Examples

Apply a filter

Keep only the vocal songs (i.e., drop instrumental pieces):

from MTCFeatures import MTCFeatureLoader
fl = MTCFeatureLoader('MTC-FS-INST-2.0')
seq_iter = fl.applyFilter('vocal')

for seq in seq_iter:
        dosomething(seq)

If a filter has arguments, these should be provided with the filter name as tuple:

seq_iter = fl.applyFilter( ('afteryear', 1950) )

seq_iter = fl.applyFilter( ('betweenyears', 1850, 1900) )

Keep only songs in tune families with more than 10 members:

seq_iter = fl.minClassSizeFilter('tunefamily', 10)

A filter can be inverted by setting argument invert to True:

seq_iter = fl.applyFilter( ('afteryear', 1950), invert=True )

A chain of filters can be applied with the applyFilters method. The filters will be applied in the order provided.

seq_iter = fl.applyFilters(
    [
        {'mfilter':'vocal'},
        {'mfilter':'freemeter', 'invert':True},
        {'mfilter':('afteryear',1850)}
    ]
)

Register a filter

To register a new filter, use method registerMelodyFilter in the MTCFeatureLoader class. The filter should be a function returning True if the melody should be kept.

fl.registerFilter('vocal', lambda x: x['type'] == 'vocal')

Register a filter with arguments:

fl.registerFilter('afteryear', lambda y: lambda x: x['year'] > y )

Various Examples

Only use the midipitch from all songs in MTC-ANN-2.0.1:

from MTCFeatures import MTCFeatureLoader
fl = MTCFeatureLoader('MTC-ANN-2.0.1')
seq_iter = fl.selectFeatures(['midipitch'])

Use midipitch and duration from all songs in MTC-ANN-2.0.1:

fl = MTCFeatureLoader('MTC-ANN-2.0.1')
seq_iter = fl.selectFeatures(['midipitch', 'duration'])

Use intervals and inter onset interval ratios from all songs in MTC-ANN-2.0.1 and get rid of the None values for the first note:

fl = MTCFeatureLoader('MTC-ANN-2.0.1')
seq_iter = fl.selectFeatures(['chromaticinterval', 'IOR'])
seq_iter = fl.replaceNone(seq_iter=seq_iter)

Use scale degree, metric contour and beat position from all songs in MTC-ANN-2.0.1:

fl = MTCFeatureLoader('MTC-ANN-2.0.1')
seq_iter = fl.applyFeatureExtractor('full_beat_str')
seq_iter = fl.selectFeatures(
        ['scaledegree','metriccontour','full_beat_str'],
        seq_iter=seq_iter
)

Get backgroundcorpus for MTC-ANN from MTC-FS-INST:

fl = MTCFeatureLoader('MTC-FS-INST-2.0')
seq_iter = fl.applyFilter('ann_bgcorpus')

Get labeled songs in Onder de groene linde:

fl = MTCFeatureLoader('MTC-FS-INST-2.0')
seq_iter = fl.applyFilters(
    [
        {'mfilter':'inOGL'},
        {'mfilter':'labeled'}
    ]
)

Keep only those in tune families with more than 2 melodies:

seq_iter = fl.minClassSizeFilter('tunefamily', 2, seq_iter=seq_iter)

Use labeled 17th and 18th century fiddle music only:

fl = MTCFeatureLoader('MTC-FS-INST-2.0')
sel_instr = fl.applyFilter('instrumental')
sel_17th18th_c = fl.applyFilter(
        ('betweenyears', 1600, 1800),
        seq_iter=sel_instr
)
sel_labeled = fl.applyFilter('labeled', seq_iter=sel_17th18th_c)

or:

seq_iter = fl.applyFilters(
    [
        {'mfilter':'instrumental'},
        {'mfilter':'labeled'},
        {'mfilter':('betweenyears', 1600, 1800)}
    ]
)

Use big tune families (>=20 melodies):

fl = MTCFeatureLoader('MTC-FS-INST-2.0')
sel_big = fl.minClassSizeFilter('tunefamily', 20)

Use small tune families (<=5 melodies) only:

fl = MTCFeatureLoader('MTC-FS-INST-2.0')
sel_small = fl.maxClassSizeFilter('tunefamily', 5)

Use only melodies with given identifiers:

fl = MTCFeatureLoader('MTC-FS-INST-2.0')
id_list = [
        'NLB125814_01',
        'NLB125815_01',
        'NLB125817_01',
        'NLB125818_01',
        'NLB125822_01',
        'NLB125823_01'
]
sel_list = fl.applyFilter( ('inNLBIDs', id_list) )

Use only instrumental melodies from tune family 2805_0:

fl = MTCFeatureLoader('MTC-FS-INST-2.0')
tf_list = ['2805_0']
sel_list = fl.applyFilter(
        ('inTuneFamilies', tf_list),
        seq_iter=fl.applyFilter('instrumental')
)

Write the result to a gzipped .jsonl file:

fl.writeJSON('2805_0.jsonl.gz', seq_iter=sel_list)

Only keep ESSEN melodies from The Netherlands (For the exact names of the origins, consult the directory tree of the **kern distribution of the Essen Folksong Collection.):

fl = MTCFeatureLoader('ESSEN')
sel_dut = fl.applyFilter( ('origin', 'europa/nederlan') )

Get vocal melodies that have a meter:

fl = MTCFeatureLoader('MTC-FS-INST-2.0')
seq_iter = fl.applyFilters(
    [
        {'mfilter':'vocal'},
        {'mfilter':'freemeter', 'invert':True}
    ]
)