I have just got back, well a couple of days ago, from my most recent conference the British Forum for Ethnomusicology and the Royal Musical Association postgraduate students conference. This is a lively conference where postgraduate researchers get together to support each other in presenting their research. This conference is a great place to showcase the early stages of your work and get feedback from students and academics within the field, in a friendly and supportive environment.
I gave my paper on the first day:
Great to see some of our #MusicatSouthampton PhD students getting involved in the BFE/RMA Research Students’ Conference @Hudds_Uni_Music this week – wishing everyone @bferma2018 an exciting few days sharing their research!@HumanitiesUoS@ArtsUniSouth#bferma2018https://t.co/yeaL61nBiv
— Music at Southampton (@UoSMusic) January 4, 2018
This was the first paper I have given on my thesis research directly. Here I explored how I am using one method of quantifying musical similairities from a melodic line. I have built upon the work of Hirata and Matsuda, and showed how we can extend their model to discuss other prominent methods of music analysis in this way. I always find the most useful components of any conference paper is the chance to discuss and answer questions on your topic. I was given interesting topics that questioned why I was using traditional music theories, and why I was not interested in the traditional ‘genre’ classification.
For anyone interested the abstract I provided for the conference was as follows:
The release of Napster in 1999 saw exponential growth of online music streaming. This has inspired the development of tools to discover new music, typically as part of applications such as Spotify. Music recommendation often suggests new music to users based on their own and others’ preferences (Herlocker et al., 2004; Celma, 2010) or audio-based methods of “similarity” analysis (Downie, 2008). I propose that score-based musical analysis can provide a more useful similarity comparison than audio analysis.
To date, there has been a lack of success in extracting high-level musical features from audio. In contrast, traditional music analysis methods (e.g. Schenker) enable the evaluation of high-level musical features, including harmony, timbre, and melody. These features could be compared to determine the similarity of two pieces of music. I will discuss how traditional music analysis methods can be utilised to determine musical similarity and propose ways of quantifying similarity to enable computation. Additionally, I will report on the provisional findings of a listener study examining the perceived audibility of theoretical definitions of musical similarity.
This work will enable cross-genre musical recommendation based on the fundamental features of a piece of music. For the consumer this means more relevant and accurate recommendations, and for the emerging artists greater exposure.
The conference was a great opportunity to begin to share my thesis research, and introduce the concepts I am developing to the wider research community. The University of Southampton was strongly represented, and we have a great time – especially enjoying afternoon tea.
My next presentation is on the 23rd January at DANS in the Netherlands.