Theme Transformer: Symbolic Music Generation with Theme-Conditioned Transformer
نویسندگان
چکیده
Attention-based Transformer models have been increasingly employed for automatic music generation. To condition the generation process of such a model with user-specified sequence, popular approach is to take that conditioning sequence as priming and ask decoder generate continuation. However, this prompt-based cannot guarantee would develop or even simply repeat itself in generated In paper, we propose an alternative approach, called theme-based conditioning, explicitly trains treat thematic material has manifest multiple times its result. This achieved two main technical contributions. First, deep learning-based uses contrastive representation learning clustering automatically retrieve materials from pieces training data. Second, novel gated parallel attention module be used sequence-to-sequence (seq2seq) encoder/decoder architecture more effectively account given decoder. We report on objective subjective evaluations variants proposed Theme conventional prompt-based baseline, showing our best can generate, some extent, polyphonic pop piano repetition plausible variations condition.
منابع مشابه
Model Transformer plugin generation
The current paper presents a new approach using generic and meta-transformations for generating platform-specific transformer plugins from model transformation specifications defined by a combination of graph transformation and abstract state machine rules (as used within the Viatra2 framework). The essence of the approach is to store transformation rules as ordinary models in the model space w...
متن کاملTheme And Variation Encodings with Roman Numerals (TAVERN): A New Data Set for Symbolic Music Analysis
The Theme And Variation Encodings with Roman Numerals (TAVERN) dataset consists of 27 complete sets of theme and variations for piano composed between 1765 and 1810 by Mozart and Beethoven. In these theme and variation sets, comparable harmonic structures are realized in different ways. This facilitates an evaluation of the effectiveness of automatic analysis algorithms in generalizing across d...
متن کاملSymbolic Implementation of the Best Transformer
This paper shows how to achieve, under certain conditions, abstract-interpretation algorithms that enjoy the best possible precision for a given abstraction. The key idea is a simple process of successive approximation that makes repeated calls to a theorem prover, and obtains the best abstract value for a set of concrete stores that are represented symbolically, using a logical formula.
متن کاملMusic Mood Classification of Television Theme Tunes
This paper introduces methods used for Music Mood Classification to assist in the automated tagging of television programme theme tunes for the first time. The methods employed use a knowledge driven approach with tailored parameters extractable from the Matlab MIR Toolbox [1]. Four new features were developed, three based on tonality and one on tempo, to enable a degree of quantified tagging, ...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: IEEE Transactions on Multimedia
سال: 2022
ISSN: ['1520-9210', '1941-0077']
DOI: https://doi.org/10.1109/tmm.2022.3161851