NAVIGATING THE LANDSCAPE OF COMPUTER AIDED ALGORITHMIC COMPOSITION SYSTEMS: A DEFINITION, SEVEN DESCRIPTORS, AND A LEXICON OF SYSTEMS AND RESEARCH Christopher Ariza New York University Graduate School of Arts and Sciences New York, New York ABSTRACT Towards developing methods of software comparison and analysis, this article proposes a definition of a computer aided algorithmic composition (CAAC) system and offers seven system descriptors: scale, process-time, idiom-affinity, extensibility, event production, sound source, and user environment. The public internet resource algorithmic.net is introduced, providing a lexicon of systems and research in computer aided algorithmic composition. 1. DEFINITION OF A COMPUTER-AIDED ALGORITHMIC COMPOSITION SYSTEM Labels such as algorithmic composition, automatic composition, composition pre-processing, computer-aided composition (CAC), computer composing, computer music, procedural composition, and score synthesis have all been used to describe overlapping, or sometimes identical, projects in this field. No attempt will be made to distinguish these terms, though some have tried (Spiegel 1989; Cope 1991, p. 220; Burns 1994, p. 195; Miranda 2000, pp. 9-10; Taube 2004; Gerhard and Hepting 2004, p. 505). In order to provide greater specificity, a hybrid label is introduced: CAAC, or computer aided algorithmic composition. (This term is used in passing by Martin Supper (2001, p. 48).) This label is derived from the combination of two labels, each too vague for continued use. The label “computer aided composition” lacks the specificity of using generative algorithms. Music produced with notation or sequencing software could easily be considered computer aided composition. The label “algorithmic composition” is likewise too broad, particularly in that it does not specify the use of a computer. Although Mary Simoni has suggested that “because of the increased role of the computer in the compositional process, algorithmic composition has come to mean the use of computers…” (2003), there remain many historical and contemporary compositional techniques that, while not employing the computer, are properly described as algorithmic. David Cope supports this view, stating that “… the term ‘computer’ is not requisite to a definition of algorithmic composition…” (1993, p. 24). Since 1955 a wide variety of CAAC systems have been created. Towards the aim of providing tools for software comparison and analysis, this article proposes seven system descriptors. Despite Lejaren Hiller’s well-known claim that “computer-assisted composition is difficult to define, difficult to limit, and difficult to systematize” (Hiller 1981, p. 75), a definition is proposed. A CAAC system is software that facilitates the generation of new music by means other than the manipulation of a direct music representation. Here, “new music” does not designate style or genre; rather, the output of a CAAC system must be, in some manner, a unique musical variant. An output, compared to the user’s representation or related outputs, must not be a “copy,” accepting that the distinction between a copy and a unique variant may be vague and contextually determined. This output may be in the form of any sound or sound parameter data, from a sequence of samples to the notation of a complete composition. A “direct music representation” refers to a linear, literal, or symbolic representation of complete musical events, such as an event list (a score in Western notation or a MIDI file) or an ordered list of amplitude values (a digital audio file or stream). Though all representations of aural entities are necessarily indirect to some degree, the distinction made here is not between these representations and aural entities. Rather, a distinction is made between the representation of musical entities provided to the user and the system output. If the representation provided to the user is the same as the output, the representation may reasonably be considered direct. A CAAC system permits the user to manipulate indirect musical representations: this may take the form of incomplete musical materials (a list of pitches or rhythms), an equation, non-music data, an image, or meta-musical descriptions. Such representations are indirect in that they are not in the form of complete, ordered musical structures. In the process of algorithmic generation these indirect representations are mapped or transformed into a direct music representation for output. When working with CAAC software, the composer arranges and edits these indirect representations. The software interprets these indirect music representations to produce musical structures. This definition does not provide an empirical measure by which a software system, removed from use, can be isolated as a CAAC system. Rather, a contextual [email protected] of scope is provided, based in part on use case. Consideration must be given to software design, functionality, and classes of user interaction. This definition is admittedly broad, and says only what a CAAC system is not. This definition includes historic systems such as the Experiments of Hiller and Isaacson (1959), Iannis Xenakis’s SMP (1965), and Gottfried Michael Koenig’s PR1 (1970a) and PR2 (1970b). In these cases the user provides initial musical and non-musical data (parameter settings, value ranges, stockpile collections), and these indirect representations are mapped into score tables. This definition likewise encompasses Xenakis’s GENDYN (1992) and Koenig’s SSP (Berg et al 1980). This definition includes any system that converts images (an indirect representation) to sound, such as Max Mathews and L. Rosler’s Graphic 1 system (1968) or Xenakis’s UPIC (1992; Marino et al. 1993). It does not matter how the images are made; they might be from a cellular automaton, a digital photograph, or hand-drawn. What matters is that the primary user-interface is an indirect representation. Some systems may offer the user