- Minimal deviations from MuseScore defaults (manual edits do not convert)
- Most systematic changes as defined in the
.mssstyle sheet:- Justify full page.
- All present instruments showing at all times (none hidden).
- Note, this applies except for in some extreme cases such as Beethoven No.9:
- Movement 2: the ATB trombones rest for 500 bars.
- Movement 4: trombones, solo voices, chorus, and percussion.
- Naturally, whether instruments are hidden from display or not can easily be adapted by the user and does not affect the retrieval.
- Note that "all present instruments" refers to the movement (i.e., "tacet" from the movement = absent from the score).
- Note, this applies except for in some extreme cases such as Beethoven No.9:
- Note: Import the
.mssstyle file in-app or with the command line:mscore <before_file_name>.mscz --style <style-file-name>.mss -o <after_file_name>.mscz.
- Every part on a separate stave (e.g., Flute 1 separate from Flute 2) for clarity and interoperability.
- Note that this is a unique feature of this corpus, making it much more usable for both part extraction and MIR tasks.
- Connect bar lines through those like instruments e.g., Flutes 1 and 2; Horns 1, 2, and 3.
- Parts and part names:
- Full part names in the format
<transposition where relevant> <instrument> <number>, e.g.,A Clarinet 2. - Abbreviated names without transposition or period character, e.g.,
Cl 2.- This deviates from common practice, but (for what it's worth) that musical practice is often wrong:
- Periods should stand for incomplete words without the final letter, so then:
Cl.forClarinetis fine- but
VlnforViolinis not.
- Periods should stand for incomplete words without the final letter, so then:
- This deviates from common practice, but (for what it's worth) that musical practice is often wrong:
- String instruments in the singular, e.g.,
Violin 1as in 'the violin 1 part'.- Transpositions as in the original, except no mid-movement transpositions:
- For example, the Beethoven symphonies do this in the horn parts (4/ii, 5/iii 4 bars before the end), and clarinets (9/iv).
- This does not convert reliably, and adding a new part creates its own problems.
- Where that is not possible in MS4, a manual workaround is required.
- Brahms 1/iii and 2/ii have horns and trumpets in B natural, which means setting the transposition manually.
- Transpositions as in the original, except no mid-movement transpositions:
- Full part names in the format
- Stave size:
- Manually set for each work to use the largest stave size that fits within the page, including the large lyric annotations while maintaining approximate consistency across multi-movement works.
- Again, if anyone uses these scores in performance, they may wish to adjust.
- More formally:
- Strong preference for not overfilling page.
- Strong preference for approximate consistency across multi-movement works.
- Medium preference for not underfilling.
- Medium preference for the same number of staves on each page and weak preference that this includes the last page.
- Copy the named (public domain) source edition exactly, except in limited cases like:
- Some variation in "crescendo" vs "cresc." vs hairpins. In future we will aim for:
- "Cresc." text with the dotted line for long-range (e.g., across a page break),
- Hairpin symbols for all short range (e.g., 1 bar),
- Some discretion for busy staves ("cresc" takes less space).
- Note that this is certainly not consistenly implemented througout at present.
- Stave and page breaks mostly follow MuseScore defaults.
- Given the stylistic choices above, we make no attempt for page breaks to follow the source.
- E.g., using Eulenberg for the Beethoven means a very different page size.
- Notable exceptions include Brahms 4/iv, which is in a completely regular 8-bar phrasing.
- We set this out with page breaks every 4 or 8 bars depending on the note density.
- Tempo markings:
- Tempos etc. exactly as in the source and properly encoded in MuseScore (mostly successful in conversion).
- Tempo text (e.g.,
Allegro) is only ever included if present in the original source. - Tempo in metronome mark is included:
- If present in the original source.
- In square brackets where clarification is needed.
- Tempo text (e.g.,
- Here, as elsewhere, avoid any "hidden" markings, except:
- 'tuplets' (e.g., Brahms 1/iv). Show the first '6' or '12' and then hide thereafter.
- ornament markings (e.g., staccato) that are missing in the source but wanted for playback.
- Tempos etc. exactly as in the source and properly encoded in MuseScore (mostly successful in conversion).
- Given the stylistic choices above, we make no attempt for page breaks to follow the source.
- Some variation in "crescendo" vs "cresc." vs hairpins. In future we will aim for:
- Non-tempo textual markings:
- Haupstimme annotations usually as lyrics, as described in this explanation doc.
Stave textorPlaying techniquefor instrument-specific markings likearco.- Note that original scores use annotations like
a2where there are two parts on one stave.- We have split those parts, so don't strictly need these annotations.
- All the same, we include them where it might plausibly help clarify solos and the moves to/from unison.
- Note that original scores use annotations like
System textfor non-tempo comments that apply to all parts e.g., in Beethoven 9/ii,ritmo de ....
- Bar (measure) counting.
- This is a perennially complex issue, and one often broken by export change of format.
- Please see our recent paper about this topic with a proposed solution here.
- For now, on MuseScore files we leave the default measure count (1, 2, 3, ...):
- MuseScore offers some functionality to 'exclude bars from the measure count'. This can lead to improved layout in the app, but experimentation shows that it does not convert well.
- We therefore avoid such manual changes and (attempt) to fix them automatically from the measure map.
- Example issues: anacruses, first/second time bars, split bars.
- MuseScore offers some functionality to 'exclude bars from the measure count'. This can lead to improved layout in the app, but experimentation shows that it does not convert well.