Diverse Melody Generation from Chinese Lyrics via Mutual Information Maximization
This work aims to improve the quality and diversity of melody generation for Chinese lyrics, which is an incremental improvement for music composition systems.
This paper addresses the challenge of generating diverse melodies from Chinese lyrics by adapting mutual information maximization. The proposed method, Diverse Melody Generation (DMG), uses scheduled sampling and force decoding to enhance alignment, resulting in more pleasing and coherent tunes according to subjective tests.
In this paper, we propose to adapt the method of mutual information maximization into the task of Chinese lyrics conditioned melody generation to improve the generation quality and diversity. We employ scheduled sampling and force decoding techniques to improve the alignment between lyrics and melodies. With our method, which we called Diverse Melody Generation (DMG), a sequence-to-sequence model learns to generate diverse melodies heavily depending on the input style ids, while keeping the tonality and improving the alignment. The experimental results of subjective tests show that DMG can generate more pleasing and coherent tunes than baseline methods.