The IRMA Community
Newsletters
Research IRM
Click a keyword to search titles using our InfoSci-OnDemand powered search:
|
Markov-Chain Melody: A Mathematical Approach to AI-Assisted Songwriting
Abstract
This study examined whether simple probabilistic models can generate coherent melodic fragments for AI-assisted songwriting. The work focused on building first- and second-order Markov chains, analyzing the statistical properties of their outputs, and comparing listener perceptions with human-composed hooks. A dataset of 20 popular MIDI melodies was standardized to C major and used to estimate transition probabilities, from which 30 sequences of 8–16 notes were generated. Structural measures showed that human melodies had higher entropy (2.29 vs. 1.80) and greater pitch diversity (6.37 vs. 5.92), while Markov outputs had higher repetition (0.43 vs. 0.34). A perceptual survey with 30 listeners rated human melodies higher in catchiness (4.19 vs. 3.78), musicality (4.33 vs. 3.63), and preference (3.98 vs. 3.68), though Markov sequences were still viewed as acceptable. Findings show that Markov chains capture short-term tendencies and can generate usable motifs but lack long-range structure, making them useful for low-resource music tasks.
Related Content
|
Randy Joy Magno Ventayen, Elmar B. Noche, Maricel Aqui.
© 2026.
34 pages.
|
|
Evren Idil Yazan.
© 2026.
30 pages.
|
|
Sonia Ahuja.
© 2026.
24 pages.
|
|
Saroj Sandeep Phalke, Gitanjali Shinde, Grishma Bobhate, Poonam Railkar, Haribhau R. Bhapkar, Sonal Fatangare.
© 2026.
28 pages.
|
|
Syrenzo Ramos Sicuan.
© 2026.
26 pages.
|
|
K. S. Jishnu, P. S. Shijukumar, G. Bhargavi, Vimal Sankar, P. S. Sujith Kumar, Nisha Thorakattu Madathil.
© 2026.
30 pages.
|
|
Vibha Tiwari, Rahul Lalwani, Akshada Telang.
© 2026.
36 pages.
|
|
|