Overview#

This notebook gives a general overview of the features included in the dataset.

Hide imports 
import os
from collections import defaultdict, Counter
from fractions import Fraction

from git import Repo
import dimcat as dc
import ms3
import pandas as pd
import plotly.express as px
import plotly.graph_objects as go

from utils import CADENCE_COLORS, CORPUS_COLOR_SCALE, STD_LAYOUT, TYPE_COLORS, color_background, corpus_mean_composition_years, value_count_df, get_corpus_display_name, get_repo_name, print_heading, resolve_dir
Hide source 
CORPUS_PATH = os.path.abspath(os.path.join('..', '..'))
ANNOTATED_ONLY = os.getenv("ANNOTATED_ONLY", "True").lower() in ('true', '1', 't')
print_heading("Notebook settings")
print(f"CORPUS_PATH: {CORPUS_PATH!r}")
print(f"ANNOTATED_ONLY: {ANNOTATED_ONLY}")
CORPUS_PATH = resolve_dir(CORPUS_PATH)

Notebook settings
-----------------

CORPUS_PATH: '/home/runner/work/workflow_deployment/medtner_tales'
ANNOTATED_ONLY: False
Hide source 
repo = Repo(CORPUS_PATH)
print_heading("Data and software versions")
print(f"Data repo '{get_repo_name(repo)}' @ {repo.commit().hexsha[:7]}")
print(f"dimcat version {dc.__version__}")
print(f"ms3 version {ms3.__version__}")

Data and software versions
--------------------------

Data repo 'medtner_tales' @ adb6f24
dimcat version 0.3.0
ms3 version 2.5.2

dataset = dc.Dataset()
dataset.load(directory=CORPUS_PATH, parse_tsv=False)

[default|all]
All corpora
-----------
View: This view is called 'default'. It 
	- excludes pieces that are not contained in the metadata,
	- filters out file extensions requiring conversion (such as .xml), and
	- excludes review files and folders.

                   has   active   scores measures           notes        expanded          chords       
              metadata     view detected detected parsed detected parsed detected parsed detected parsed
corpus                                                                                                  
medtner_tales      yes  default       19       19     19       19     19       19     19       19     19

133/437 files are excluded from this view.

133 files have been excluded based on their subdir.


There are 1 orphans that could not be attributed to any of the respective corpus's pieces.

N = 19 annotated pieces, 76 parsed dataframes.
Hide data loading 
all_metadata = dataset.data.metadata()
assert len(all_metadata) > 0, "No pieces selected for analysis."
print(f"Metadata covers {len(all_metadata)} of the {dataset.data.count_pieces()} scores.")
all_notes = dataset.get_facet('notes')
all_measures = dataset.get_facet('measures')
mean_composition_years = corpus_mean_composition_years(all_metadata)
chronological_order = mean_composition_years.index.to_list()
corpus_colors = dict(zip(chronological_order, CORPUS_COLOR_SCALE))
corpus_names = {corp: get_corpus_display_name(corp) for corp in chronological_order}
chronological_corpus_names = list(corpus_names.values())
corpus_name_colors = {corpus_names[corp]: color for corp, color in corpus_colors.items()}

Metadata covers 19 of the 19 scores.

Composition dates#

This section relies on the dataset’s metadata.

valid_composed_start = pd.to_numeric(all_metadata.composed_start, errors='coerce')
valid_composed_end = pd.to_numeric(all_metadata.composed_end, errors='coerce')
print(f"Composition dates range from {int(valid_composed_start.min())} {valid_composed_start.idxmin()} "
      f"to {int(valid_composed_end.max())} {valid_composed_end.idxmax()}.")

Composition dates range from 1904 ('medtner_tales', 'op08n01') to 1925 ('medtner_tales', 'op48n01').

Mean composition years per corpus#

Hide source 
summary = all_metadata.copy()
summary.length_qb = all_measures.groupby(level=[0,1]).act_dur.sum() * 4.0
summary = pd.concat([summary,
                     all_notes.groupby(level=[0,1]).size().rename('notes'),
                    ], axis=1)
bar_data = pd.concat([mean_composition_years.rename('year'), 
                      summary.groupby(level='corpus').size().rename('pieces')],
                     axis=1
                    ).reset_index()
fig = px.bar(bar_data, x='year', y='pieces', color='corpus',
             color_discrete_map=corpus_colors,
            )
fig.update_traces(width=5)
fig.update_layout(**STD_LAYOUT)
fig.update_yaxes(gridcolor='lightgrey')
fig.update_traces(width=5)

Composition years histogram#

Hide source 
hist_data = summary.reset_index()
hist_data.corpus = hist_data.corpus.map(corpus_names)
fig = px.histogram(hist_data, x='composed_end', color='corpus',
                   labels=dict(composed_end='decade',
                               count='pieces',
                              ),
                   color_discrete_map=corpus_name_colors,
                  )
fig.update_traces(xbins=dict(
    size=10
))
fig.update_layout(**STD_LAYOUT)
fig.update_yaxes(gridcolor='lightgrey')
fig.show()

