The keynote speaker of BENELEARN will be Tijl de Bie from the University of Bristol, UK.
The title of his talk is “Machine Learning for Music Analysis“.
As humans we take our ability to instinctively ‘understand’ music in terms of rhythm, harmony, melody, mood,… for granted. However, it is remarkably hard to explain in algorithmic terms how we do this, in a way that is implementable on a computer. Yet, the creative and commercial potential of being able to do this makes this a task worth pursuing.
In this talk I will survey the state-of-the-art on various music analysis tasks, from automated chord recognition, over mood detection, to scoring the hit potential of a pop song. I will highlight that the best performing algorithms tend to be data-driven, machine learning algorithms. In doing this, it will become clear how this research area has successfully overcome the scarcity of high quality label data by utilising high volumes of poor quality data instead. I will end by discussing some future challenges and opportunities this vibrant research area holds for the machine learning research community.