Research Paper 進化言語学の方法による音楽の文脈自由規則の発見

東条, 敏

音楽の和音進行のカデンツ規則は,自然言語の文脈自由文法のクラスに属する.本研究では楽曲から文脈自由規則を発見する試みとして,音楽理論GTTMに基づくタイムスパン木に係り受け関係を見出すこと,進化言語学のモデルILMによる文法発見の規則を効率化することの両面から研究を行った.まずタイムスパン木にはピッチ(音高)の概念を組み込み,それにより和音が発見できることから,和音連鎖列の妥当性をTPS理論から評価した.さらにタイムスパンの構成から楽曲間の距離を測る方法を考案し,この数値的距離と心理的類似性との関係を研究した.進化言語学の方法においては文脈自由規則の短縮化などを用いILMの効率化を図った. : There exist context-free grammar (CFG) rules in music, e.g., for sequences of chords as cadences. In this research, we have investigated the formalisms to detect such CFG rules in music from two different aspects. First, we relied on the Generative Theory of Tonal Music (GTTM) and found a long-distance dependency, that is a witness for being CFG, in its time-span trees. Second, we have tried a method on evolutionary linguistics called the iterated learning model (ILM), and improved the efficiency of the grammar detection. In time-span tree, we introduced the notion of pitch, and thus we could detect the most plausible sequence of chords, employing the theory of tonal pitch space (TPS). In addition, we have defined the distance between two time-span trees, by the sum of different time-spans, and then we have evaluated the correlation between this distance and our psychological similarity. As for ILM, we have invented a method of string clipping, and thus we could improve the efficiency.

Number of accesses :  

Other information