Argentine Tango is famous for being an expressive and emotional dance form. The word alone carries with it images of sensuality, passion, power and drama. As an improvised dance form, Tango allows the dancers to express what they feel from the music through their movement.
My BSc Dissertation aims to discover what it is in this music that makes us dance to it the way we do, and to see if we can extract this meaning from the music in a formalised manner, so that we can use techniques from the field of artificial intelligence to choreograph a performance. People often talk of the “universal language of music” – so why not treat it as such? One area I looked into was whether we can derive some form of grammar and/or semantic meaning from Tango music. This would allow us to use Natural Language Processing methods to understand how the music fits together, and make decisions based on that.
In order to aid in the analysis, I developed a simple prototype tool called LISA: LISten and Annotate. LISA will open a music file and allow notes to be made. So that these notes are aligned correctly, LISA performs automatic beat detection and will “snap” annotations to the nearest beat detected. The waveform can be viewed in one of two ways: as a classic stereo amplitude/time graph, or a spectral frequency analysis. LISA was developed in C# using the FMOD sound library.
I found this an absolutely fascinating project to work on. It covered so many different disciplines, and drew them together for a cohesive purpose. Of course, this document was just preliminary research: a summary of some of the literature and an attempt to select current techniques and theories which would be relevant to this application. I hope to experiment some more with these sorts of ideas in the years to come. I would love to see how close machines can come to human creativity.
Two documents were written for the final hand-in: the report itself, and an executive summary, which summarises the contents of the full report in four sides. These documents are now available for download.
 Executive Summary |
Final Report |