Dimensions#

Overview#

Hide source 
corpus_metadata = summary.groupby(level=0)
n_pieces = corpus_metadata.size().rename('pieces')
absolute_numbers = dict(
    measures = corpus_metadata.last_mn.sum(),
    length = corpus_metadata.length_qb.sum(),
    notes = corpus_metadata.notes.sum(),
    labels = corpus_metadata.label_count.sum(),
)
absolute = pd.DataFrame.from_dict(absolute_numbers)
absolute = pd.concat([n_pieces, absolute], axis=1)
sum_row = pd.DataFrame(absolute.sum(), columns=['sum']).T
absolute = pd.concat([absolute, sum_row])
relative = absolute.div(n_pieces, axis=0)
complete_summary = pd.concat([absolute, relative, absolute.iloc[:1,2:].div(absolute.measures, axis=0)], axis=1, keys=['absolute', 'per piece', 'per measure'])
complete_summary = complete_summary.apply(pd.to_numeric).round(2)
complete_summary.index = complete_summary.index.map(dict(corpus_names, sum='sum'))
complete_summary
absolute per piece per measure
pieces measures length notes labels pieces measures length notes labels length notes labels
Medtner Tales 19 2464 6598.0 42929 6508 1.0 129.68 347.26 2259.42 342.53 2.68 17.42 2.64
sum 19 2464 6598.0 42929 6508 NaN NaN NaN NaN NaN NaN NaN NaN

Measures#

print(f"{len(all_measures.index)} measures over {len(all_measures.groupby(level=[0,1]))} files.")
all_measures.head()

2472 measures over 19 files.
mc mn quarterbeats duration_qb keysig timesig act_dur mc_offset numbering_offset dont_count barline breaks repeats next quarterbeats_all_endings volta markers jump_bwd jump_fwd play_until
corpus fname interval
medtner_tales op08n01 [0.0, 2.0) 1 1 0 2.0 -3 4/8 1/2 0 <NA> <NA> <NA> <NA> firstMeasure (2,) NaN <NA> NaN NaN NaN NaN
[2.0, 4.0) 2 2 2 2.0 -3 4/8 1/2 0 <NA> <NA> <NA> <NA> <NA> (3,) NaN <NA> NaN NaN NaN NaN
[4.0, 6.0) 3 3 4 2.0 -3 4/8 1/2 0 <NA> <NA> double <NA> <NA> (4,) NaN <NA> NaN NaN NaN NaN
[6.0, 8.0) 4 4 6 2.0 -3 4/8 1/2 0 <NA> <NA> <NA> <NA> <NA> (5,) NaN <NA> NaN NaN NaN NaN
[8.0, 10.0) 5 5 8 2.0 -3 4/8 1/2 0 <NA> <NA> <NA> line <NA> (6,) NaN <NA> NaN NaN NaN NaN 
print("Distribution of time signatures per XML measure (MC):")
all_measures.timesig.value_counts(dropna=False)

Distribution of time signatures per XML measure (MC):

4/4    562
2/4    552
3/8    519
3/4    425
6/8    235
4/8     81
2/2     80
2/8     17
9/8      1
Name: timesig, dtype: int64

Harmony labels#

All symbols, independent of the local key (the mode of which changes their semantics).

try:
    all_annotations = dataset.get_facet('expanded')
except Exception:
    all_annotations = pd.DataFrame()
n_annotations = len(all_annotations.index)
includes_annotations = n_annotations > 0
if includes_annotations:
    display(all_annotations.head())
    print(f"Concatenated annotation tables contains {all_annotations.shape[0]} rows.")
    no_chord = all_annotations.root.isna()
    if no_chord.sum() > 0:
        print(f"{no_chord.sum()} of them are not chords. Their values are: {all_annotations.label[no_chord].value_counts(dropna=False).to_dict()}")
    all_chords = all_annotations[~no_chord].copy()
    print(f"Dataset contains {all_chords.shape[0]} tokens and {len(all_chords.chord.unique())} types over {len(all_chords.groupby(level=[0,1]))} documents.")
    all_annotations['corpus_name'] = all_annotations.index.get_level_values(0).map(get_corpus_display_name)
    all_chords['corpus_name'] = all_chords.index.get_level_values(0).map(get_corpus_display_name)
else:
    print(f"Dataset contains no annotations.")
mc mn quarterbeats quarterbeats_all_endings duration_qb mc_onset mn_onset timesig staff voice ... phraseend chord_type globalkey_is_minor localkey_is_minor chord_tones added_tones root bass_note pedalend volta
corpus fname interval
medtner_tales op08n01 [0.0, 2.0) 1 1 0 0 2.0 0 0 4/8 2 1 ... { MM7 True True (-4, 0, -3, 1) () -4 -4 NaN <NA>
[2.0, 3.5) 2 2 2 2 1.5 0 0 4/8 2 1 ... <NA> Mm7 True True (6, -4, 0, 2) (-1,) 2 6 NaN <NA>
[3.5, 4.0) 2 2 7/2 7/2 0.5 3/8 3/8 4/8 2 1 ... <NA> Mm7 True True (2, 6, 3, 0) (-3,) 2 2 NaN <NA>
[4.0, 5.0) 3 3 4 4 1.0 0 0 4/8 2 1 ... <NA> Mm7 True True (1, 5, 2, -1) () 1 1 NaN <NA>
[5.0, 9.5) 3 3 5 5 4.5 1/4 1/4 4/8 2 1 ... } m True True (0, -3, 1) () 0 0 NaN <NA>

5 rows × 33 columns

Concatenated annotation tables contains 6502 rows.
60 of them are not chords. Their values are: {'{': 59, '|PAC}': 1}
Dataset contains 6442 tokens and 1332 types over 19 documents.