Phillip Kirlin

My research interests lie in the field of music informatics, an interdisciplinary area that uses and develops new computational methods and models for processing and producing music and musical information. Specifically, I am interested in advancing the state of research in artificial intelligence, machine learning, and probabilistic modeling by using music-related problems as motivators. I am intrigued by this field precisely because it combines computational study with an art form that requires a great deal of contextual knowledge.

Combining artificial intelligence and music leads to a wide variety of problems for study. My focus is on computational music theory and analysis. Problems I'm interested in include

• Computational studies of hierarchical music analysis: I'm analyzing Schenkerian analysis using probabilistic models to uncover the statistical regularities in the way people analyze music. Check out my dissertation for the most recent findings.
• Harmonic and melodic analysis: I'm interested in going beyond chord labeling. Can we get a complete functional harmonic analysis with Roman numerals, figured bass, and identification of all non-harmonic tones with their functions? What about analyzing the voice-leading of a piece; can we identify the melodic connections between notes at multiple structural levels?
• Finding a grammar of music: Does it exist? Is it possible to algorithmicize Lerdahl and Jackendoff's A Generative Theory of Tonal Music?
• Algorithmic composition and machine creativity

Besides music informatics, I'm also interested in a broad range of AI topics, especially machine learning. I've also done some work in sabermetrics and geographic information systems.