Composers: Pop* II
Metamusicians: Paul Bodily, and Dan Ventura, CS Department,
BYU, Provo, Utah
Description of the Work
Pop* (pronounced Pop-Star) II, which was featured in the 2017 MuMe concert, is an automated pop lead sheet composer. It uses a modular framework to generate verse-chorus structure, rhyme-scheme, lyrics, harmony, and melody. Pop* creates novel full-length pop songs in lead sheet format with no external input beyond an inspiring set of pop lead sheets. To concretely render compositions, we generate both printed sheet music and MP3 audio recordings. MP3 audio files feature computer-sung lyrics accompanied by synthesized piano and bass comping chords.
Technical Description
Pop* II uses a hierarchical Bayesian program learning model, meaning that the concept of a pop composition is factored into subconcept models such as structure, lyrics, harmony, melodic pitch, and melodic rhythm. These subconcepts are further factored until subconcepts represent simple enough ideas to be approximated using data-driven (conditional) probability distributions. Generation of novel compositions is achieved by combining subconcept values as they are probabilistically sampled from subconcept distributions.
Currently in its second iteration, the system uses probabilistic constrained Markov models to generate sequences for each musical viewpoint. In their traditional form, constrained Markov models allow structure to be imposed on sequential data using unary constraints at any of several sequence positions. We have expanded these models to allow for binary relational constraints which enables the system to impose meaningful patterns of repetition—including motifs, verse-chorus structures, and rhyme schemes—while still sampling composition-length sequences as a single Markov process. The relational constraints for these models are also automatically learned from data. This is done by inferring viewpoint-specific structural repeats from existing lead sheets using a self-alignment technique and then converting this structure into a set of relational constraints.
The system incorporates added elements of autonomy, inspiration, self-awareness, and framing using semantic analysis and social networking to intelligently choose an inspiring set.
Begin the Movement
Begin the Movement is the first song to be composed by Pop* II and was created using structure learned from Twinkle, Twinkle, Little Star; harmonic and melodic models from Over the Rainbow; and a lyric model from Imagine and Hey, Jude.
Audio recording
Lead sheet
Biography
Paul Bodily is a PhD candidate in the CS department at Brigham Young University (BYU). Under the advisement of Dr. Dan Ventura, his research focuses on machine learning in pop music with the intent of building data-driven generative systems.
Dr. Dan Ventura is a CS professor at BYU whose focus is on computational creativity systems generally. Students under his advisement have published systems in domains such as artistic image generation (DARCI), recipe generation (PIERRE), jazz lead sheet composition (CARL), and neology (Nehovah).
