1
Petri Toiviainen & Tuomas Eerola University of Jyväskylä, Department of Music At the end of the 19th century, the Finnish Literary Society (SKS) collected tens of thousands of folk tunes from all over the Finland. Ilmari Krohn published about 9000 of these between the years 1898-1933. This collection is known as the Finnish Folk Songs (Suomen Kansan Sävelmiä). It consists of several subcollections: Folk Songs (Laulusävelmät) Spiritual Folk Songs (Hengelliset sävelmät) Folk Dances (Kansantanssit) Kantele and Jouhikko Tunes Runo-tunes 1. Digitizing the Finnish Folk Tunes 2. Building a digital archive 3. Using the archive for research data mining comparative research: musical features lyrics Development of tools for computational analysis (MidiToolbox) We aim to provide public access to the collection by publishing the archive on the internet. In this application, individual tunes can be searched according to various criteria (villages, regions, lyrics, musical features, collectors). The search engine returns musical notation, MIDI file and textual information for each tune fulfilling the search criteria. We also plan to develop search methods based on musical content (e.g. query-by-whistling). E-mail: [email protected] Web: http://www.jyu.fi/musica/sks/ Financial support: The Virtual University of Finland Music Education and Research on the Net (MOVE) Background Aims Geographical differences Musical features User-interface of the search engine. Search engine Contact information Music Cognition Group UNIVERSITY OF JYVÄSKYLÄ Lyrics Pitch-class distributions Prior to the analysis the tunes have been transposed to either C major or C minor according to the key signature. Overall, the distributions for both subcollections resemble those observed in other types of tonal music (European Folk Songs, Western Art Music, Bebop Jazz, etc.). In Runo-tunes the minor third is more common than the major third while in Folk Songs the reverse holds. A similar difference can be observed in the frequencies of the tonic and the dominant. Comparison of subcollections Locations of origin The location of origin (village, municipality etc.) of each tune has first been converted to latitude and longitude values. Using this information the density of tune locations has been displayed graphically for each subcollection. Proportion of tunes in ternary meter Variation (entropy) of note durations Average interval size (semitones) Melodic range (semitones) Proportion of tunes starting with the dominant Proportion of tunes starting with the tonic Folk Songs N=2772 N=280 N=1017 N=4842 Sorrow The Folk Song collection contains the lyrics of the first verse of each song. There are a total of 74596 words, of which 13857 are unique (incl. inflectional forms). Frequencies of words related to some interesting themes are portrayed here. Drinking Love Joy C C# D D# E F F# G G# A A# B 0 0.05 0.1 0.15 0.2 0.25 Sävelluokka Osuus C C# D D# E F F# G G# A A# B 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 Sävelluokka Osuus -P8 -M6 -d5 -m3 P1 +m3 +d5 +M6 +P8 0 0.05 0.1 0.15 0.2 0.25 Intervalli Osuus -P8 -M6 -d5 -m3 P1 +m3 +d5 +M6 +P8 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 Intervalli Osuus Folk Songs Folk Songs Runo-tunes Runo-tunes A comparison of musical features of Folk Songs and Runo-tunes Interval distributions The interval distributions for both subcollections are typical for Folk music; small intervals are predominant and descending scalar motion more frequent than ascending. In Folk Songs tone repetition is more common than in Runo-tunes. Large intervals occur less frequently in Runo-tunes than in Folk Songs. Kantele artist Teppana Jänis. Photo: A. O. Väisänen (Kantele and Jouhikko Tunes, XXVI, © SKS, 1928). Application of Musical Data Mining Proportion of tunes in minor key Geographical variation of various musical features in Folk Songs (N=4842). The colorbar on the right of each picture shows the range of variation. © 2003 Petri Toiviainen & Tuomas Eerola The collection is musically comprehensive, nationally important and well documented. Finland Russia Sweden Estonia Runo-tunes Kantele and Jouhikko Tunes Spiritual Folk Tunes Proportion Proportion Interval Interval Proportion Proportion Pitch-class Pitch-class

Application of Musical Data Mining - Jyväskylän yliopistoesavelmat.jyu.fi/pdf/a4_eng.pdf · Society (SKS) collected tens of thousands of folk tunes from all over the Finland

Embed Size (px)

