Harmonic Training Dataset

Representation of harmonic data

The development of the presented melodic harmoniser incorporates statistical learning on different harmonic levels (chord transitions, cadences and voice leading) from a data pool with ‘real-world’ representations of historical traditions of music creation.

Musical pieces are manually annotated such that structural harmonic features may be extracted at various hierarchic levels. Specifically, for each piece/excerpt the following structural aspects are manually annotated:

•    Original musical piece score/transcription (musical surface: ms0)
•    Harmonic reduction(s) of each musical work/excerpt so that structural harmonic / non-harmonic notes are explicitly marked (musical surface: ms1)
•    Tonality and local scale/key changes are determined so that harmonic concepts relating to modulations can be learnt (indicated by a cluster of all notes included in the scale, with the lowest note designating the tonic)
•    Grouping structure is given so that cadential patterns at various hierarchic levels can be inferred (pitch height of notes is irrelevant; more notes correspond to higher level boundary).

Training template

Harmonic dataset

The dataset comprises seven broad categories of musical idioms, further divided into sub-categories, and presented in the following list:

1. Modal harmonisation in the Middle Ages (11th – 14th centuries): includes subcategories of the Medieval harmonic styles of Organum and Fauxbourdon.
    1.1. Organum excerpts or short pieces (11th-12th centuries).
    1.2. Fauxbourdon excerpts or short pieces (13th-14th centuries, Dufay, Binchois, et al).

2. Modal harmonisation in the Renaissance (15th – 17th centuries): includes modal music from the 16th – 17th centuries along with modal chorales.
    2.1. Collection of homophonic vocal music (madrigal, frottola, et al) by Arcadelt, Desprez, Gallus, Morley.
    2.2. Selected homophonic excerpts from Palestrina's "Stabat Mater".
    2.3. Selected modal homophonic chorales by Osiander, Praetorius, Scheidt, Hassler, Vulpius, Lasso, Walter, et al.     Further categorization by mode is possible.

3. Tonal harmonisation (17th – 19th centuries): includes a set of Bach Chorales, the Kostka-Payne corpus ( and tonal harmonisation sets from the 18th – 19th centuries.
    3.1. Selection of 35 Bach chorales from the Breitkopf edition. This set represents baroque homophonic harmonic         style (17th-18th century). Further categorization to major or minor is possible.
    3.2. The dataset of Kostka-Payne corpus (Tonal Harmony workbook, 3rd edition, McGraw-Hill, 1995), produced by     David Temperley (chord-list file) and Bryan Pardo (midi files with chord's quality). This set represents classical and     romantic harmonic style (18th-19th century).
    3.3. Selected excerpts or short pieces by Beethoven, Chopin, Schubert, Brahms, Liszt, Grieg, Mussorgsky, et al.         This set represents 19th century chromatic harmony.

4. Harmonisation in National Schools (19th – 20th centuries): includes 19th – 20th century harmonisation of folk songs from Norway, Hungary and Greece.
    4.1. Selection of harmonizations of Norwegian folk songs by E. Grieg from his opp. 17 & 66 (tonal chromatic                 harmony).
    4.2. Selection of harmonisations of Hungarian and Romanian folk songs by B. Bartok from his 10 easy pieces, 14     Bagatelles, Hungarian peasant songs & Six romanian dances (free chromatic harmony).
    4.3. Selection of harmonisations of Greek folk songs by G. Labelet (diatonic modal harmony).
    4.4. Selection of harmonisations of Greek folk songs by Y. Constantinidis from his "44 piano miniatures" (chromatic     modal/impressionistic harmony).

5. Harmonisation in the 20th century: includes mainly vocal music by Cl. Debussy, P. Hindemith, E. Whitacre, I. Stravinsky, among others. Also, includes 20th-century harmonic concepts extracted from short musical excerpts.
    5.1. Selected short excerpts of 20th century harmonisation concepts from the textbooks of Stefan Kostka, Kent             Williams, Walter Piston and various other sources, further categorized by harmonisation technique (chords with         added notes, diatonic modality, pandiatonic, impressionistic, octatonic, parallel chromatic, polychords, quartal             harmony, whole-tone harmony, secundal harmony)
    5.2. Selected choral music by Cl. Debussy (Trois Chansons), extended tonal/impressionistic harmony
    5.3. Selected choral music by P. Hindemith (Six Chansons), neotonal harmony (not publishable due to copyright         restrictions)
    5.4. Selected choral music by E. Whitacre (A boy and a girl, Sleep, Water Night), neotonal harmony (not                         publishable due to copyright restrictions)
    5.5. Selection of other vocal neotonal works by I. Stravinsky and A. Schnittke (not publishable due to copyright             restrictions)

6. Harmonisation in folk traditions: includes Tango (classical and nuevo styles), Epirus polyphonic songs and Rebetiko songs.
    6.1. Selection of Argentinian Tango music, in the classical style and the nuevo style of A. Piazzolla, extended tonal     harmony
    6.2. Selection of 3-voice and 4-voice polyphonic songs from Epirus (transcriptions by K. Lolis), minor pentatonic         harmony
    6.3. Selection of Greek Rebetiko songs (transcribed as melodies with chord symbols), idiosyncratic modal                     harmony, triadic harmonisation of oriental modes

7. Harmonisation in 20th-century popular music and jazz: includes mainstream jazz, piano pieces by Bill Evans and a collection of songs by The Beatles.
    7.1. Selection of jazz standards from the Real Book (melodies with chord symbols), mainstream jazz harmony
    7.2. Selected jazz harmonizations by Bill Evans (transcriptions by J. Reilly), extended jazz harmony (not                     publishable due to copyright restrictions)
    7.3. Selection of bossanova jazz style tunes (Songs by A. C. Jobim, Real Book style transcriptions)
    7.4. Selection of 10 songs by The Beatles (not publishable due to copyright restrictions)

The above dataset is actively updated, corrected and revised as the harmoniser is being tested and explored. The manual annotation and analysis of the musical pieces for the production of the dataset xml files was carried out by a research team coordinated and supervised by Costas Tsougras. The team comprises Costas Tsougras, Andreas Katsiavalos, Kostas Giannos and Nikolaos Diminakis.

CCM  group
CCM group website
AUTH website
COINVENT website