Mirror of https://musescore.com/openscore-lieder-corpus
Collection of songs by 19th century composers in MuseScore format with associated data.
Scores can be downloaded individually in PDF, MIDI, MusicXML, MP3 and other formats from their official pages on MuseScore.com. Alternatively, scores can be converted to other formats en masse using MuseScore's free desktop software using either the Batch Convert Plugin or the command line interface as explained below.
To convert the corpus using the command line interface, you need a json file with in / out path pairs.
We provide such a file here with the latest contents of the corpus, ready for converting the scores as given (mscx files) to mxl.
Implement the batch conversion with the current
corpus_conversion.json
from this folder with the command:
mscore -j corpus_conversion.json
We aim to keep this contents file up to date so you should not need to adjust it for simple conversions (of whole corpus from mscx to mxl). That said, we also provide the means to do so if necessary: see this script.
To create a new / updated version of the corpus_conversion.json
, run python3 corpus_conversion.py
from this folder.
For conversion to another file format, before the relevant step/s above,
replace .mxl
with the desired format (.pdf
or .mid
) either directly for every instance in the corpus_conversion.json
file or
in the corpus_conversion.py script (as the out_format
in the prep_conversion_doc
function).
Score and lyric files are arranged in the following directory structure:
<composer>/<set>/<song>
Directories:
<composer>
- composer's name in the form Last,_First_Second...
.<set>
- name of the extended work that the song belongs to, if any.
_
(i.e. a single underscore)<song>
- name of the song, including a possible prefix.
Score files within each song directory are named as follows
lc<id>.mscx
Filename components:
lc
- "Lieder Corpus"<id>
- the score's unique Musescore ID
https://musescore.com/openscore-lieder-corpus/scores/<id>
https://musescore.com/score/<id>
.mscx
- the file extension for MuseScore's uncompressed score format.Lyric files follow the same naming format, with the extension .txt
:
i.e. lc<id>.txt
.
These files provide the lyrics (a.k.a. poem, text) for each song separately to assist with cross-referencing and text-based studies. The lyrics are as automatically extracted from each relevant part of every score. For a typical song, this means exactly one line of text.
There are some exceptions.
For the occasional part-songs in the collection (with 2 or more vocal parts) piece,
there will be one line for each part.
In these cases, the text is often identical or nearly so
(see, for instance, Coleridge-Taylor,_Samuel/_/Viking_Song/
).
Ossia parts complicate this somewhat: they are included if they contain
lyrics, again on a separate line (first) line (above the 'real' text).
For instance, in Gabriel,_Virginia/_/Only/
, the text file is:
so I must strive to for mine for ever and aye.
1. Only a face at the window, Only a face, nothing more; Yet the look in the eyes, as they meet mine, Still comes to me o’er to and o’er Only a word, man, of greeting, Only a word that was all; Yet all day, in my heart it conechoed, Like the sound of an anman gel’s call.`
Here the first line comes from an extra stave for a third verse alternative part, and the second provides the main text (first verse only). Automatic processes seeking exactly one line from each lyric files should use the lowest line.
Finally, in just one case is there are 'song' with no lyrics:
Kralik,_Mathilde/Jugend-Lieder/00_Einleitung/
.
This is a piano introduction to a song cycle.
For consistency, we include an empty .txt
file.
See issue #1 for known challenges.
We also provide automatic harmonic analyses for each score in the
human- and computer-readable "Roman text" standard.
Files are named analysis_automatic.rntxt
.
For more information and more analyses in this format,
see the When in Rome meta-corpus repository.
With the exception of a few unsafe or illegal characters, names of songs, sets and composers have been left in their original forms.
Modern filesystems should have no problems with Unicode characters in
file paths. If the paths are displayed incorrectly by git
, try setting:
git config core.quotePath false
Users on macOS may also need to set:
git config core.precomposeunicode true
Tip: Add --global
after config
in the above commands to make git
behave this way by default for all repositories on your local machine.
The Data/
directory contains the following:
corpus_conversion.json
file.Summative plots of the corpus contents:
The number of corpus composers alive and active over time.
The number of songs by the top 15 most represented composers.
The number of songs per 'set' (a.k.a. 'collection', 'cycle').
These scores are released under Creative Commons Zero (CC0). See LICENSE.txt.
We kindly ask that you credit OpenScore Lieder and provide a link to https://musescore.com/openscore-lieder-corpus or this repository for any public-facing use of these scores.
For academic publications, please cite the report on this mirror we published in M.E.C.:
@inproceedings{GothamJonas2022,
abstract = {The OpenScore Lieder Corpus is a collection of over 1,200 nineteenth century songs encoded by a dedicated team of mostly volunteers over several years. Having reported on the initial phase, motivations, design, and community-oriented aspects of the project before, we present here the first, stable, large-scale release of this corpus specifically designed for MIR researchers, complete with comprehensive, structured, linked metadata. The corpus continues to be available under the open CC0 licence and represents a compelling dataset for a range of MIR tasks, not least given its unusual balance of large-scale with high-quality encoding, and of diversity (songs by over 100 composers, from many countries, and in a range of languages) with unity (centred on the nineteenth-century lieder tradition).},
author = {Gotham, Mark Robert Haigh and Jonas, Peter},
title = {{The OpenScore Lieder Corpus}},
keywords = {mec-proceedings, mec-proceedings-2021},
pages = {131--136},
publisher = {{Humanities Commons}},
isbn = {978-84-1302-173-7},
editor = {M{\"u}nnich, Stefan and Rizo, David},
booktitle = {{Music Encoding Conference Proceedings 2021}},
year = {2022},
doi = {10.17613/1my2-dm23},
bibbase_note = {<span style="color: green; font-weight: bold">Best Poster Award.</span>},
displayby = {Contributions from MEC 2021}
}
For some use / interest cases, you may also wish to check out either or both of: our initial 2018 report on the project, and a 2021 discussion:
These scores were transcribed by contributors to OpenScore Lieder, many of them after vocal line templates by Leigh Van Handel et al., and moderated by a professional team of proofreaders.
For more about this corpus, its motivations, how to contribute, who has contributed, and the funding that has supported this social initiative, please see https://fourscoreandmore.org/scores-of-scores/