[Paper Review] Computoser - rule-based, probability-driven algorithmic music composition

Bozhidar Bozhanov|arXiv (Cornell University)|Dec 9, 2014

Music Technology and Sound Studies5 references3 citations

TL;DR

Computoser proposes a hybrid algorithmic music composition system that integrates rule-based composition with data-driven probability models derived from analyzing over 500 musical pieces. By combining statistical probabilities of musical elements (e.g., scale, tempo, intervals) with structured composition rules from established practices, the system generates musically coherent and varied pieces, achieving 58.4% listener approval in evaluations.

ABSTRACT

This paper presents the Computoser hybrid probability/rule based algorithm for music composition (http://computoser.com) and provides a reference implementation. It addresses the issues of unpleasantness and lack of variation exhibited by many existing approaches by combining the two methods (basing the parameters of the rules on data obtained from preliminary analysis). A sample of 500+ musical pieces was analyzed to derive probabilities for musical characteristics and events (e.g. scale, tempo, intervals). The algorithm was constructed to produce musical pieces using the derived probabilities combined with a large set of composition rules, which were obtained and structured after studying established composition practices. Generated pieces were published on the Computoser website where evaluation was performed by listeners. The feedback was positive (58.4% approval), asserting the merits of the undertaken approach. The paper compares this hybrid approach to other approaches to algorithmic composition and presents a survey of the pleasantness of the resulting music.

Motivation & Objective

• To address the limitations of existing algorithmic composition methods, which often produce unpleasant or monotonous music.
• To reduce lack of variation in rule-based systems by incorporating statistical probabilities derived from real musical data.
• To develop a hybrid approach that balances musical coherence with creative diversity through data-informed rule application.
• To evaluate the perceptual quality of generated music through listener feedback, ensuring practical musicality.

Proposed method

• The system analyzes a corpus of 500+ musical pieces to extract statistical probabilities for musical characteristics such as scale, tempo, and intervallic relationships.
• It applies a large set of composition rules derived from established music theory and pedagogical practices to guide structural and harmonic decisions.
• Musical elements are selected probabilistically based on the derived statistics, ensuring data-informed variation while maintaining tonal and rhythmic consistency.
• The algorithm generates complete musical pieces by iteratively applying probabilistic selection and rule-based constraints across multiple musical dimensions.
• The implementation is publicly available as a reference, enabling reproducibility and further research.
• Listener evaluation was conducted via the Computoser website to assess perceived musical quality and acceptability.

Experimental results

Research questions

• RQ1 Can a hybrid approach combining rule-based composition with data-derived probabilities produce musically coherent and varied compositions?
• RQ2 How does the integration of statistical probabilities improve the perceived pleasantness of algorithmically generated music compared to purely rule-based systems?
• RQ3 To what extent do listeners perceive the generated pieces as musically valid and enjoyable?
• RQ4 Does the use of real musical data to inform probability distributions enhance the diversity and expressiveness of generated compositions?
• RQ5 How does the system's output compare to existing algorithmic composition methods in terms of listener approval and musical quality?

Key findings

• The system achieved a 58.4% approval rate among listeners, indicating a positive perception of the generated music's quality and pleasantness.
• The hybrid model successfully reduced the monotony and unpleasantness often found in purely rule-based or probability-only approaches.
• Statistical analysis of 500+ musical pieces provided reliable probability distributions for key musical parameters such as scale, tempo, and interval usage.
• The integration of music theory rules with data-driven probabilities enabled the generation of structurally coherent and rhythmically varied compositions.
• The reference implementation is publicly available, supporting reproducibility and further development in algorithmic composition.