Citation preview

Petri Toiviainen & Tuomas Eerola University of Jyväskylä, Department of Music

At the end of the 19th century, the Finnish LiterarySociety (SKS) collected tens of thousands of folk tunes from all over the Finland. Ilmari Krohn published about 9000 of these between the years1898-1933. This collection is known as the FinnishFolk Songs (Suomen Kansan Sävelmiä). It consistsof several subcollections:

� Folk Songs (Laulusävelmät)� Spiritual Folk Songs (Hengelliset sävelmät)� Folk Dances (Kansantanssit)� Kantele and Jouhikko Tunes� Runo-tunes

1. Digitizing the Finnish Folk Tunes2. Building a digital archive3. Using the archive for research

� data mining� comparative research:

• musical features• lyrics

� Development of tools for computational analysis (MidiToolbox)

We aim to provide public access to the collection by publishing the archive on the internet. In this application, individual tunes can be searched according to various criteria (villages, regions, lyrics, musical features, collectors). The search engine returns musical notation, MIDI file and textual information for each tune fulfilling the search criteria. We also plan to develop search methods based on musical content (e.g. query-by-whistling).

E-mail: [email protected]: http://www.jyu.fi/musica/sks/Financial support:

The Virtual University of Finland

Music Education and Research on the Net (MOVE)

Background

Aims

Geographical differences

Musical features

User-interface of the search engine.

Search engine

Contact information

Music Cognition Group

UNIVERSITY OF JYVÄSKYLÄ

Lyrics

Pitch-class distributions

Prior to the analysis the tunes have been transposed to either C major or C minor according to the key signature.

Overall, the distributions for both subcollections resemble those observed in other types of tonal music (European Folk Songs, Western Art Music, Bebop Jazz, etc.).

In Runo-tunes the minor third is more common than the major third while in Folk Songs the reverse holds.

A similar difference can be observed in the frequencies of the tonic and the dominant.

Comparison of subcollections

Locations of origin

The location of origin(village, municipalityetc.) of each tune hasfirst been converted to latitude and longitudevalues.

Using this informationthe density of tunelocations has beendisplayed graphicallyfor each subcollection.

Proportion of tunes in ternarymeter

Variation (entropy) of notedurations

Average interval size (semitones)Melodic range (semitones)

Proportion of tunes starting withthe dominant

Proportion of tunes starting withthe tonic

Folk Songs

N=2772 N=280 N=1017

N=4842

Sorrow

The Folk Song collectioncontains the lyrics of the first verse of each song. There are a total of 74596 words, of which 13857 areunique (incl. inflectionalforms).

Frequencies of wordsrelated to some interestingthemes are portrayed here.

Drinking

Love

Joy

C C# D D# E F F# G G# A A# B0

0.05

0.1

0.15

0.2

0.25

Sävelluokka

Osu

us

C C# D D# E F F# G G# A A# B0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

Sävelluokka

Osu

us

-P8 -M6 -d5 -m3 P1 +m3 +d5 +M6 +P80

0.05

0.1

0.15

0.2

0.25

Intervalli

Osu

us

-P8 -M6 -d5 -m3 P1 +m3 +d5 +M6 +P80

0.05

0.1

0.15

0.2

0.25

0.3

0.35

Intervalli

Osu

us

Folk Songs

Folk Songs

Runo-tunes

Runo-tunes

A comparison of musical features of Folk Songs and Runo-tunes

Interval distributions

The interval distributions for both subcollections are typical for Folk music; small intervals are predominant and descending scalar motion more frequent than ascending.

In Folk Songs tone repetition is more common than in Runo-tunes.

Large intervals occur less frequently in Runo-tunes than in Folk Songs.

Kantele artist Teppana Jänis. Photo: A. O. Väisänen (Kantele and Jouhikko Tunes, XXVI, © SKS, 1928).

�������������������� ���� �������������������� ������������� ���� �Application of Musical Data Mining

Proportion of tunes in minor key

Geographical variationof various musicalfeatures in Folk Songs(N=4842).

The colorbar on the right of each pictureshows the range of variation.

© 2003 Petri Toiviainen & Tuomas Eerola

The collection is musicallycomprehensive, nationallyimportant and welldocumented.

Finl

and

Rus

sia

Sw

eden

Estonia

Runo-tunes Kantele and Jouhikko Tunes

Spiritual Folk Tunes

Pro

por

tion

Pro

por

tion

Interval

Interval

Pro

por

tion

Pro

por

tion

Pitch-class

Pitch-class