樂與科技研究與實踐中心（PRiSM）。我在研究中心裡有機會了解到他們令人贊嘆的超級電腦「PRiSM sample RNN」的功能，並與學者討論它在輔助作曲上的可能性。我們有討論我初步的研究方向，並規劃進一步的討論。

In January 2022 I was invited by Emily Howard, director of PRiSM (Practise and Research in Science and Music) at the Royal Northern College of Music, to visit the centre.  I had the chance to learn about the functionality of their breathtaking supercomputer “PRiSM sample RNN”, and to talk to their Post-Doctoral Research Fellow about its potential applications as a compositional aid.  We have already discussed a tentative direction for the research, and have organised further discussions.

PRiSM sample RNN與其他的音樂研究中心之神經網絡不同，它是專門分析音頻數據（而不是midi或樂譜）。因此它可以提供許多新鮮的研究方向。我初步的研究方案是試圖運用PRiSM sample RNN的神經網絡來分析並模仿（由南華大學學生錄製的）國樂或雅樂的錄音，接著將神經網絡所產生的風格模仿成果裡面有趣的內容記譜下來，重新用實際演奏來演奏此成果的。我會在此網站上發佈未來的發展。

PRiSM sample RNN seems to be somewhat different from neural networks in other music research centres, in that it processes audio date, as opposed to midi or notated scores.  It therefore affords several fresh directions for research.  My initial research idea is to try to use PRiSM sample RNN’s neural net to analyse and imitate recordings (made by students at Nanhua University) of Chinese traditional music and Yayue (Tang dynasty ceremonial court music), in turn taking the most interesting portions of the imitative results produced by the neural net and renotating them for actual performers.  I shall post further developments on my website.