Author ORCID Identifier
https://orcid.org/0009-0005-8188-5422
Date of Award
Spring 5-12-2025
Document Type
Thesis (Ph.D.)
Department or Program
Computer Science
First Advisor
Michael A. Casey
Abstract
Listening to fast-tempo piano sonatas of the Classical period (circa 1750-1820) has been shown to have therapeutic effects for neurological disorders such as epilepsy. The limited existing repertoire of music in this style motivates the creation of more long-form, coherent compositions with clearly defined structure. Despite the long history of computer-based music generation and recent progress in deep learning, particularly transformer-based models, generating structurally coherent long-form music remains a major challenge. This difficulty stems from the scarcity of reliable structural annotation datasets, the computational demands of modeling very long musical sequences, and the lack of effective structural encoding in both symbolic music representations and model architectures.
The goal of this thesis is to design, implement, and evaluate a framework for generating coherent Classical sonata-style piano music, with a particular focus on extended duration and well-defined structural organization. The main contributions of this work are as follows: (i) AutoStruc, an unsupervised structural annotation system that integrates multi-modal music representations to construct phrase-section structural annotations from Classical piano sonatas; (ii) REMI-Lite, a novel symbolic music tokenization scheme that emphasizes a hierarchical organization of beat, bar, phrase; (iii) a generative framework that models the compositional process of Classical piano sonatas by assembling phrase-level components---generated by two transformer-based architectures, nextGEN and accGEN---into coherent, well-structured sections or movements; and (iv) task-specific evaluation methods, including a novel objective metric, Melody Retrieval, to assess model correctness, and behavioral studies to evaluate the musicality, rhythmic regularity, and structural coherence of the generated compositions.
To our knowledge, this is the first attempt to apply transformer-based architectures with a strong emphasis on structural modeling for the generation of Classical piano sonatas.
Recommended Citation
Feng, Yijing, "Computational Modeling and Structural Generation of Piano Music in the Classical Style" (2025). Dartmouth College Ph.D Dissertations. 394.
https://digitalcommons.dartmouth.edu/dissertations/394
