## HIERARCHICAL MARKOV MODELING FOR GENERATIVE MUSIC

### BibTeX

@MISC{Thornton_hierarchicalmarkov,

author = {Chris Thornton},

title = {HIERARCHICAL MARKOV MODELING FOR GENERATIVE MUSIC},

year = {}

}

### OpenURL

### Abstract

The paper describes a hierarchical Markov modeling strategy that offers the advantages of a statistical approach without constraining the level of analysis. The method can mediate music generation from sample compositions. Some illustrative examples are described. 1.

### Citations

1777 | An introduction to Kolmogorov complexity and its applications
- Li, Vitányi
- 1997
(Show Context)
Citation Context ...→ Y, Z Y → A, B/C Z → D form a hierarchical model for the sequences A B C and A C D because both can be generated from X. 2 Although the task of finding the simplest model is noncomputable in general =-=[9]-=- in the particular case of a hierarchical rewrite model, there is the possibility of breaking the task down into tractable steps. The task of finding the smallest hierarchical model can be decomposed ... |

72 | Multiple viewpoint systems for music prediction
- Conklin, Witten
- 1995
(Show Context)
Citation Context ...eling should have some way of discovering and adapting to different levels of analysis, so as to incorporate the kinds of hierarchical functionality associated with, e.g., the viewpoints framework of =-=[3]-=- and analytic applications of classification [4; 5], style analysis [6] and segmentation [7]. But if this involves introduction of domain knowledge, the question is raised of how we can be confident i... |

66 | Folk music classification using hidden Markov models
- Chai, Vercoe
- 2001
(Show Context)
Citation Context ...apting to different levels of analysis, so as to incorporate the kinds of hierarchical functionality associated with, e.g., the viewpoints framework of [3] and analytic applications of classification =-=[4; 5]-=-, style analysis [6] and segmentation [7]. But if this involves introduction of domain knowledge, the question is raised of how we can be confident in our choice of musical descriptors. 1 In principle... |

46 |
Automatic genre classification of music content: a survey
- Scaringella, Zoia, et al.
- 2006
(Show Context)
Citation Context ...apting to different levels of analysis, so as to incorporate the kinds of hierarchical functionality associated with, e.g., the viewpoints framework of [3] and analytic applications of classification =-=[4; 5]-=-, style analysis [6] and segmentation [7]. But if this involves introduction of domain knowledge, the question is raised of how we can be confident in our choice of musical descriptors. 1 In principle... |

31 |
Using Machine-Learning Methods for Musical Style Modeling
- Dubnov, Assayag, et al.
- 2003
(Show Context)
Citation Context ...els of analysis, so as to incorporate the kinds of hierarchical functionality associated with, e.g., the viewpoints framework of [3] and analytic applications of classification [4; 5], style analysis =-=[6]-=- and segmentation [7]. But if this involves introduction of domain knowledge, the question is raised of how we can be confident in our choice of musical descriptors. 1 In principle, hidden Markov mode... |

22 | Representation and discovery of vertical patterns
- Conklin
- 2002
(Show Context)
Citation Context ... Hierarchical approaches offer flexibility in the level of analysis but may require use of special-purpose sampling and discovery, which may necessitate bringing domain-specific descriptors into play =-=[10]-=-. The proposed approach combines the sampling advantages of a statistical/informational approach with the flexibility of a hierarchical method. No special-purpose sampling is involved. As in hierarchi... |

11 | Adaptive text mining: Inferring structure from sequences
- Witten
- 2004
(Show Context)
Citation Context ...ving the element of choice. Hierarchical knowledge is not imposed on the statistical process. Rather, it emerges, much in the way that grammatical structure emerges in hierarchical dictionary methods =-=[8]-=-. The method decomposes modeling into two steps. In the initial step, an informationally optimal hierarchical model (based on rewrite rules) is derived. In the second step, a Markov model is derived f... |

5 |
A Generative Model for Rhythms
- Paiement, Grandvalet, et al.
- 2007
(Show Context)
Citation Context ...e form of Markov model, e.g., [1], and use of transition probabilities for purposes of building up a novel sequence. A problem with such approaches is the difficulty of choosing the level of analysis =-=[2]-=-. A statistical model is necessarily constructed in terms of whatever entity is selected for sampling; the level of analysis is then fixed in advanced. If we construct a statistical model on the basis... |

4 |
Extracting patterns in music for composition via markov chains
- Verbeurgt, Dinolfo, et al.
- 2004
(Show Context)
Citation Context ... by examining the relative probabilities of musical subsequences and generating a piece which conforms to the observed distribution. This often involves derivation of some form of Markov model, e.g., =-=[1]-=-, and use of transition probabilities for purposes of building up a novel sequence. A problem with such approaches is the difficulty of choosing the level of analysis [2]. A statistical model is neces... |

3 |
An Information-dynamic Model of Melodic Segmentation
- Müllensiefen, D, et al.
- 2008
(Show Context)
Citation Context ...s to incorporate the kinds of hierarchical functionality associated with, e.g., the viewpoints framework of [3] and analytic applications of classification [4; 5], style analysis [6] and segmentation =-=[7]-=-. But if this involves introduction of domain knowledge, the question is raised of how we can be confident in our choice of musical descriptors. 1 In principle, hidden Markov models have the potential... |