Seminar by Daniel Wolff "Spot The Odd Song Out: Similarity models in analysis of corpora and listener groups"
Daniel Wolff, currently visiting researcher at MTG, gives a talk about "Spot The Odd Song Out: Similarity models in analysis of corpora and listener groups." on Thursday, November 26th 2015, at 3:30pm in room 52.S31.
The concept of similarity can be applied to music in a multitude of ways. Applications include systems which provide similarity estimates depending on the specific user and context as well as analysis tools that show similarity of music with regards to specified compositional, physical or contextual features. The ASyMMuS project allows musicologists to apply similarity analysis to musical corpora on a big-data infrastructure - allowing for a comparison of e.g. the works of a certain composer.
For analysis of music reception, perceived similarity is of interest. It is specific to individuals and influenced by a number of factors such as cultural background, age and education. We will discuss how to adapt similarity models to the relative similarity data collected in the game with a purpose "Spot The Odd Song Out" (http://mirg.city.ac.uk/casimir/game). Models are parametrised with the help of machine learning technique. Experiments show that models can be adapted to this very general similarity data - depending on the amount of data available. With transfer learning, methods can be used on smaller, e.g. group-specific datasets, which we utilise for a comparative analysis of similarity models between different user groups.