Fig. 13

Absolute and relative occurrences of chord classes in the original annotations (left, centre-left), and after removal of consecutively repeated chords (right, centre-right). Absolute occurrences are counted and accumulated throughout the corpus, whereas relative occurrences are first aggregated per annotation, as frequencies, then averaged across the whole dataset. Note that the “N” chord class denotes the “silent chord” as per the Harte notation (obtained for all subsets).